Datasets:

medarc

/

clinical_pile_v1_minhash_deduped

like 1

direct_quantification_of_natural_moisturizing_factors_in_stratum_corneum_using_direct_analysis_in_re

## Abstract: in some diseases, such as ichthyosis and atopic dermatitis 43,44 . Therefore, the mechanisms of NMF production and the relevant enzymes associated with them have become a prominent area of focus and study. In measuring NMF, SC samples are collected simply by tape-stripping and are quantified using liquid chromatography-mass spectrometry (LC-MS) or gas chromatography-mass spectrometry (GC-MS). Tape-stripping is a technique to take a SC sample using adhesive tape. The procedure is as follows. A tape is firmly adhered to the skin, pressed with the fingers over the entire area covered by the tape, and then removed from the skin. While a large quantity of sample skin is obtainable thanks to this simple method, the chromatographic techniques require lengthy analyses and cannot meet current levels of demand. In order to overcome this issue, we used DART coupled with time-of-flight (TOF) MS and were able to quantify 12 NMF in the previous report 31 .However, there were still 2 issues to be tackled. One was that only 12 NMF among 26 were quantified because some components were too few to detect in the samples. In addition, leucine and isoleucine were not separable without chromatographic techniques. The second was the lower accuracy resulting from the addition of SIL-IS. Ambient mass spectrometry methods, such as DART-MS, are more susceptible to ion suppression/ enhancement 45,46 , and the tissue-specific ion suppression effect still provides challenges for quantitative analysis 47 . Although SIL-IS are thought to be essential for quantitative analysis, they are difficult to coat evenly as the SC samples are solid.We employed a triple quadrupole (TQ) MS suitable for quantitative analysis, owing to its high sensitivity using multiple reaction monitoring (MRM). NMF that are undetectable using TOF MS can possibly be detected using TQ. In addition, more selective identification can be achieved using the product ions formed via the collisions. For example, leucine and isoleucine have the same molecular composition. Thus, they cannot be distinguished without chromatographic separation. However, the product ions induced by collision-induced dissociation are different. They can be separated in an identical fashion using TQ. Furthermore, an inkjet printing technique was used to realize homogeneous deposition and coating of the SC samples to improve the accuracy. This technique has been already utilized in the field of imaging MS (IMS) for matrix additions 48 , calibration standards 49 , and SIL-IS 50 . However, our inkjet technique is unique and superior in terms of both the precision and volume of ejected droplets, allowing for highly customized coating patterns.Described below are the steps to be taken to establish the quantification method for NMF in the SC using the inkjet printing technique and TQMS. Introduced is a novel procedure for high-throughput and quantitative analysis using DART-MS. ## Direct Quantification of Natural Moisturizing Factors in Stratum Corneum using Direct Analysis in Real Time Mass Spectrometry with Inkjet-Printing Technique Katsuyuki Maeno Proper hydration of the stratum corneum, the skin's outermost layer, is essential for healthy skin. Water-soluble substances called natural moisturizing factors (NMF) are responsible for maintaining adequate moisture in the skin and are closely associated with a variety of the skin's functions. Therefore, quantitative analysis methods for NMF are indispensable when attempting to clarify one of the mechanisms of hydration and its effect on the skin. This study sought to develop a quick and simple analytical technique, which can quantify NMF from the skin without the need for extraction or separation, using direct analysis in real time-mass spectrometry (DART-MS). The goal was to deliver a high quantitative capability, so a unique inkjet printing technique was employed to evenly coat a stable isotope-labeled internal standard (SIL-IS) on tape-stripped skin. This technique allowed for the quantification of 26 NMF with established calibration curves and comparatively high linear correlations. The speed of measurement was found to be advantageous as 100 strips of tape can be measured in roughly 2 hours. The effectiveness of the inkjet coating was also verified by comparing its precision with that of conventional pipetting. This new technique can be an alternative method to quantify NMF rapidly and perhaps allow for a clearer elucidation of their function in skin. Ambient mass spectrometry (MS), such as direct analysis in real time (DART) 1,2 , desorption electro spray ionization (DESI) 3,4 , low temperature plasma (LTP) 5 , and paperspray 6 , is an ambient ionization method that reduces the time, equipment, and expertise needed for sample preparation and chromatography separation. There are portable applications with ambient MS because measurement procedures are simple and quick to execute . Solid or liquid samples can be measured simply by placing them in the gap between the ion source and the inlet of the mass spectrometer. DART-MS is considered an effective ambient plasma ionization method that can dramatically reduce analysis time for the routine screening of samples. Therefore, DART-MS has been successfully applied in a variety of fields such as foods 11 , food packaging 12 , forensic analysis 13 , additives in plastics 14 , contaminants in soil 15 , pesticides 16 , metabolites 17,18 , drugs 19 , nucleotides 20 , and mycotoxins 21 . The DART ion source can supply a stream of electrically discharged gas (typically excited He or N 2 ) and samples are hit directly by those gases or indirectly via other reactions with water molecules to form the ionized components that lead to the MS 1,2 . Since the time of the DART method's development, a lot of qualitative applications have arisen by combining DART with high resolution MS and ion mobility 25 . While quantitative applications were initially few, they are now increasing through the use of stable-isotope-labeled (SIL) internal standards (SIL-IS) and other techniques 22, . We developed a method to quantitatively analyze natural moisturizing factors (NMF) using DART-MS and SIL-IS in our previous study 31 . NMF, such as urea, pyrrolidone carboxylic acid, lactic acid, urocanic acid, and various amino acids, are water-soluble compounds with low molecular weights that exist in the epidermis 32,33 . The epidermis, the upper layer of skin, has the ability to produce NMF and they are responsible for maintaining adequate hydration of the stratum corneum (SC), the outermost part of the epidermis . Proper hydration is essential for healthy skin in terms of elasticity, enzyme activity, and barrier function . In addition, NMF are involved ## Results and Discussion Workflow of the NMF quantification. The goal of this study was to establish and optimize a procedure for NMF quantification in the SC which is faster and simpler than conventional techniques and which includes all the steps needed from sample collection to data analysis. The established workflow is shown in Fig. 1. First, SC samples were collected from various parts of the body from volunteers with tape stripping using D-squame tape (Cuderm Corporation, Dallas, TX, USA) with a diameter of 22 mm (Fig. 1a). The tape was originally slitted so that the tapes could easily be cut into rectangular strips for DART-MS measurement later. The tape stripping was conducted several times on the same area of skin to examine the depth profile of NMF. Then, the amount of total proteins in the SC adhered to the tape was measured using a SquameScan 850 A instrument (Heiland electronic, Wetzlar, Germany, Fig. 1b) 51 . Since the amount of SC taken from the skin varies, it is not possible to attribute the amount of NMF to the quality of SC or the amount of SC. Knowing the weight of SC removed is beneficial to distinguish those 2 factors, and we used the amount of total proteins as a key indicator in this study. Since the amount of total proteins measured is considered to be proportional to the amount of SC removed, that value was used to normalize the amount of NMF measured using DART-MS 21 . After that, the tape was placed in the inkjet device and coated with a SIL-IS mixture solution (Fig. 1c). A square shape area with one side of 11 mm (a green square area in the tape in Fig. 1c) was coated so that the area used for the DART-MS measurement could be included. One nL of the SIL-IS mixture solution was ejected from the inkjet head and dropped on the surface of the tape containing the SC. The droplets were designed to lie in a laterally and longitudinally aligned grid pattern of spots with a distance of 220 μm between each droplet (Fig. S1). Each droplet quickly evaporated and dried. The final concentration of the coated SIL-IS was 1 nmol/cm 2 on the tape for each SIL-IS. 81 tapes (9 × 9) can be accommodated on the stage part of this device and coated within 1 hour at one time. Then, the tape was cut into 2 rectangular strips (20 × 2 mm) (Fig. 1d). The strip was attached to the top of a quartz prism (Fujiwara, Tokyo, Japan) with the adhesive side facing up. 10 strips were attached onto the same prism at intervals of about 10 mm between each strip. The prism was placed between the DART ion source and the MS detector so that the excited and heated helium gas from DART hit the surface of the tape strips on the prism. The helium gas reacts with atmospheric water molecules to produce ionized water clusters [(H 2 O) n + H] + . These protonated water clusters can then react with the NMF and SIL-IS to form protonated cations and then flowed into the ceramic tube inlet for MS analysis (Fig. 1e). Heated helium is needed to evaporate NMF in the SC and SIL-IS coated on the SC adhere to the strips before ionization. The prism moved automatically at a speed of 0.2 mm/s in a direction perpendicular to the flow of helium gas. We chose a distance of 10 mm between 2 tapes and set the speed of the prism at 0.2 mm/s so that the signals of all the NMF and SIL-IS were consumed before the peak of the next tape appeared. The area hit by the helium gas was approximately 0.1 cm 2 (10 × 2 mm), which was measured using heat-sensing tapes (NiGK, Saitama, Japan) placed on the quartz prism instead of the D-squame tape. Fig. S2 shows a typical total ion current chromatogram for one prism. MRM chromatograms measured with optimized MRM methods for each NMF were produced, and the peak area was used to determine the amount of NMF. 10 tapes are measured at one time within about 10 minutes. MRM method selection for NMF and SIL-IS. A number of recent studies have shown that DART forms abundant (de)protonated analytes [M ± H] ± via a proton transfer from background ions such as H 3 O + (H 2 O) n and O 2 to the analytes M with relatively low internal energy 29,52 . That is also applicable to NMF. Another group has confirmed that almost all amino acids produce (de)protonated molecules at the dominant ion peaks 52 . Our study also obtained the same result. Therefore, (de)protonated molecules of NMF and SIL-IS were used as precursor ions for the MRM conditions. However, specific MRM methods for NMF and SIL-IS must be applied because chromatographic separation is not possible and they are only distinguishable using MS. It is probable that a method optimized for a specific NMF or SIL-IS will detect others because most NMF and SIL-IS have a carboxylic group and an amino group in common and their molecular composition or structures, such as aspartic acid and asparagine, glutamic acid and glutamine, are similar to each other. We discovered about 10 optimized MRM conditions for each NMF and SIL-IS through an MRM optimization tool installed in LCMS8040 and confirmed the selectivity of each MRM method as follows. A solution containing 100 ppm of each of the NMF and SIL-IS was measured under the MRM conditions optimized for each NMF and SIL-IS by placing a capillary tube with a small amount of each solution between the DART ion source and the MS. We managed to find specific MRM conditions for almost all NMF and SIL-IS except for leucine and isoleucine, as shown in Table 1. Leucine and isoleucine are structural isomers that share the same chemical formula. Therefore, both molecules produce similar product ions and are difficult to distinguish without chromatographic separation. Fig. S3 shows the specificity result of leucine and isoleucine with the best MRM methods optimized for each of the 2. The MRM method optimized for isoleucine (m/z 132.1 > m/z 57.1) detects isoleucine by the product ion m/z 57.1 originating from the branched structure. However, it also detects leucine with a relatively small peak compared to isoleucine. Leucine and isoleucine are considered to be present in SC in the same representative amounts, as shown in the GC-MS result, so we decided to use this method for isoleucine. On the other hand, the MRM method optimized for leucine (m/z 132.1 > m/z 43.1) detects only leucine by the product ion m/z 43.1. In addition, some NMF, such as cysteine, glycine, lactic acid, urea, alanine, do not give enough product ions to be detectable. Therefore, selected ion monitoring (SIM) for (de)protonated ions was used instead of MRM. In this study, we employed some SIL-IS with only one isotope label because of the low price. However, a mass of a naturally occurring isotope in the NMF to be measured can overlap the mass of its SIL-IS. Therefore, theoretical isotope ratios of NMF were calculated and the amounts of NMF affecting the amounts of SIL-IS were taken into consideration to make the calibration curves. Calibration curves for NMF. A summary of the calibration curves of the 26 NMF is shown in Table 1. A validation process was performed by determining the linear range, precision, and limit of quantification (LOQ) based on the procedure defined by the Food and Drug Administration (FDA). Blank tapes coated with the mixture solutions of NMF and SIL-IS by LaboJet were used to determine the validations. Matrix-specific validation is often desired owing to the presence of different interfering components. Due to the presence of unknown amounts of endogenous NMF in this study, the SC cannot be used directly as a blank. Furthermore, the SC is a heterogeneous solid sample and cannot be separated into several even portions. Therefore, different approaches, www.nature.com/scientificreports www.nature.com/scientificreports/ such as surrogate analytes and standard additions, could not be employed. The slopes, intercepts, and correlation coefficients are shown as characteristic parameters of linearity in the range from 0.05 to 25 nmol/cm 2 . All the NMF were detectable (S/N > 3) at 0.05 nmol/cm 2 , except for lactic acid, urea, arginine, glycine, cysteine, and glutamic acid, which were detectable at 0.25-0.5 nmol/cm 2 . The low detectability can be ascribed to their low volatility or SIM methods. Good linear correlations for all of the NMF in a given concentration range were obtained between the peak area ratio of the NMF to IS and the amount of NMF applied to each piece of tape. Repeatability (intraday) was also assessed by measuring the tape 8 times. The relative standard deviation (RSD) was calculated for all concentration levels as an indicator of the intra-assay precision and only the maximum RSD values among www.nature.com/scientificreports www.nature.com/scientificreports/ all of the concentrations are shown. The data indicates relatively good intraday precision. The LOQ was determined based on a signal-to-noise ratio (S/N) of 10, where the signal is the peak intensity of each NMF extracted from the chromatogram. The amount of NMF depends on the place of the body and also the amount of SC taken. Looking at the SC in the forearm and cheek, the amounts of arginine and lactic acid present were lower than their LOQ, respectively, so these must be improved in the future. ## Comparison of addition techniques of the SIL-IS mixture solution (LaboJet vs Micropipetto). To demonstrate how effective the LaboJet coating is, we compared the IS addition by Labojet with the Micropipetto used in the previous study 31 . D-squame tapes coated with a concentration of 1 nmol/cm 2 on the tape for each NMF were prepared in the same way as in Fig. 2. Then, the SIL-IS mixture solution was added onto the tape using 2 different coating methods, LaboJet and Micropipetto. For the LaboJet addition, the SIL-IS mixture solution was applied onto the tape in the same way as in Fig. 2. For the Micropipetto addition, 1 μL of 100 nmol/ mL of the SIL-IS mixture solution was applied within the central 2 × 5 mm area (0.1 cm 2 ) of the tape, which is the area He gas hit, as shown in the previous report 31 . So, the final amounts of SIL-IS added by both methods on the tapes were the same, 1 nmol/cm 2 . After that, both tapes were set on the prism and measured by DART-MS in the same way as described in Fig. 1 6 times, respectively. %RSD for all NMF measured with the LaboJet addition www.nature.com/scientificreports www.nature.com/scientificreports/ technique were smaller than those with the Micropipetto addition (Fig. 3, Table S1). He gas hits an area of about 0.1 cm 2 (2 mm × 5 mm) on the tape. For the addition by Micropipetto, it is possible that the added SIL-IS mixture solution is not evenly coated and the quantified values are likely to be higher than the actual values or fluctuate as shown in Fig. 3. In addition, the area of tape He gas hit is not necessarily constant. The DART ion source (He gas), the top of the prism, and the MS detector are designed to lie on a straight line. If the position of the DART is higher, even by a few mm, than the optimized position, the He gas passes over the tape on the prism and results in an area of tape smaller than 0.1 cm 2 being hit by the gas. On the contrary, if the position of DART is lower than the optimized position, the He gas hits a lower position of the tape and the area getting hit becomes larger than 0.1 cm 2 . In both cases, the ratio of the amount of NMF to the amount of SIL-IS could vary because SIL-IS is not coated evenly on the tape and exists at random at certain positions in the area the He gas is hitting. Therefore, the fluctuation of the height of DART leads to a larger %RSD as well as higher quantified values. Lysine, tyrosine, and serine showed much larger %RSD. The values of the area of NMF divided by the area of SIL-IS were higher than those of other NMF. Theoretically, the ratio of the area of NMF to the area of SIL-IS should be one. That has still not been elucidated and is under investigation. However, our hypothesis is that the physical characteristics of NMF have something to do with the result (isoelectric point, volatility, polarity, etc.). On the other hand, the ratio of the amount (area) of SC to the amount of SIL-IS measured by LaboJet is always the same regardless of the position that the He gas hits. This result proved that Labojet coating is essential for a precise, direct quantification of solid samples. ## Comparison of quantitative NMF values obtained by DART-MS and GC-MS. A standard addi- tion technique is commonly used to overcome any matrix interferences occurring between the target molecules and matrix in a sample. However, this technique involves adding known amounts of standard to one or more aliquots of the processed samples. It cannot be applied to solid samples, such as the SC because the SC is a heterogeneous sample and identical samples cannot be prepared. Therefore, we employed another analytical method for the NMF quantification to verify our technique. The amounts of NMF measured by DART-MS were compared with those measured by a conventional GC-MS method. This GC-MS method has been previously validated, but it is designed to measure only amino acids . Therefore, the amounts of amino acids obtained by the 2 methods were compared. A tape-stripped SC sample was cut in 2, with one half measured by DART-MS and the other by GC-MS, as described in Materials and Methods. These 2 areas were next to each other. Therefore, we assumed that these 2 areas contained the same levels of amino acids. Figure 4 shows the amounts of amino acids with profiling obtained by DART-MS and GC-MS. We measured 12 samples in total (the 2 nd , 3 rd , and 4 th tapes from 4 volunteers) with each method. Some amino acids showed good agreement in the values between DART-MS and GC-MS. However, the ratio of the recovery of amino acids obtained by DART-MS to the one obtained by GC-MS was not consistent. Most of the amino acids showed that the amounts obtained by DART-MS were 2 to 5 times larger than those obtained by GC-MS. One possible explanation for this difference is that amino acids are extracted from a liquid for the GC-MS, while they are simultaneously evaporated and ionized by heated and excited He gas for the DART-MS. It is possible that the latter method has a higher extraction efficiency. Another explanation for the result is that a difference in the states of NMF and SIL-IS in the SC solid sample affects the recovery obtained by DART-MS because the NMF are originally present in the SC while the SIL-IS are added on the SC and penetrate into the SC. Therefore, the positions at which NMF and SIL-IS exist cannot be the same. Basically, NMF should be in the same depth and state as SIL-IS. Further study is necessary to elucidate the difference of the recovery. An example of NMF compositional profile in the SC taken from the cheek using DART-MS is shown in Fig. 5a. All values are shown in Table S2. Among the 26 NMF, the values of arginine, cysteine, and lactic acid were lower than the LOQ and were undetectable. Therefore, the remaining 23 NMF were quantified using the calibration curves described above. The NMF compositional profile was not so different from the one which was already reported 56 . Serine, pyroglutamic acid, citrulline, glycine, alanine and urea were present in the SC as major components. An NMF depth profile in the SC was also examined (Fig. 5b). Most of the NMF showed an increasing trend with the depth of the SC. In contrast, urea, which is the NMF originating from sweat, showed a decreasing trend. Although these trends have already been shown by other studies, our DART-MS technique provided further confirmation of these trends 35,56 . ## Conclusions A novel analytical method for the rapid and easy quantification of NMF in the SC was successfully achieved by combining the DART-MS technique and SIL-IS coating technique using LaboJet. DART-MS is a rapid and easy-to-use analytical technique. However, quantification precision is a disadvantage of DART-MS. We improved that by employing a new SIL-IS coating technique with LaboJet and proved that SIL-IS, coated evenly in a reticular pattern with nanoliter droplets, generated low %RSD of quantitative values and could be used to enhance the precision of the DART-MS quantification technique. Good linear correlations were also obtained for the calibration curves of 26 NMF with this coating technique. The developed DART-MS technique does not require pretreatment, such as liquid extraction and chromatographic separation. This result means a reduced analysis time compared to conventional techniques, with only 2 hours being required to measure 100 samples. To the best of our knowledge, other than this study, there is no report on direct quantification of NMF in the SC. At present, we are aiming to advance this DART-MS technique to on-site analysis by shortening the SIL-IS coating time. Furthermore, DART-MS is not limited to NMF analysis. It can detect other components, such as lipids, in the SC. ## Materials and preparation. All of the amino acids (alanine, arginine, asparagine, aspartic acid, cysteine, glutamic acid, glutamine, glycine, proline, serine, tyrosine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, valine, ornithine, and citrulline), pyroglutamic acid, urocanic acid, urea, and lactic acid were purchased from Wako (Osaka, Japan). All of the SIL-IS were purchased from JUNSEI (Tokyo, Japan). Deionized water was produced in-house using a Milli-Q gradient water purification system (Merck Millipore, Darmstadt, Germany). The SIL-IS mixture solution containing all of the SIL compounds was prepared in deionized water with each having a final concentration of 0.5 μmol/mL. The NMF mixture solutions containing all of the amino acids and pyroglutamic acid, urocanic acid, urea, lactic acid were prepared in deionized water with a final concentration ranging between 0.025-12.5 μmol/mL for each compound. ## DART-MS. DART-MS measurements were performed using a DART ion source (IonSence, Saugus, MA, USA) and a triple quadruple mass spectrometer LCMS8040 (Shimadzu, Kyoto, Japan). The ion source was operated in positive ion mode at 500 °C. Helium was used as the ionizing gas at a flow rate of 3 L/min. A MS ion source was not used (Ionization voltage, off; Nebulizing gas, off; Drying gas, off). The desolvation line temperature was set at 250 °C. All spectra were acquired in a mass range of m/z 50-600. The positive/negative switching mode was used in correspondence to the components. A dwell time of all of the MRM and SIM methods for NMF and SIL-IS was 10 ms and a loop time, which is the time between data points, was 0.8 s. One peak consists of about 12 data points on average, although peak width varies depending on the molecule. LaboJet. The SIL-IS mixture solution and the NMF mixture solution were printed on the SC samples evenly using LaboJet2000 (MicroJet, Nagano, Japan). It has a piezoelectric system that can dispense a droplet ranging from 1 pL to 1 nL precisely. The volume of the droplet can be varied by changing the electric voltage applied to the piezoelectric dispenser. Operating conditions for the piezoelectric dispenser varied slightly day-to-day depending especially on atmospheric pressure and temperature. To demonstrate the homogeneity and stability, and therefore resulting high accuracy, of the amount ejected with LaboJet, masses of 100,000 droplets of water were determined by weighing them with an analytical microbalance each time before the coating; the mass of 100,000 droplets of water is 100 mg. We always made sure that the value was within a 1% error margin. This inkjet device not only dispenses a precise amount of fluid but places the droplet on an accurately aimed location. Therefore, droplets

chemsum

{"title": "Direct Quantification of Natural Moisturizing Factors in Stratum Corneum using Direct Analysis in Real Time Mass Spectrometry with Inkjet-Printing Technique", "journal": "Scientific Reports - Nature"}

geminal_bis-borane_formation_by_borane_lewis_acid_induced_cyclopropyl_rearrangement_and_its_frustrat

## Abstract: Cyclopropylacetylene reacts with two molar equivalents of Piers' borane [HB(C 6 F 5 ) 2 ] under mild conditions by an addition/rearrangement sequence with cyclopropyl ring opening to give a mixture of two a-B(C 6 F 5 ) 2 substituted tetrahydroboroles. This compound forms an active frustrated Lewis pair with P t Bu 3 that heterolytically splits dihydrogen and adds carbon dioxide as a geminal chelate bis-boryl component. The respective reactions of the two-fold HB(C 6 F 5 ) 2 addition to Ph-CH 2 CH 2 C^CH were studied as a geminal Lewis acid reference. Most of the products were characterized by X-ray diffraction. ## Introduction Bis-boranes featuring pairs of strongly Lewis acidic B(C 6 F 5 ) 2 groups should be ideally matching templates for binding of CO 2 under frustrated Lewis pair (FLP) conditions. Although such geminal bis-boranes are principally readily available from terminal alkynes by sequential hydroboration reactions with two molar equivalents of HB(C 6 F 5 ) 2 , as it has been shown by Piers et al., 1,2 surprisingly little is known about this CO 2 -trapping reaction. Stephan et al. had used Siebert's unsaturated geminal BCl 2 compound 1 (ref. 3) and the corresponding B(C 6 F 5 ) 2 analogue 2, which was derived from 1 by treatment with Zn(C 6 F 5 ) 2 , for FLP/CO 2 scavenging, 4 but the vast majority of FLP/CO 2 chemistry used non-chelate Lewis acidic binding motifs 5,6 (Scheme 1). We have now investigated the t Bu 3 P/CO 2 trapping reaction using a pair of geminal C 6 F 5 containing bis-boranes. Both were obtained by the treatment of the respective terminal acetylene starting materials with two molar equivalents of Piers' borane [HB(C 6 F 5 ) 2 ]. While we observed the expected normal behaviour upon reacting the alkyne Ph-CH 2 CH 2 C^CH (5a) with the hydroboration reagent, we observed a rather complex rearrangement behaviour that took place upon the treatment of cyclopropylacetylene (5b) with the HB(C 6 F 5 ) 2 borane. The characterization of the resulting special rearrangement product, its formation and its FLP reaction with CO 2 in the presence of a tert-phosphine will be presented and discussed in this account. ## Results and discussion The Ph-CH 2 CH 2 C^CH/2HB(C 6 F 5 ) 2 system Terminal acetylenes undergo regioselective 1,2-hydroboration with the HB(C 6 F 5 ) 2 reagent to yield the respective substituted vinyl boranes. 7 When the reaction is carried out in a 1 : 2 molar ratio of alkyne and [B]H borane, the respective saturated geminal bis-borane is obtained in many cases under kinetic control. 1 This typical reaction path was also observed when we treated the alkyne 5a with HB(C 6 F 5 ) 2 in a 1 : 2 ratio in toluene solution at r.t. (1 hour reaction time). Workup gave the product 6a, which we isolated as a white solid with a 76% yield. The compound was characterized by C,H-elemental analysis and by spectroscopy, and we carried out some characteristic reactions. Compound 6a shows a single 11 B NMR resonance at d ¼ 72.1 ppm, which is typical for Lewis acidic planar tricoordinate R-B(C 6 F 5 ) 2 situations. 8 Consequently, we observed the three 19 F NMR signals of the symmetry-equivalent C 6 F 5 groups at boron. They show a typical large meta/para fluorine NMR chemical shift difference (Dd 19 F m,p ¼ 13.7 ppm). The mixture of compound 6a with the bulky phosphine P t Bu 3 (1 : 1) Scheme 1 Geminal bis-boranes and their FLP reactions with CO 2 . represents a reactive frustrated Lewis pair that is able to heterolytically split dihydrogen 9 under mild conditions (r.t., 2.0 bar of H 2 , overnight in pentane). The product precipitated from the reaction mixture and was isolated as a white solid with an 87% yield. Compound 7 was characterized by X-ray diffraction (single crystals were obtained from pentane/dichloromethane at 35 C by the diffusion method). Compound 7 shows a fully extended all anti-periplanar C 4chain featuring the phenyl substituent at one end and the geminal pair of boryl groups at the other. The C1-B1/B2 bonds are almost of the same length and the pair of boron atoms is bridged by the hydride (see Fig. 1). In the crystal there is an independent HP t Bu 3 + countercation. In solution, compound 7 shows hydride gives rise to a broad 1 H NMR signal at d ¼ 2.64 ppm (Scheme 2). We then treated the bis(boryl)alkane/phosphine FLP [6a/ P t Bu 3 ] with carbon dioxide. Exposure of the 6a/P t Bu 3 mixture in pentane at r.t. to CO 2 (2.0 bar) quickly (in 2 hours) resulted in the formation of a white precipitate of compound 8, which was isolated with an 81% yield. Compound 8 is sensitive in solution (CD 2 Cl 2 ) and decomposed above 0 C. Single crystals of the FLP/ CO 2 adduct 8 suitable for characterization by X-ray diffraction were obtained from pentane/dichloromethane at 35 C by the diffusion method (see Fig. 2). In the crystal, compound 8 shows a gauche/anti-periplanar conformation of the Ph-CH 2 CH 2 CH 2 -CH-chain. The geminal pair of B(C 6 F 5 ) 2 substituents at carbon atom C1 has taken up the CO 2 molecule in a rather symmetric way by forming two boron-oxygen bonds of almost the same length, and also the C5-O1/O2 bonds are almost equal in length, indicating a fully delocalized structure for this submoiety of compound 8. The resulting six-membered heterocycle features an almost coplanar arrangement of the BOCOB unit with only the carbon atom C1 being localized markedly outside of this plane. The bulky P t Bu 3 group is found attached at the central carbon atom C5 of this heterocyclic subunit of the overall molecular zwitterionic FLP/CO 2 addition product 8. In solution we observe the 13 C NMR resonance of the scavenged CO 2 molecule at d ¼ 172.7 ppm with a 1 J PC coupling constant of 92. The cyclopropylacetylene/2HB(C 6 F 5 ) 2 system: rearrangement to tetrahydroborole derivatives We next reacted cyclopropylacetylene (5b) with two molar equivalents of Piers' borane [HB(C 6 F 5 ) 2 ] (toluene, r.t., 1 hour). In Fig. 1 A projection of the molecular structure of the FLP dihydrogen splitting product 7 (thermal ellipsoids are shown with a 30% probability level). Selected bond lengths () and angles (degrees): B1/B2 1.945( 5), C1-C2 1.526(4), C1-B1 1.609( 5), C1-B2 1.598( 5), B1-H01 1.26(4), B2-H01 1.34(3), B2-C1-B1 74.7(2), B2-H01-B1 96.9. Scheme 2 Frustrated Lewis pair reactions with the geminal bisborane. 6a. Fig. 2 A view of the molecular structure of the zwitterionic FLP/CO 2 adduct 8 (thermal ellipsoids are shown with a 50% probability level; hydrogen atoms of the P t Bu 3 group and the C 6 F 5 substituents at boron atoms B1 and B2 are omitted for clarity: for details see the ESI †). Selected bond lengths () and angles (degrees): B1-O1 1.628(4), B2-O2 1.635(4), O1-C5 1.265(4), O2-C5 1.264(4), B2-C1-B1 108.6(2), O2-C5-O1 127.7 (3). this case, we did not obtain the simple cyclopropyl-CH 2 CH [B(C 6 F 5 ) 2 ] 2 product (6b), but found that a rearrangement had occurred. In situ NMR spectroscopy revealed the formation of a ca. 7 : 1 mixture of the a-boryl-tetrahydroborole products cis-9 and trans-9. We isolated the compound cis-9 in an almost pure condition (96 : 4) after workup as a pale yellow solid with a 67% yield (see Scheme 3). Compound cis-9 was characterized by X-ray diffraction using single crystals that were grown from a pentane solution of the compound at 35 C (see Fig. 3, left). The X-ray crystal structure analysis has shown that a fve-membered saturated tetrahydroborole framework had been formed in the reaction, bearing a B(C 6 F 5 ) 2 substituent at the a-position C1, a methyl substituent at C3 and a C 6 F 5 substituent at C4. The pair of substituents in cis-9 at carbon atoms C1 and C4 are cis-oriented; both are in a trans-arrangement with the methyl substituent at carbon atom C3. The plane of the C 6 F 5 group at C4 is oriented markedly away from the mean heterocyclic core [dihedral angle q B1-C4-C21-C22 119.6( 1) ], whereas the C 6 F 5 group at the adjacent boron atom B1 is rotated slightly in the opposite direction [q C1-B1-C11-C12 49.0( 7) , C4-B1-C11-C12 136.2( 5 We tried to fnd a mechanistic rationale for the formation of the boryl tetrahydroborole product 9 in the reaction of cyclopropylacetylene (5b) with two HB(C 6 F 5 ) 2 equivalents. It is known that cyclopropanes are often readily opened to the respective olefn isomers upon exposure to boron Lewis acids. 10 Therefore, we briefly checked whether the opened isomer of 5b, 2-methyl-1-buten-3-yne, might be involved in this reaction. However, this was not the case. Its reaction with two equivalents of HB(C 6 F 5 ) 2 took a different course (for details, see the ESI †). Therefore, we assumed a reaction pathway as outlined in Scheme 4. It is known that 5b undergoes a single hydroboration with Piers' borane to give 10, so we assume it to be the initial intermediate. 11 With a second HB(C 6 F 5 ) 2 equivalent this can then undergo the subsequent hydroboration reaction to give the geminal bis-boryl substituted compound 6b. In the in situ NMR experiment we observed an intermediate which is likely 6b (for details, see the ESI †). This is not stable under our typical reaction conditions but undergoes Lewis acid induced cyclopropyl ring opening, potentially leading to 11 which is subsequently stabilized by a sequence of hydride/C 6 F 5 1,2-shifts to result in the observed product 9. We must stress that we so far have no information about the alleged intermediates on the way and we cannot convincingly explain, let alone predict, the preferred stereochemical outcome, aside from the assumption that the formation of the observed cis-9 product is following a pathway of least steric hindrance on the way (Scheme 5). The geminal bis-boryl compound contains a pair of Lewis acidic boron atoms and, consequently, it may serve as a chelate boron Lewis acid component in FLP chemistry. The isolated cis-9 in conjunction with the phosphorus Lewis base P t Bu 3 served as an active dihydrogen splitting reagent. Thus, treatment of a 1 : 1 mixture of cis-9 and P t Bu 3 with dihydrogen (2.0 bar) in pentane solution overnight produced the dihydrogen splitting product cis-12 as a precipitate. The salt cis-12 was isolated as a white solid with a 71% yield. We obtained single crystals of compound cis-12 from pentane/dichloromethane by a diffusion method which were suitable for characterization by X-ray Scheme 3 Formation of the tetrahydroborole derivative 9 (with unsystematical atom numbering scheme as used in Fig. 3). diffraction (see Fig. 4). In the crystal, we see the typical r-1-boryl, t-3-methyl, c-4-C 6 F 5 arrangement 12 of the substituents on the tetrahydroborole framework. There is now a hydride bridging between the two boron atoms. 13 Consequently, both the boron atoms B1 and B2 have attained distorted tetrahedral coordination geometries ( P B1 ccc ¼ 345.3 , P B2 ccc ¼ 349.8 ), and we found the HP t Bu 3 + cation in the crystal. The bulk isolated product cis-12 (in CD 2 Cl 2 ) contained ca. 15-20% contamination of the isomer trans-12 since we had started from a not completely pure starting material (for details, see the ESI; † the characterization of the independently synthesised isomer trans-12 will be described below). Compound cis-12 shows a pair of 11 B NMR signals in the typical borate chemical shift range (d ¼ 14.5 ppm, 19.7 ppm). It shows a 31 P NMR phosphonium doublet at d ¼ 60.6 ppm with 1 J PH $ 428 Hz. We also exposed the cis-9/P t Bu 3 FLP (again contaminated with a small amount of trans-9) to carbon dioxide (2.0 bar, r.t., overnight) in pentane solution. Under the typical conditions, the zwitterionic FLP/CO 2 addition product precipitated and was recovered by fltration to give cis-13 as a white solid with a 73% yield. The NMR analysis (in THF-d 8 ) again showed the presence of a second isomer (trans-13, see below, ca. 3%). Single crystals of cis-13 suitable for X-ray crystal structure analysis were obtained from pentane/dichloromethane at 35 C by the diffusion method (see Fig. 5). The compound contains a central heterocyclic six-membered ring that was formed by double chelate coordination of the geminal bis-boryl acceptor with the oxygen atoms of the phosphine activated carbon dioxide molecule. The structure of this subunit is largely delocalized with similar bond lengths in the B1-O1/B2-O2 pair as well as the C6-O1/O2 pair of carbon-oxygen bonds. Carbon atom C6 has the P t Bu 3 group attached to it. This chelate heterocycle is interlocked with the fve-membered tetrahydroborole framework, which has the boron atom B1 incorporated in it. This section of the molecule shows the same characteristic stereochemical features as we had found for its precursor cis-9. The hydrogen atoms at C1/C4 and the methyl substituent at carbon atom C3 are all in a cis-arrangement on this fve-membered ring. The boryl tetrahydroborole system 9 contains three independent carbon chirality centres. Therefore, there is the possibility of forming four diastereoisomers. So far our rearrangement reaction was rather stereoselective and produced the major product cis-9 with the relative stereoselectivity r-1, t-3, c-4 plus a small amount of a minor isomer which probably represents one of the other three diastereoisomers, but whose relative stereochemistry we did not know. We have now prepared and characterized the isomer "trans-9" (of relative r-1, c-3, t-4 stereochemistry) by a selective isomerization process at the saturated central heterocyclic framework. For that purpose, we treated the substituted tetrahydroborole product cis-9 [r-1, t-3, c-4] with a catalytic amount (20 mol%) of the persistent nitroxide radical TEMPO (pentane, r.t., 4 days). 11,14 This reaction apparently proceeded with reversible H-atom abstraction at the activated C1 position of the heterocycle and we isolated the trans-9 epimer [r-1, c-3, t-4] as a colourless solid with a 74% yield. This compound was characterized by C,H-elemental analysis, by NMR spectroscopy ( 11 B: d ¼ 79.6 ppm, 72.9 ppm, for details see the ESI †) and by Xray diffraction. Single crystals suitable for the X-ray crystal structure analysis of compound trans-9 were obtained from a pentane/dichloromethane mixture at 35 C (see Fig. 3, right). It shows the typical fve-membered tetrahydroborole framework with the B(C 6 F 5 ) 2 and C 6 F 5 substituents at carbon atoms C1 and C4 now in a trans relationship. The methyl group at C3 has remained trans oriented to the C 6 F 5 group at C4. Compound trans-9 also formed an active frustrated Lewis pair with P t Bu 3 . The system heterolytically cleaved dihydrogen at near to ambient conditions (pentane, r.t., 2.0 bar H 2 , overnight), and we isolated the hydridoborate/phosphonium salt with a 62% yield. It shows typical 11 B NMR signals at d ¼ 13.3 ppm and 17.1 ppm and a 31 P NMR feature at d ¼ 60.7 ppm ( 1 J PH $ 428 Hz). Compound trans-12 was characterized by X-ray diffraction (single crystals were obtained from pentane/dichloromethane at r.t. by the diffusion method). The X-ray crystal structure analysis (see Fig. 6) showed the presence of the hydride bridged pair of boron atoms inside the anion and Compound trans-12 shows a relative stereochemistry of r-1, c-3, t-4 (see Scheme 6 and Fig. 6). Compound trans-9 also reacts with carbon dioxide in the presence of P t Bu 3 . Exposing a mixture of trans-9 and tris(tertbutyl)phosphine in pentane solution overnight at r.t. to a CO 2 atmosphere gave the FLP/CO 2 adduct trans-13 as a white precipitate with a 76% yield. The compound turned out to be only sparingly soluble in many solvents. However it could be characterized by X-ray diffraction using single crystals that were directly obtained from the reaction mixture of trans-9/P t Bu 3 with CO 2 . The structure (see Fig. 5, right) confrmed the stereochemical assignment of the backbone of the compounds of this trans-series: in compound trans-13 the boryl substituent at carbon atom C1 is in a trans relationship with the C 6 F 5 substituent at the distal ring carbon atom C4, and the latter is oriented trans relative to the methyl group at C3. Consequently, the relative positions of the three substituents at the central tetrahydroborole framework in compound trans-13 are r-1boryl, c-3-methyl, t-4-C 6 F 5 confgured. The CO 2 oxygen atoms are found to be bonded to the pair of boron Lewis acid sites and the phosphorus atom is coordinated to the CO 2 carbon atom. The CO 2 bonding to the geminal bis(borane) acceptor is slightly unsymmetrical with the B1-O1 bond in the central position being markedly longer than the lateral B2-O2 contact and also the P1-C6 linkage is rather long (see Fig. 5). Compound trans-13 was just sufficiently soluble in d 8 -THF to allow the recording of most of its NMR features. The actual sample used was ca. 90% pure, and it contained a minor compound of unknown composition. Compound trans-13 shows a 31 P NMR resonance at d ¼ 57.4 ppm. The 13 C NMR signal of the CO 2 derived moiety Fig. 5 Projection of the molecular structures of the FLP/CO 2 addition product cis-13 [left, thermal ellipsoids are shown with a 50% probability level; hydrogen atoms of the P t Bu 3 group and the C 6 F 5 substituents at boron atoms B1, B2, and at carbon atom C4 are omitted for clarity: for details see the ESI; † selected bond lengths () and angles (degrees): P1-C6 1.905(2), B1-O1 1.657(2), B2-O2 1.636(2), O2-C6 1.258(2), O1-C6 1.255(2), O1-C6-O2 128.0(2), B2-C1-B1 110.9(2)] and trans-13 [right, the independent synthesis of trans-13 is described below; thermal ellipsoids are shown with a 30% probability level; hydrogen atoms of the P t Bu 3 group and the C 6 F 5 substituents at boron atoms B1, B2, and at carbon atom C4 are omitted for clarity: for details see the ESI; † selected bond lengths () and angles (degrees): P1-C6 1.913(10), B1-O1 1.717(13), B2-O2 1.634(13), O2-C6 1.248( 12), O1-C6 1.271( 12), O1-C6-O2 128.7( 9), B2-C1-B1 118.1( 9)]. ## Conclusions We have shown in this study that the reaction of cyclopropylacetylene with two molar equivalents of Piers' borane [HB(C 6 F 5 ) 2 ] takes an unusual course. We assume that initially the usual two-fold hydroboration reaction of the terminal alkyne takes place with the anti-Markovnikov orientation generating the respective geminal bis-boryl compound. This is apparently not stable under the applied mild reaction conditions, but undergoes an intramolecular rearrangement process initiated by cyclopropyl ring opening by the adjacent strong borane Lewis acid. This initiates a series of 1,2-migration reactions involving the migration of one C 6 F 5 group from boron to carbon which eventually yields the a-boryl tetrahydroborole system 9. This is obtained with a rather high diastereoselectivity from this rearrangement process. The major compound cis-9 is an active FLP dihydrogen cleavage reagent in the presence of the bulky P t Bu 3 Lewis base. The cis-9/P t Bu 3 FLP also sequesters CO 2 cleanly in a chelate fashion, similar to the here studied more Lewis acidic geminal R-CH[B(C 6 F 5 ) 2 ] 2 reference systems, despite the loss of one electron withdrawing C 6 F 5 substituent at a boron atom. This probably indicates the favourable influence of the geminal bis-boryl situation for both chelate hydride and chelate CO 2 binding. ## Preparation of compound 6a A solution of 4-phenyl-1-butyne (5a, 65.0 mg, 0.50 mmol) in toluene (1.0 mL) was added to a suspension of bis(penta-fluorophenyl)borane (345 mg, 1.00 mmol) and toluene (3.0 mL). The reaction mixture was stirred at room temperature for 1 hour and then the suspension was fltered by cannula fltration. The volatiles of the obtained fltrate were removed in vacuo to give a colorless oil. Subsequently pentane (4.0 mL) was added and the mixture was stored at ca. 35 C overnight. The formed white powder was isolated by fltration, washed with pentane (2 1 mL) and dried in vacuo to give compound 6a (312 mg, 0.38 mmol, 76%) as a white solid. Anal. calc. for C 34 H 12 B 2 F 20 : C, 49.68%; H, 1.47%. Found: C, 49.40%; H, 1.40%. For the NMR data see the ESI. † ## Preparation of compound 7 A solution of compound 6a (82.2 mg, 0.10 mmol) and tri-tertbutylphosphine (20.5 mg, 0.10 mmol) in pentane (3.0 mL) was exposed to a hydrogen atmosphere (2.0 bar) at room temperature and stirred overnight. The resulting white precipitate was collected by cannula fltration and washed with pentane (3 2 mL). After the removal of all volatiles in vacuo, compound 7 was obtained (88.6 mg, 0.087 mmol, 87%) as a white solid. Anal. calc. for C 46 H 41 B 2 F 20 P: C, 53.83%; H, 4.03%. Found: C, 53.81%; H, 4.01%. Single crystals suitable for the X-ray crystal structure analysis were obtained by the slow diffusion of pentane into a solution of compound 7 in dichloromethane at 35 C. ## Preparation of compound 8 A solution of compound 6a (123.3 mg, 0.15 mmol) and tri-tertbutylphosphine (30.3 mg, 0.15 mmol) in pentane (5.0 mL) was exposed to CO 2 (2.0 bar) at room temperature and then stirred for 2 hours. The resulting white precipitate was isolated by cannula fltration and washed with pentane (3 1 mL). After drying the solid in vacuo, compound 8 (129.4 mg, 0.12 mmol, 81%) was obtained as a white powder. Anal. calc. for C 47 H 39 B 2 F 20 O 2 P: C, 52.84%; H, 3.68%. Found: C, 53.21%; H, 3.91%. Single crystals of compound 8 suitable for the X-ray crystal structure analysis were obtained by the slow diffusion of pentane into a solution of the white powder in dichloromethane at 35 C. ## Preparation of compound cis-9 A solution of compound 5b (33.0 mg, 0.50 mmol) in toluene (1.0 mL) was added to a suspension of bis(pentafluorophenyl)borane (345 mg, 1.00 mmol) and toluene (3.0 mL). After stirring the reaction mixture at room temperature for 1 hour, the solution was separated from the resulting suspension by cannula fltration. Then all volatiles of the fltrate were removed in vacuo to give a yellow oil, which was dissolved in pentane (2.5 mL) and stored at 35 C overnight. The precipitated pale yellow solid was isolated by fltration and washed with cold pentane (2 0.5 mL). The removal of all volatiles in vacuo gave a pale yellow solid (253 mg, 0.34 mmol, 67%). Anal. calc. for C 29 H 8 B 2 F 20 : C, 45.95 %; H, 1.06%. Found: C, 45.74%; H, 1.07%. Crystals of compound cis-9 suitable for the X-ray crystal structure analysis were obtained from a solution of the yellow solid in pentane at 35 C. Preparation of compound trans-9 TEMPO (16.6 mg, 0.11 mmol) was added to a solution of compound cis-9 (400 mg, 0.53 mmol) in pentane (15 mL). After stirring the reaction mixture at r.t. for 4 days, the resulting suspension was concentrated to about 2.0 mL, and stored in the fridge (35 C) overnight. The precipitated white powder was isolated via cannula fltration, and washed with cold pentane (2 1.0 mL). The removal of all volatiles under reduced pressure gave product trans-9 (296 mg, 0.39 mmol, 74%) as a white solid. Anal. calc. for C 29 H 8 B 2 F 20 : C, 45.95%; H, 1.06%. Found: C, 45.45%; H, 0.95%. Crystals suitable for the X-ray crystal structure analysis were obtained from a solution of compound trans-9 in pentane (1.5 mL) and CH 2 Cl 2 (0.5 mL) at 35 C. ## Preparation of compound cis-12 A solution of compound cis-9 (cis/trans z 96/4, vide supra) (113.7 mg, 0.15 mmol) and tri-tert-butylphosphine (30.3 mg, 0.15 mmol) in pentane (5.0 mL) was exposed to dihydrogen (2.0 bar) at room temperature and then stirred overnight. The formed white precipitate was collected by cannula fltration and washed with pentane (3 1 mL). After the removal of all volatiles in vacuo, a white solid was obtained (101.4 mg, 0.11 mmol, 71%). Anal. calc. for C 41 H 37 B 2 F 20 P: C, 51.17%; H, 3.88%. Found: C, 50.96%; H, 3.76%. Single crystals of compound cis-12 suitable for the X-ray crystal structure analysis were obtained by the slow diffusion of n-pentane into a solution of the white solid in dichloromethane at room temperature. ## Preparation of compound trans-12 A solution of compound trans-9 (75.8 mg, 0.10 mmol) and tritert-butylphosphine (20.2 mg, 0.10 mmol) in pentane (4.0 mL) was exposed to a dihydrogen atmosphere (2.0 bar) at room temperature and stirred overnight. The formed white precipitate was collected by cannula fltration and washed with npentane (2 1 mL). The removal of all volatiles in vacuo gave compound trans-12 (59.2 mg, 0.062 mmol, 62%) as a white solid. Anal. calc. for C 41 H 37 B 2 F 20 P: C, 51.17%; H, 3.88%. Found: C, 51.09%; H, 3.67%. Single crystals suitable for X-ray crystal structure analysis were obtained by the slow diffusion of pentane into a solution of compound trans-12 in dichloromethane at room temperature. ## Preparation of compound cis-13 A solution of compound cis-9 (cis/trans z 96/4, vide supra) (113.7 mg, 0.15 mmol) and tri-tert-butylphosphine (30.3 mg, 0.15 mmol) in pentane (5.0 mL) was exposed to a CO 2 atmosphere (2.0 bar) and then stirred overnight at room temperature. The formed white precipitate was collected by cannula fltration and washed with pentane (3 1 mL). After the removal of all volatiles in vacuo, a white solid was obtained (108.3 mg, 0.11 mmol, 73%). Anal. calc. for C 42 H 35 B 2 F 20 O 2 P: C, 50.23%; H, 3.51%. Found: C, 50.07%; H, 3.36%. Single crystals of compound cis-13 suitable for the X-ray crystal structure analysis were obtained by the slow diffusion of n-pentane into a solution of the obtained white solid in dichloromethane at 35 C. ## Preparation of compound trans-13 A solution of compound trans-9 (75.8 mg, 0.10 mmol) and tritert-butylphosphine (20.2 mg, 0.10 mmol) in pentane (5.0 mL) was exposed to CO 2 (2.0 bar) at room temperature and then stirred overnight. The formed white precipitate was collected by cannula fltration and washed with pentane (3 1 mL). After the removal of all volatiles in vacuo, compound trans-13 (76.2 mg, 0.076 mmol, 76%) was obtained as a white solid. Anal. calc. for C 42 H 35 B 2 F 20 O 2 P: C, 50.23%; H, 3.51%. Found: C, 49.96%; H, 3.27%. Single crystals of compound trans-13 suitable for the X-ray crystal structure analysis were obtained directly from a reaction solution of compound trans-9 (37.9 mg) and tritert-butylphosphine (10.1 mg) and dichloromethane (1.0 mL) in a CO 2 atmosphere (2.0 bar) at room temperature.

chemsum

{"title": "Geminal bis-borane formation by borane Lewis acid induced cyclopropyl rearrangement and its frustrated Lewis pair reaction with carbon dioxide", "journal": "Royal Society of Chemistry (RSC)"}

full_spectrum_raman_excitation_mapping_spectroscopy

## Abstract: A generalization of the Raman scattering (RS) spectrum, the Raman excitation map (ReM) is a hyperspectral two-dimensional (2D) data set encoding vibrational spectra, electronic spectra and their coupling. Despite the great potential of REM for optical sensing and characterization with remarkable sensitivity and selectivity, the difficulty of obtaining maps and the length of time required to acquire them has been practically limiting. Here we show, with a simple setup using current optical equipment, that maps can be obtained much more rapidly than before (~ms to ~100 s now vs. ~1000 s to hours before) over a broad excitation range (here ~100 nm is demonstrated, with larger ranges straightforward to obtain), thus taking better advantage of scattering resonance. We obtain maps from different forms of carbon: graphite, graphene, purified single walled carbon nanotubes (SWCNTs) and chirality enriched SWcnts. the relative speed and simplicity of the technique make ReM a practical and sensitive tool for chemical analysis and materials characterization.RS spectra are one dimensional (1D) plots of intensity versus wavelength shift providing a fingerprint for chemical analysis 1 (including chemometrics 2 ), widely applied to nanocarbons 3 , and many other sample types. In RS, incident monochromatic light is scattered by phonons to produce peaks shifted by the phonon frequency. In micro-RS, a microscope objective focuses light onto a sample, elastically (Rayleigh) scattered light is filtered out, and the scatter is analyzed (inset, Fig. 1). This is usually limited to a single laser wavelength (λ), or a few discrete wavelengths.However, the precise choice of λ is consequential for the signal intensity because it has non-resonant RS and resonant RS (RRS) components 4 . Scattering efficiencies can rise by orders of magnitude (~10 3 -10 8 ×) for λ near a resonance. The intensity of a Raman mode vs. λ is its resonant excitation profile (REP) 5 . Peaks in the RS spectra are vibrational, while features in the REP are electronic in origin. Plotting the intensity versus Raman shift and λ makes a REM, with horizontal slices that are RS spectra and vertical slices that are REPs. Ordinarily, to obtain a REM, many laser wavelengths are taken sequentially with a tunable laser to build up a two-dimensional (2D) map 6,7 .While these techniques are very general, RRS (and REM) is particularly important for nanocarbons 3,8 . For SWCNTs, the precise chemical structure is linked to the optical spectrum by the Kataura 9 plot, with which RRS can be used to identify nanotube species 10 . Such data helps detect trace metals and qualify/quantify the metal/ semiconducting purity in enriched SWNCT materials, an issue in nanoelectronics 6,11,12 . In spite of its proven analytical value, because it is slow and technically cumbersome 6,13,14 REM is not widely used, even for nanocarbons.The proposed hyperspectral "full spectrum" approach makes REM rapid and relatively simple, having no moving parts, and relatively low cost to implement. By "full spectrum" here, we mean this can be done in principle for any color, and that many wavelengths are used all at once. Updating a classic approach to parallelization 15 , this is akin to "line illumination" RS 1 but rather than illuminating monochromatically, a color gradient is used. Recent photoluminescence (PL) excitation experiments use such an approach 16,17 . Challenges for using such methods for RS as opposed to PL include the weak signal (~10 −6 vs. ~10 −1 efficiency for PL 18 ), the strong, unwanted Rayleigh background, the need for higher (~10×) spectral resolution, the potential for complications due to laser heating, and the more complicated data processing that is required. All these challenges are overcome here with current generation optical components.It is technically difficult and costly to span a large range of excitation wavelengths for RS with conventional monochromatic lasers which are ordinarily wavelength tuned for REM. Instead, here a supercontinuum (SC) light source provides broadband (~450 nm-2200 nm) light for all wavelengths, all at once. Unlike other broadband sources, SC has laser-like collimation and so can be tightly focused and otherwise manipulated, but it still has good spectral power density (~0.1 mW/cm −1 ). Some important steps have been demonstrated with broadband light sources. Recently, SC has been used to map graphene by stepping serially in time 19 . Also, notably, (non-SC) diode illumination has been spectrally dispersed in a line to obtain 1D RS spectra 20 . Here, SC is nearly monochromatic in a spatially compact spot (~10 μm), and spatially dispersed (~mm scale) to cover a large excitation bandwidth (here ~100 nm), so that spectra are high resolution, but such that sample heating is insignificant. Unlike conventional RRS, an entire range of excitation wavelengths is used simultaneously. ## Methods Setup. See Fig. 1 for a schematic illustration of the setup. The supercontinuum light source (SC) was an NKT Photonics SuperK Extreme High Power Super Continuum White Light Laser (EXR-15). The cold filter (CF) stage, depicted here as a single filter, was actually a folded cavity with a visible band high reflecting mirror slightly tilted from normal. This allowed multiple reflections before transmission and so acted as a high rejection bandpass filter, passing the visible light but not the strong NIR which is a source of heat, and background for RS. A short wave pass edge filter designed for anti-Stokes RS spectroscopy at 633 nm (Iridian Spectral Technologies) was used as an excitation filter (ExF). The grating to disperse the exciting white light (ExG) was a 1200 lines/mm holographic transmission grating (Wasatch Photonics). The dispersed excitation light was focused to a line with a 10 × 0.26 numerical aperture long working distance NIR microscope objective (ExL, Mitutoyo). Samples were mounted by taping them to a glass slide and clamping them on a micrometer driven xyz stage. Collection was by a 10 × 0.45 numerical aperture long working distance microscope objective (EmL, Edmund Optics). A long wave pass edge filter designed for Stokes Raman spectroscopy at 633 nm was used as an emission filter (EmF, Semrock). A 1200 lines/mm holographic transmission grating (Wasatch Photonics) was used to disperse the scattered light. A 75 mm focal length achromatic lens (Thorlabs) was used as a tube lens (TL). The camera (C1) was a 5.5 megapixel cooled CMOS detector (Andor Neo 5.5CL). The other camera (C2) was a low cost room temperature megapixel webcam (Edmund Optics). ## Materials. The HOPG (Structure Probe, Inc.), graphene (Structure Probe, Inc.), and unsorted SWCNTs (Raymor Nanointegris) were obtained commercially. The HOPG was taped to a glass slide with double sided tape, and a second glass slide with double sided tape was pressed against it to remove the top surface and reveal a clean layer. The graphene sample was supported by silicon substrate which was taped to a glass slide. The unsorted SWCNTs were aqueous suspensions and were drop cast directly onto a glass slide producing a black, optically thick film. The chiral and diameter sorted SWCNTs (6,5), (9,8) and (7,5) were sorted from as-prepared bulk SWCNTs using polymer wrapping procedures (See Supplementary Information for details). Sorted SWCNT dispersions were drop cast on PTFE membranes (0.2 µm pore size) until optically thick films were formed, and were rinsed in toluene to remove the excess polymer. The PTFE membranes were taped onto glass slides for measurement. ## Spectral calibration. The emission wavelength was calibrated with three narrow band edge filters placed at the "EmF" position at 659.2 nm (Alluxa), 706.5 nm (Alluxa) and 752 nm (Iridian Spectral Technologies). The excitation wavelength was calibrated to the band edge of short wave pass filter near 633 nm and by setting the G band RS peak resonance to 1590 cm −1 . Spatial resolution. This instrument did not reach diffraction-limited spatial resolution. For our setup, the illumination line on the sample was wide (~43 μm) and this limits the spatial resolution horizontally -i.e. perpendicular to the illumination line axis. (See Supplementary Information Fig. S1). Parallel to the illumination line, intensity fluctuations as small as ~1 pixel in height were observable (See Supplementary Information Fig. S2). This is 6.5 μm on the spectroscopy camera, which, at the magnification of ~3.8×, corresponds to a spatial resolution of 1.7 μm on the sample surface. This is the limiting factor for the spatial resolution along the vertical axis. With appropriate components, we expect that much higher spatial resolution could be obtained from a similar instrument. For example, using higher numerical apertures and pixel sizes adapted to the focal spot will increase the spatial resolution. ## Spectral resolution. There are two spectral resolutions to consider in REM, the Raman spectral resolution (x-axis) and the excitation spectral resolution (y-axis). The spectral resolution can be estimated with a spectral line which is narrower than the resolution. The full width at half maximum (FWHM) of G band of HOPG was ~28 pixels wide on the spectroscopy camera, corresponding to ~180 μm on the camera and ~47 μm on the sample. This is similar to the spatial size of the illumination line, providing evidence that the spatial linewidth is setting the spectral resolution of this instrument. (See Supplementary Information Fig. S3 for more detail.) Given the dispersion of ~0.67 pixels/cm −1 this corresponds to a spectral resolution of ~41 cm −1 . This is also the effective limit on the spectral resolution with respect to excitation wavelength. The specifications of the supercontinuum light beam can limit the spectral resolution. First, the beam diameter is specified at ~1 mm at 530 nm increasing to ~2 mm at 1100 nm. This beam diameter is smaller than microscope objective's pupil size. So, it is this beam diameter that determines the effective numerical aperture for focusing. The spatial resolution scale d is given by the Abbe relation d = λ/(2n sinθ), with λ the wavelength, n the index of refraction and θ the collection angle. Using 0.5 mm beam radius at a working distance of 30 mm from sample to microscope objective gives θ ≈ 1° which, using n = 1 for air, gives a resolution of ~17 μm at 530 nm. This is comparable in scale to the experimental resolution, and so may already be a limiting factor here. Steps that would improve this include using a microscope objective with higher numerical aperture/shorter working distance, using a beam expander on the supercontinuum beam, or using a different supercontinuum source that better fills the objective. A separate consideration is the pulsed nature of the light source. Here, the supercontinuum light source is seeded by a laser with 5 ps duration pulses. Although the supercontinuum light is not by any means a monochromatic pulse, it is instructive to consider how monochromatic the spectrally dispersed light could actually be if it had this same 5 ps pulse duration. A Fourier transform limited pulse has Δτ•Δν = P where P is a pulse shape dependent and is ≈0.4 for Gaussian pulse shapes. For a Δτ = 5 ps Gaussian pulse then, Δν = 80 GHz. At 633 nm (474 THz), 80 GHz corresponds to 0.11 nm. In cm −1 , 633 nm corresponds to 15798 cm −1 , so that results in a spectral resolution of ~3 cm −1 . This would be a best case resolution spectral resolution for a 5 ps Gaussian monochromatic pulse. Although the spectrally dispersed visible light from the supercontinuum is far removed from this simple situation, it does suggest that the spectral resolution could be limited to ~ a few cm −1 due to the pulsed nature of the beam. If this is the case, the spectral resolution can be improved by using a source with a longer pulse duration, or, ideally, a continuous wave (cw) source. ## Spectral bandwidth. Since the supercontinuum light is extremely broadband, what determines the wavelength range is the dispersion of the excitation grating (ExG) and the size of the excitation microscope objective entrance pupil, or any other such obstruction in the optical path. Physically, the grating disperses the collimated white light supercontinuum beam, converting wavelength into angle. The range of wavelengths in the illumination line is set geometrically by this angle, the diameter of the beam, and the size of the entrance pupil going into the microscope objective. The smaller the grating dispersion the larger the bandwidth, and the larger the entrance pupil, and the closer it is to the grating, the larger the bandwidth. Higher magnification objectives tend to have smaller entrance pupils, so spatial resolution is reduced if bandwidths are increased by increasing entrance pupil diameters. The instrument we demonstrate here has a fairly high grating dispersion, and a fairly large entrance pupil. Both the grating dispersion and the entrance pupil can be made larger or smaller, if desired, with off the shelf optical components. Lower dispersion gratings lead to smaller angles, and so higher bandwidth, but lower dispersion gratings have lower spectral resolution, so there is the usual trade-off between resolution and bandwidth in grating spectrometers if the grating dispersion is reduced to obtain a larger range of wavelengths. Wavelength dependence of intensity. Any optical system will have some wavelength dependence. For the most part, the wavelength dependence of this system is gradually changing with wavelength, so any significant (2020) 10:9172 | https://doi.org/10.1038/s41598-020-65757-9 www.nature.com/scientificreports www.nature.com/scientificreports/ changes in intensity here can be attributed to the sample, except, obviously, near the cut-on or cut-off edges of the interference filters. The supercontinuum source, camera, and gratings have the most variation with wavelength, but all are quite gradual. The supercontinuum light intensity primarily increases gradually with increasing wavelength, and is evaluated explicitly below. The camera specifications give a quantum efficiency near 60% at 600 nm dropping to ~33% at 800 nm. The detection grating specifications have the opposite trend, with a smoothly varying efficiency curve near ~60% efficiency at 700 nm and ~85% efficiency at 800 nm. The interference filters have transmissions specified as better than 93% and so only introduce minor ripple effects (a few %) away from their cut-off wavelengths. The result of all this is that REMs we obtain are meaningful, even without any calibration, keeping in mind these gradual, smaller variations, and the filter cut-offs. The rigorous calibration of Raman scattering intensity as a function of Raman shift is not simple, even for standard, monochromatic laser Raman spectroscopy. Procedures for calibrating conventional monochromatic laser Raman spectra are described in detail in ref. 1 and codified into standard protocols as described in ref. 21 The reason for the complexity is not only that the collection optical system will have some wavelength and polarization dependence, but also because Raman scattering matrix elements depend on the spatial orientation with respect to the polarization of light. In fact, all aspects of the metrology of the calibration of Raman spectra continue to be actively developed by standards organizations such as the Versailles Advanced Materials and Standards Organization (VAMAS) in its technical working area (TWA) 42 "Raman Spectroscopy" 22 . As indicated in ref. 1 , and still the case now, most published Raman spectra are not corrected for the instrumental response. To make such calibrations, the protocols described in ref. 21 use classic optical spectroscopic methods of "standard candles" where either a known, calibrated light source (e.g. blackbody of fixed temperature) or a substrate with a known fluorescence spectrum (e.g doped glass) is measured by the optical system. The wavelength dependent response of the detection system is determined from the ratio of the known signal and the measured signal. This ratio is applied as a factor (by assuming linearity) to subsequently measured data to obtain Raman spectra with calibrated relative intensity as a function of wavelength. These approaches can be extended to REM with the complication that the laser intensity (either in power or photon incident rate) is not generally constant as the wavelength is changed. For conventional REM, with laser wavelength tuned serially in time, the laser power can be actively controlled to maintain fixed power (or fixed photon/rate), or the laser power can be allowed to vary with wavelength and the spectra can be normalized with the assumption of linear scaling of scattering intensity with the incident photon rate -ordinarily a good approximation when the laser power is low. Similar monitoring of the laser power at all wavelengths could be accomplished here by directing part of the beam or spectrally dispersed Rayleigh scattering to a spectrometer or spectrum analyzer. There is a more current discussion of intensity calibration by such methods in tunable laser Raman excitation mapping (for lasers tuned serially in time) in refs. 5,6 . Here, with a "full spectrum" approach, intensity calibration is a particular challenge because two dimensions -excitation and collection -each with their own variations must be accounted for together, at the same time. The detection side will have an instrumental spectral response, which ideally would not -but possibly could -vary spatially across the focal plane of the camera. In addition, the supercontinuum light source varies in intensity with wavelength. Therefore it is simpler here, and probably more common in REM in general, to use a known Raman spectrum as a benchmark and compare intensities of the test sample to the known Raman spectrum. This is different from the standard procedures of ref. 21 where the samples have certified fluorescence spectra when illuminated by a single, specific laser wavelength. The fluorescence peak covers only a fairly limited range of wavenumbers and its shape is not generally the same, or certified for other wavelengths. As opposed to using fluorescence or illumination "standard candles", a good approach is to measure a known non-resonant Raman scatterer and obtain its REM, and use this known spectrum to calibrate the variation with laser power. Non-resonant Raman scattering intensities have the characteristic ~1/λ 4 scattering dependence of wavelength (in units of power). So, if one has a Raman scatterer that is known to be non-resonant, the evolution of the spectral lines over the map can be corrected by a multiplication factor at each point to reproduce the expected ~1/λ 4 non-resonant REP. A strength of this approach is its practical simplicity, as it relies only on measuring one extra sample or compound to obtain the correction factors. A weakness is that the response is only rigorously measured at the Raman bands, not over the entire map, so it works best for reference compounds with many spectral lines, and depends on having a sufficiently smoothly varying collection optics side spectral response. Here, we could correct these SWCNT spectra by dividing through by such correction factors derived from the HOPG or graphene map intensities. However, those maps are noisy, and so would more ideally be integrated longer. Also HOPG and graphene have relatively few spectral lines, although fortunately they are very close to some of the SWCNT lines of interest. So, we prefer not to explicitly make a correction like this, but rather we prefer to simply use the HOPG and graphene peak intensities as a guide to give us a good idea how much impact variations in illumination power and instrument response have on the spectra. Since the illumination and detection efficiencies are the same independent of sample -they all are subject to the same instrumental efficiencies -it is meaningful to compare different maps for different samples. Resonant structure can be confirmed in the SWCNT samples by comparing to the more gradually evolving HOPG or graphene intensities. The HOPG G band is an excellent benchmark for comparing to the SWCNT G bands. As an extra check on the meaning of the intensity variation in these maps, the wavelength dependence of the illumination intensity was also measured separately. (See Supplementary Information, Fig. S4) The intensity, in counts, increased with wavelength, roughly by a factor of two over the scan range, similar to a blackbody source but with an increase just before the cut-off wavelength (633 nm). Therefore, if illumination intensity were the only factor, the scattering intensity would increase with wavelength. Of course, the non-resonant Raman scattering would have the ~1/λ 4 wavelength dependence, which would lead to a ~30% reduction in scattering probability over a change of ~50 nm in the visible range. (2020) 10:9172 | https://doi.org/10.1038/s41598-020-65757-9 www.nature.com/scientificreports www.nature.com/scientificreports/ In the future it will be important to develop rigorous and practical intensity calibration procedures for this method, just as is it important in more established methods of Raman spectroscopy. ## Signal-to-noise. To understand the performance of this particular instrument, and the visibility of the RS signal, it is useful to determine the signal-to-noise level. A detailed calculation, made by comparing two datasets from the same area is presented in the Supplementary Information (Fig. S5 and Table S1). For 500 ms acquisitions the most visible and prominent features had a signal to noise ratio (SNR) ratio of 18 and the three strongest SWCNT Raman bands had SNR ratio of 10 or greater. This is the SNR of the equivalent of a monochromatic laser's horizontal slice in the REM. Here, the incident power was ~10 mW dispersed over the band ~540 nm to ~630 nm. More qualitatively, the brightest bands were still visible and potentially usable for 5 msec integrations, and weak bands were easily visible with longer integrations, as might be expected. (Supplementary Information Fig. S6). ## Data processing. Acquisition was with Andor Solis software, using rolling shutter mode, and no binning. One minute integrations were taken by summing six 10 second integrations. (Integration times as short as 5 ms are demonstrated in the Supplementary Information Fig. S6) Images were converted to Ascii format. The data was imported using Pandas into Python 3. Data was manipulated with the numpy package and plotted using matplotlib. A constant base-plane background level has been subtracted from all maps (i.e. constant regardless of horizontal or vertical position on the map). To convert from horizontal pixel to Raman shift, the intensity at each Raman shift at 2 cm −1 intervals from 100 cm −1 to 3600 cm −1 was calculated by linearly interpolating the intensity measured in counts on the nearest neighbor pixels with calibrated shifts just to the left and right of these values. Essentially, the signal is recorded by pixel -which maps very nearly to wavelength because of the near-linearity of the grating dispersion -and is then processed into Raman shift. This is stated explicitly only to be clear: The intensity, although calculated at each Raman shift, is an intensity per pixel, which most closely approximates a bin of fixed interval in wavelength, as opposed to a bin of fixed interval in Raman shift. These are, of course, different because Raman shift is an energy, not a wavelength. In other words, there is no adjustment in the intensity for the variation in the bin size of each pixel in terms of Raman shift. ## Results and Discussion Figure 1 shows a schematic of the setup. Details are in the Methods section. In essence, a SC source is filtered to block unwanted heating and background. A grating disperses the SC light spectrally and a microscope objective focuses it to a color-graded line. A separate objective collects the scattered light, with the excitation angled outside its collection cone. An edge filter blocks the Rayleigh scatter, and a grating angled orthogonally to the first disperses the signal, which is imaged by a camera. So, the vertical axis encodes the excitation wavelength, and the horizontal axis encodes the scattered wavelength. Figure 2 shows the result for a film of (7,5) SWCNTs. (See Methods). The incident power was ~10 mW from ~540 nm to ~630 nm. Peak intensities were ~10 5 counts/pixel in 60 s. Figure 2(a) shows the real-time camera image. Figure 2(b) shows this transformed to nm (excitation) and shift (cm −1 ). The filters create the triangular shape with a short wave pass excitation filter (horizontal) and a long wave pass (diagonal). Nine strong peaks are visible, and can be identified with tables. 3 Higher order combination modes area also visible in (a). Figure 2(c) shows a horizontal slice, corresponding to a conventional RS spectrum (623 nm). Figure 2(d) shows vertical slices corresponding to the D, G + and 2D band REPs. This is the integrated intensity of each band, by summing over a bin in Raman shift containing the band. Importantly, Fig. 2 recovers all the information that one gets from conventional monochromatic laser Raman spectrum. However, since a supercontinuum laser has been used, there is no restriction on the incident light due to the choice of the source (this source emits ~450 nm to ~2 μm), as there is for monochromatic lasers or even for tunable lasers, which generally cover a more limited range. In addition to the Raman spectrum the REPs of many bands can be evaluated simultaneously. Thus, a "full spectrum" REM can be obtained. This, our first demonstration of the method, uses a simple Rayleigh scattering rejection scheme. That is, we use a single short wave pass filter on the excitation optics and a single matched long wave pass filter on the collection optics side. The result is that the map is restricted to the triangular shape bounded by the cut-off wavelength of the filters. However, with no change to the configuration, the boundaries can easily be changed by swapping edge filters. The edge filters can also be angle tuned, a scheme used in some tunable filter Raman excitation mapping systems 6 . With a single short wave pass/long wave pass filter combination the bandwidth of REPs is small for small Raman shift, and increases as the Raman shift becomes large. So, for example, the 2D band has quite a wide REP while other bands may not. It must be noted that the ideal sample is flat and homogeneous. If the sample is sufficiently flat and homogeneous, a single acquisition is sufficient to obtain a representative REM. This is easily confirmed experimentally by translating the sample spatially in the plane, while obtaining real-time REMs. So, to obtain the resonant excitation profile (REP) in a single shot we are using spatial homogeneity of the sample. It is true that inhomogeneity of the sample will be reflected in the measured data. However, if desired, it is possible to scan the sample spatially to build up a multidimensional data set that includes REMs of each point on the substrate. However, these samples are sufficiently homogeneous that we have not done so here. Figure 3 shows different carbons under identical conditions. Horizontal lines are artifacts of scattering from particulates. The graphite (a) and graphene (b) signals are weak, so the color scale has been magnified 100× relative to the other samples. All samples show a G band (~1590 cm −1 ) -called G + for SWCNTs -and 2D band (~2600 cm −1 , also called G'). The G band for graphene is imperceptibly weak on this scale. (But it is shown in Supplementary Information Fig. S7). The graphene 2D band is only ~200 counts/pixel, with the strongest peaks coming from the silicon substrate underneath. The intensities for HOPG are ~400 counts/pixel (G and 2D (2020) 10:9172 | https://doi.org/10.1038/s41598-020-65757-9 www.nature.com/scientificreports www.nature.com/scientificreports/ bands). The SC power, camera response, and transmission are wavelength dependent, however they have the same dependencies for all the maps. Also, since the bands for HOPG are flat, they can be used as a benchmark for comparison. (See Methods for a detailed discussion). The SWCNT films produce much stronger signals. Figure 3(c) shows unsorted SWCNTs, and (d)-(f) show different species, all on the same scale. The unsorted sample has strong G + and 2D bands, but other bands are barely visible. Of the sorted, chirality enriched samples, the (7,5) in (f) is most intense at ~10 5 counts/pixel (G band), (~250×), greater than HOPG, and broader. The (6,5) sample reaches similar intensities, but the REP is shifted to smaller wavelength, as expected [~566 nm for (6,5) ~645 nm for (7,5) 24 ] The (9,8) species, which is far from resonance, is weaker, but it is still much more intense than HOPG (~10 2 ×). For REPs it is most meaningful to compare similar bands of different samples obtained under identical conditions. This accounts for the wavelength dependence of the instrument, since it is the same for all samples. Figure 4 shows extracted REPs for the G band region of all samples, except graphene which is barely above the noise (See Supplementary Information Fig. S7). are integrated REPs, obtained from integrating the maps over the entire G band region. The HOPG sample is much weaker than the others, and is not even visible on a linear scale in (a) and is therefore replotted in (b). Because the HOPG signal is relatively weak, a more careful background subtraction was used, with a sloped straight line background used in the signal area, and a linear segment background used in the blocked region. Unlike SWCNTs, HOPG does not have sharp one-dimensional resonances in the density of states and so its REP is relatively featureless, and its intensity variations arise primarily from the instruments spectral characteristics. Between the filter cut offs, the signal is largely flat except for a gradual decrease after 610 nm. Sharp spikes such as that at 612 nm are due to light scattering from particulates on the surface. The REPs of the chirality sorted samples can be compared to recently published REPs obtained with tunable Raman instruments. 5,23,25 The (6,5) SWCNT G + band REP is shown in these references, having a principal peak at ~575 nm and a smaller peak or shoulder at the G + phonon energy above the resonance at ~527 nm. References 23,25 also show the (7,5) SWCNT G + band, with a similar shape, having a principal REP peak at ~651 nm, and a shoulder or peak at the G + phonon resonance above at ~591 nm. References 5,25 show peak widths of ~0.1 eV FWHM. Reference 23 has a sharper REP peak of width ~25 meV, possibly narrower because of a more homogenous chemical environment, or more crystalline, or less defective nanotube structure. www.nature.com/scientificreports www.nature.com/scientificreports/ The REPs for chirality sorted SWCNTs we see here are broadly consistent. The (6,5) REP peak is at 580 nm or shorter wavelength, cut off by the filter. The (7,5) REP peak appears at ~615 nm but, from the HOPG spectrum it is clear that the sensitivity of the system falls as the wavelength is increased from this point, so this is a lower bound, and it could easily match the 651 nm reported in the other references. Interestingly the (7,5) REP spans the expected range for the two G + REP peaks. We do not observe a distinct high energy peak, but the inflection point near 580 nm may originate from this shoulder. This is consistent with the observation that in these references the low energy REP tends to be well defined and the high energy one much less so. The REPs peaks we observe have the large ~0.1 eV broadening of refs. 5,25 , not the sharper structure of ref. 23 . Finally it is interesting that the (6,5) shows a possible small shoulder just before the small wavelength cut-off. Speculatively, this may be the manifestation of the "bundling" REP reported in ref. 5 . Future experiments with broader scans and/or higher resolution should be able to confirm this. As we have shown, REMs and REPs can be obtained essentially in real time -orders of magnitude more rapidly than before. (See Supplementary Fig. S6 for examples of usable maps obtained in timescales as short as 5 ms.) The speed of mapping is not just a minor convenience but markedly changes the significance of the technique: it can be used as a routine characterization tool. Furthermore, two independent dimensions of data (vibrational, electronic) are obtained simultaneously, making these maps much more specific fingerprints than conventional one-dimensional RS spectra. They are therefore well suited as inputs into chemometric analysis systems. Commonly, in RS one has to choose an instrument that operates at an appropriate wavelength. The supercontinuum light with its extreme broadband wavelength range provides great versatility. Moreover, it becomes possible, as demonstrated here at least for some range, to obtain many or even all wavelengths of interest (a "full spectrum") all at once. Both these aspects help take full advantage of resonance to not only obtain stronger signals, but also obtain expanded spectral fingerprints. From a photophysical perspective, RRS cross-sections are particularly difficult to evaluate, not only because of the need to characterize instrumental throughput and response, but because, being resonant, they can have sharp, non-linear laser wavelength dependence. The exact resonance wavelength and the breadth of the resonance can be sample environment dependent. In a single fixed wavelength measurement, intensity changes arise from many causes, including this wavelength dependence, and having a continuous REM (in real-time) will help reveal these effects and take them into account. We believe this method has great potential, and it will benefit from further technical improvements. It is straightforward to extend the bandwidth, or to increase the resolution beyond what we have shown here. We have demonstrated the method on various advanced nanocarbon materials, but it is very general, being relevant to any sample that shows RRS. Given its capabilities, we speculate that full spectrum REM could become a practical tool for chemical analysis and photophysics.

chemsum

{"title": "Full Spectrum Raman Excitation Mapping Spectroscopy", "journal": "Scientific Reports - Nature"}

depletion_sphere:_explaining_the_number_of_ag_islands_on_au_nanoparticles

## Abstract: We report multi-site nucleation and growth of Ag islands on colloidal Au nanoparticles. By modifying a single factor, a range of products from Janus nanoparticles to satellite nanostructures are obtained.The identification of these key factors reveals the correlation between the concentration gradient and the choice of nucleation sites. In contrast to the inhibited homogeneous nucleation in the bulk solution, we argue that the non-steady-state concentration gradient plays a critical role in inhibiting nucleation within nanometer distance during the initial stage of growth-an essential but not yet recognized factor in colloidal synthesis. A depletion sphere model was developed, so that the multi-site nucleation is well integrated with the classic theory of nucleation and growth. Alternative explanations are carefully examined and ruled out. We believe that the synthetic know-how and the mechanistic insights can be broadly applied and are of importance to the advance of nanosynthesis. ## Introduction Nanoparticles are the basic building blocks of nanotechnology. Despite tremendous synthetic progress, the number of structural types is still far less than the molecules in the molecular world and the nuts and bolts in the macroscopic world. Hence, accumulation of synthetic know-how and mechanistic insights is essential for future development of functional architectures. Going beyond simple structures, 1 nanohybrids have attracted great attention for the promise of multi-functionality and synergistic effects. 2 Valence is an important chemistry concept defning an atom's combining power with other atoms. Similar concepts can be defned for satellite nanostructures with a precise number of island domains. 3 Typically, nanoparticles with a single island are called Janus nanoparticles, 4 whereas those with more than two islands are called satellite structures. The few examples of satellite structures in the literature were synthesized by polymer attachment, 5 colloidal assembly, 6 and growth. 3c,7 Defning the number of islands in such structures (i.e., achieving valency control) is a critical step towards "total synthesis" 3a of functional architectures. In this work, we demonstrate the control of valency in the colloidal growth of Au-Ag satellite nanostructures (Fig. 1), where the valency depends on the surface ligand density and the rate of Ag reduction. A depletion sphere mechanism was proposed to explain the choice of nucleation sites as the result of non-steady-state concentration gradient. ## Results and discussion The multi-island growth was an extension from the previous growth of a single Ag island on a Au seed. 8 Basically, the ligand 2-mercaptobenzoimidazole-5-carboxylic acid (MBIA) has a -SH group and a diametric -COOH group, allowing it to interact strongly both with the Au core and the subsequent Ag island. Fig. 1 Schematics illustrating the effects of depletion sphere (transparent sphere) in controlling the Ag nucleation on Au seeds and their later growth into Au-Ag Janus/core-satellite structures with different Au : Ag island ratio: 1 : 1; 1 : 2; 1 : 3; and 1 : 4. New nucleation sites can start when the depletion sphere does not fully cover the seed surface. Scale bar: 50 nm. Thus, partial embedding of this molecule in between the Au-Ag layers 9 creates defects and strains, making it possible to tune the Au-Ag interfacial energy. 8,10 As a result of the "invisible hands of thermodynamics", 11 various Au-Ag hybrids ranging from concentric core-shell, eccentric core-shell, acorn, and heterodimer nanostructures have been obtained. We made an interesting observation that improved ligand packing on the seed can lead to a greater number of Ag islands per seed. In a typical synthesis, as-synthesized 70 nm citratestabilized Au nanoparticles (Fig. S1 †) 8 were used as seeds. They were incubated with MBIA (20 mM in water) at elevated temperatures (60-100 C) to allow the ligands to pack well on the nanoparticle surface. After cooling to room temperature, the reductant hydroquinone (HQ) was added to the solution followed by AgNO 3 . With the start of Ag reduction, the resulting Ag atoms were deposited on the seed surface, forming islands. To prevent the product nanoparticles from aggregation, they were encapsulated in polystyrene-block-poly(acrylic acid) (PSPAA) shells. 8,12 As shown previously, the polymer encapsulation was only a method of preservation and did not alter the product structures (Fig. S5 †). In the TEM images, the Ag domains often have a lighter contrast than the Au seeds. The polymer shells appear as a white "halo" against the negatively stained background (Fig. 2a-c). With the shells, the nanoparticles are well separated from each other without further aggregation during the drying stage of TEM preparation, making it possible to distinguish the Ag islands. Energy-dispersive X-ray spectroscopy (EDS) mapping and line scans (Fig. 2g-i) verifed the Ag islands (red color) on the Au seed (green color). In contrast to the literature works where Janus/satellite nanostructures were developed in discrete systems, our system can give a range of products by modifying a single factor, for example, the ligand incubation step (Fig. 2). Au-Ag Janus nanoparticles of 95.7% purity were obtained when the Au seeds were incubated with MBIA at 60 C for 2 h (Fig. 2a). 8 When this incubation step was carried out at 80 or 100 C for 2 h, multiple Ag islands were grown on each Au seed, forming satellites. Fig. 2b shows the 80 C sample with 9.7% Au-Ag dimers, 56% Au(Ag) 2 trimers, and 35% Au(Ag) 3 tetramers; Fig. 2c shows the 100 C sample where most of the seeds have grown 4-6 islands. Across the samples of Fig. 2a-c, the color changed from brown to dark brown, and then deep green (Fig. S2a and b †), though the total amount of Au and Ag was the same. As shown in Fig. 2j, the UV-Vis spectrum of the Janus nanoparticles (Fig. 2a) showed three peaks at 420, 537, and 700 nm (curve (i) in red color), corresponding to the Ag and Au transverse absorption bands, and the Au-Ag longitudinal absorption, respectively. 8 In comparison, the presence of Au(Ag) 2 trimers in Fig. 2b led to a red shift of the longitudinal absorption band to 760 nm (curve (ii) in blue). For the satellite structures as shown in Fig. 2c, the overall round shape caused the longitudinal absorption to shift back to around 700 nm (curve (iii) in dark cyan). This behavior is similar to the multi-layer core-satellite assemblies reported by the Cheng group, 13 where the broad plasmonic absorption depends on the size and the number of satellite nanoparticles. The SERS signals of the samples roughly followed the same trend. As shown in Fig. 2k, sample 2b with the longest wavelength of the longitudinal absorption has the strongest SERS signal (curve (ii) in blue), likely due to the partial plasmon resonance with 785 nm incident light. 14 Prolonged incubation with ligands (at a constant temperature of 60 C) also caused the increase of satellite islands. With the increase of incubation time, the core-satellite structures showed a slight increase of valency (Fig. S6 †). At 15 h of incubation, the sample contained 52% Janus nanoparticles and 48% Au(Ag) 2 trimers (Fig. 3a). Among the trimers, 57% of them were roughly straight (with the Ag-Au-Ag angle $ 120 ), whereas 43% were more asymmetric with the Ag-Au-Ag angle < 120 (Fig. 3c). Further increase of incubation time to 2 days did not cause obvious increase of valency, suggesting that the effects have reached a plateau. In these experiments, the incubation step only pre-treats the seeds and all chemical reactions were carried out at room temperature. With the ligand concentration kept the same, it is unlikely that it would affect the degree of Ag + coordination, the rate of Ag reduction, or the rate of dynamic re-adsorption on the freshly generated Ag surface. 15 It is conceivable that the ligand density on the seed surface would be improved with longer incubation time and higher temperature, both of which can allow the ligand with flat geometry to move around and achieve more orderly packing. 16 But the two factors are obviously inequivalent in that even the (a-c) and 50 nm in (g-i). longest incubation time was less effective than the higher temperatures, presumably because some of the packing processes require a higher thermal energy to overcome the kinetic barriers. 16c Further exploration of the preparative conditions revealed that higher concentrations of HQ and NaOH can both promote multi-island growth. In the absence of NaOH and when all other conditions were kept the same, the increase of HQ concentration from 1.6 to 6.6, 13.2, and 19.9 mM caused a monotonous increase in the number of Ag islands per seed (Fig. 4a-h), with the average valency increased from 1 to 1.9, 2.5, and 3.6. Similarly, with HQ concentration (1.1 mM) and all other conditions kept the same, increase of NaOH concentration from 0.18 to 0.36, 0.54, and 0.72 mM led to increase of product valency (Fig. 4i-l). At high NaOH concentrations (0.54 and 0.72 mM), the crowded small Ag islands (>10 per seed) merged with each other to make a continuous shell with a rough surface (Fig. 4k-l). It is apparent that the valency of the satellite structures depends on the rate of Ag reduction. Both the higher reductant concentration and more basic conditions 17 promote Ag reduction, providing a greater amount of growth materials to the Ag islands and leading to a faster rate of color change. The effects of HQ and NaOH are inequivalent, as even the highest HQ concentration cannot achieve the high valency achieved by the high NaOH concentration (Fig. 4k-l). The multi-island growth can be achieved on seeds of various sizes. In the above discussion, 70 nm seeds are chosen to demonstrate a complete range of valency. As shown in Fig. 5a and b, the 15 and 25 nm Au seeds can also lead to a multi-island structure with modifed reaction conditions. 18 A signifcantly higher NaOH concentration was necessary for the small 15 nm seeds. We expect that the average valency should depend on the relative size of the seeds under the same reaction condition. To test this expectation, a control experiment was carried out using pre-mixed seeds of 15, 25 and 70 nm, giving average valencies of 1.1, 1.6 and 3.8, respectively (Fig. 5c). It should be noted that the size of the Ag islands on all seeds was similar, indicating that the initial Ag nucleation occurred roughly at the same time. It appears that the Ag islands form very quickly in the reaction. Fig. S7 † shows the kinetic UV-Vis absorption spectra during the growth of the sample in Fig. 3a. The appearance of Ag transverse absorption at 410 nm and the longitudinal absorption at >750 nm indicates that signifcant Ag islands have already formed within the frst scan (<1 min) and that the islands were fully grown by 5 min. It is conceivable that the initial island formation should occur much earlier than any Ag island of plasmonic signifcance. Previously, the mechanisms of Janus nanoparticles in the literature focused on the symmetry-breaking growth behaviorwhy an island is formed instead of a core-shell structure. In this branch of mechanistic discussion, symmetry-breaking can be achieved when: (1) part of the seed is covered by a special capping ligand 5a,19 or polymer coating; 20 (2) growth is preferred at the defect sites, such as stacking faults, twin plane, or dislocation; 7c,21 (3) primary seed particles aggregate with a special manner followed by oriented attachment; 22 or (4) various types of kinetic control, including supply of growth material via reaction rates, 23 and ligand kinetics. 15 We have discussed the detailed thermodynamic and kinetic arguments in our previous works. 8,11,15 In this work, we focus the discussion on the next step-why only one island is grown and what are the underlying reasons for the lack of islands on the remaining seed surface. In the literature, control of reactant concentrations or their rates of addition is a common method to exert synthetic control. The resulting satellite structures 7b,24 or partially encapsulated structures 23b,g were often broadly explained by "kinetic control", but there is a lack of theory to explain the detailed steps from the rate of reactions to the shape of nanostructures. Our ability in controlling the "on/off" of multi-island growth permits new mechanistic enquiries. We believe that further exploration of the detailed mechanistic scenario (as opposed to identifying a single factor) would greatly enhance our understanding and permit new synthetic designs. Throughout the images in Fig. 2-4, the Ag islands on any Au seed are well separated from each other, as if there is repulsion among them. Intuitively, one may invoke steric or charge repulsion. However, the Ag islands are neither liquid nor incoming particles, 6a incapable of moving to a different location under the influence of repulsion (Fig. 6a). At the point of initiation, the Ag islands must be ultrasmall and the repulsion among them needs to be extraordinary to exert sufficient influence. In the literature, polymer nanodroplets can grow on the surface of a particle or substrate, forming symmetrical satellite islands with similar appearance. 5c While it is possible that precursor droplets or highly swollen polymer domains may adjust their locations, it is impossible for the Ag islands here. Should the nucleation occur randomly on the Au surface, the probability of nucleation should be proportional to the surface area. For a Janus nanoparticle (Fig. 2a), at the point of initiation, the ultra-small Ag nucleus should occupy less than 10% of the nanoparticle surface. If we set the probability of the initial nucleation as P, the probability of a seed with second nucleation should be >P 0.9P. No matter what the P value is, it is virtually impossible to achieve one Ag island per seed (with eventual 95.7% probability) and then stop there (Fig. 6c-e). Thus, the absence of high-valency products in Fig. 2a is strong evidence against the notion of random nucleation. For growing the second Ag island and making Au(Ag) 2 trimers, it appears more probable (Fig. 3d) to achieve the :Ag-Au-Ag angle < 120 (75% probability) than >120 (25%). This is proven wrong by the experiments, where the latter structure ($120 ) was adopted by over 57% of the Au(Ag) 2 products (Fig. 3a and c). For the above theories of "repulsion" and "random nucleation", it is critical to fnd the detailed steps linking those mechanisms to the observations. For example, how exactly are the different degrees of repulsion affected by the surface ligand density or the rate of Ag reduction; and how those factors can control the valency of the fnal product. However, no plausible links could be found. After repeated experiments, we realized that both the surface ligand density and the rate of Ag reduction would lead to increased oversaturation during the initial nucleation stage, thus providing a possible link to a plausible mechanism. Specifcally, a dense layer of ligand would force the reduced Ag atoms to build up in the solution. With ligands nearly covering all seed surfaces, the increased surface energy 11 is manifested by the gradual decrease of "wetting" of Ag domains on Au, 8 and this would increase the cost of initial nucleation. 25 On the other hand, a faster rate of Ag reduction would supply more growth material to initiate and sustain the higher oversaturation. These arguments explain well the lack of multiisland growth in the presence of no/low ligand concentrations (giving core-shell structures), 8 as the rapid growth on the seeds would consume the growth material (Ag atoms) preventing their build-up. In order for heterogeneous nucleation to occur on the ligandcovered Au seed, the oversaturation of the Ag 0 in the solution must surpass a certain threshold (C hetero ), 25 which is in general lower than the threshold for homogeneous nucleation (C homo , Fig. 7). Basically, the random collision of Ag 0 atoms must reach a nucleus of critical size. With partial bonding with the seed surface, the critical nucleus can be smaller on the seed than that formed in middle of the solution. 11 On these bases, we speculate that a new island cannot occur too closely to an old island because the growth at the existing site would deplete the growth materials nearby, prohibiting new nucleation sites. Previously, the concept of a "depleted region" was invoked to explain the controlled growth of CaCO 3 crystals on a flat substrate with patterned self-assembled monolayer (SAM). 26 At steady-state, the depleted radius was estimated to be about 80-100 mm, about 4 orders of magnitude larger than the short-distance (10-20 nm) nucleation in this work. In comparison, the Ag deposition on the colloidal Au seeds occurs in (c-e) schematics illustrating the distribution of nucleation sites (blue dots) under random nucleation, after 3 rounds of nucleation with (c) P ¼ 0.3; (d) P ¼ 0.6; and (e) P ¼ 0.9. As Janus nanoparticles build up in concentration, random nucleation would start to make Au(Ag) 2 trimers, so on and so forth. It would be impossible to achieve highpurity Au-Ag Janus nanoparticles regardless of the P value. This journal is © The Royal Society of Chemistry 2017 Chem. Sci., 2017, 8, 430-436 | 433 a 3-dimensional space, and the seeds undergo constant Brownian motion. The seed concentration is estimated to be 5.66 10 11 M, that is, each seed occupies an average space of about 3 3 3 mm 3 . The fact that pure Ag nanoparticles were not formed suggested that, at some point, the depleted region (with C homo as boundary) must have extended over the average space and inhibited homogeneous nucleation. This, however, cannot explain the control of heterogeneous nucleation at short distance. It is also clear that the depleted region at such a large radius must be severely disrupted by the rapid Brownian motion of the seeds. Hence, we propose a depletion sphere model to explain the effects of a non-steady-state concentration gradient at the nanometer range, which occurs faster than the time-scale of Brownian motion. More specifcally, the initial nucleation on a seed and the subsequent growth are expected to partially deplete the local Ag 0 atoms, leading to a concentration gradient (Fig. 7a). The chemical reaction producing the Ag 0 atoms should occur homogeneously throughout the solution (3 3 3 mm 3 per seed), but their depletion only occurs at the growth site on the seeds (70 nm). Thus, the chemical reaction at the vicinity of the seed is clearly insufficient to sustain the Ag growth. With the concentration gradient, a spherical boundary can be defned as the depletion sphere, within which the oversaturation is too low (<C hetero ) to achieve heterogeneous nucleation. Thus, if the depletion sphere does not fully cover the seed surface, a second nucleation site could occur, and the probability should depend on the exposed area and the degree of oversaturation (Fig. 1). The size of the depletion sphere relative to the size of seed (Fig. 5) determines the number of nucleation sites, and thus, the number of islands after the growth. Once a Ag nucleus forms on the seed surface, the absence of ligands on the "fresh" surface would make it greatly more favorable for Ag 0 deposition than the "old" seed surface with dense ligands. 8 As previously discussed, the dynamic ligand readsorption has to compete with the Ag deposition in order to inhibit the "fresh" surface. 15 Hence, the "ligand inhibition" in this work refers to a dynamic and competitive inhibition, rather than a static and absolute inhibition. In an ideal model, efficient growth at the Ag island would deplete the Ag 0 atoms at their immediate vicinity, reaching saturation (C s ), i.e., it is a perfect "sink" and the diffusion does not cause build up of Ag 0 . The depleted region rapidly expands, creating a concentration gradient at the radial direction. The diffusion can be described by the classic Fick's laws, 27 where the change of concentration with time is proportional to the second derivative of the concentration (eqn (1)) and D is the diffusion coefficient. To simplify, eqn (1) can be expressed at the radial direction as eqn (2). In an ideal model, we can set a constant C bulk . Solving the equation leads to the Gauss error function with a sigmoid shape concentration gradient (Fig. 6b), which depends on the time and boundary conditions. A few basic conclusions can be drawn: (1) the depletion sphere as defned by the concentration gradient would expand with time; (2) at the steady state (vC/vt ¼ 0), a constant gradient would be established from the origin to the bulk solution; (3) the higher the initial oversaturation is, the steeper the concentration gradient becomes, and thus, the depletion sphere defned by C hetero is smaller in radius (Fig. 6b, C bulk 2 > C bulk 1 , thus r 2 < r 1 ). This simplifed model of diffusion has to unite with the classic model of nucleation and growth, to reflect the drastic change of concentration during the critical stage of the initial nucleation (i.e., the non-steady-state concentration gradient). Fig. 7b depicts the concentration trace across the radius: before heterogeneous nucleation, the Ag 0 concentration at all locations is forced to build up, surpassing C hetero . After nucleation, the vicinity (location 1) is most rapidly depleted and the farther The initial stage of Ag 0 build-up is the same for all locations. Once a nucleation site initiates, its vicinity (pink) is depleted faster than the locations further away, generating a concentration gradient (the red line on the vertical plane). Within the depletion sphere, the Ag 0 concentration is below C hetero (the grey transparent plane), whereas it is above the plane outside the depletion sphere, but still below C homo . In other words, heterogeneous nucleation can occur outside the sphere (requires seed surface) but no homogeneous nucleation. locations are gradually depleted due to the diffusion across the concentration gradient. Only the locations 6 and 7 outside the depletion sphere are able to maintain a higher concentration above C hetero for a sizable period. In order to explain the inhibition at the nanometer range, the critical point of choice should occur during the initial expansion of depletion sphere, rather than after establishing steady-state concentration gradients (slower than the time-scale of Brownian motion). At the outskirts of the depletion sphere, the probability of the second nucleation is not a constant. It has to be "integrated", taking into consideration both the changing size of the "exposed" area and the changing degrees of oversaturation (Fig. 7b). The probability of nucleation should increase with the degree of oversaturation (at some point it would approach 1), but the detailed dependence is still unknown in the literature. It should be also noted that C bulk is affected not only by diffusion (depletion) but also by the chemical reaction (supply) which varies with time. For making Janus nanoparticles, the lower oversaturation means that the depletion sphere of the frst nucleation has time to expand given the lower probability of the second nucleation. For multi-island (>3) satellite structures, there is a sequence of nucleation sites and depletion spheres. The steeper concentration gradient and higher nucleation probability would cause larger variance in terms of total nucleation sites per seed. This explains the broadening distribution of island number in Fig. 2a-f and 4a-h. ## Conclusions With this growth/diffusion model, we provide detailed steps linking the initial conditions (incubation time and rate of reduction) to the valency of the resulting nanostructures: both the higher ligand density on the seed and faster Ag reduction rate lead to higher initial Ag 0 oversaturation in the bulk (C bulk ), which in turn leads to a steeper concentration gradient and higher probability of nucleation. The resulting smaller depletion sphere thus leads to more Ag islands (Fig. 1). Most importantly, the inhibitive role of the depletion sphere can be broadly applied to explain the lack of new nucleation sites during the growth of Janus nanoparticles (i.e., the resulting high purity) and the mysterious spacing among the islands of satellite nanostructures. It is conceivable that all nucleation processes would deplete their immediate vicinity and inhibit new nucleation within the radius. Recognizing this critical inhibitive role is, in our opinion, essential for rational design of colloidal syntheses. Rather than identifying a critical growth factor, we endeavour to give a complete mechanistic scenario consistent with fundamental growth principles. Admittedly, this opens up a broad frontier, many aspects of which are still unknown in the literature. Nevertheless, we believe that the new synthetic know-how and the mechanistic insights in this work would contribute to the advance of nanosynthesis.

chemsum

{"title": "Depletion sphere: Explaining the number of Ag islands on Au nanoparticles", "journal": "Royal Society of Chemistry (RSC)"}

micronutrients_decline_under_long-term_tillage_and_nitrogen_fertilization

## Abstract: tillage and nitrogen (n) fertilization can be expected to alter micronutrient dynamics in the soil and in plants over time. However, quantitative information regarding the effects of tillage and N application rates on micronutrient dynamics is limited. The objectives of this study were (a) to determine the longterm effect of different tillage methods as well as variation in N application rates on the distribution of Mehlich III extractable manganese, copper, zinc, boron, and iron in soils and (b) to assess accumulation of the same nutrients in wheat (Triticum aestivum L.) tissues. The system studied was under a dryland winter wheat-fallow (WW-F) rotation. Tillage methods included moldboard (MP), disk (DP) and sweep (SW), and the N application rates were 0, 45, 90, 135, and 180 kg ha −1 . The concentration of soil manganese was greater under DP (131 mg kg −1 ) than under MP (111 mg kg −1 ). Inorganic N application reduced extractable soil copper while, it increased manganese accumulation in wheat grain over time. Comparison of micronutrients with adjacent long-term (since 1931) undisturbed grass pasture revealed that the WW-F plots had lost at least 43% and 53% of extractable zinc and copper, respectively, after 75 years of N fertilization and tillage. The results indicate that DP and inorganic N application could reduce the rate of micronutrient decline in soil and winter wheat grain over time compared to Mp and no n fertilization.Nitrogen fertilization plays a significant role in the dynamics of soil organic matter (SOM). Most of the micronutrients are largely SOM bound and will be released when SOM decomposition is stimulated 1,2 . The decomposition of SOM is stimulated by tillage through changes in soil water, aeration, temperature, and nutritional environment 3,4 . No-tillage or reduced tillage accumulates SOM in the upper surface whereas SOM are uniformly mixed to a plow depth under a conventional tillage. The stratification of SOM can lead to varying distribution of micronutrients in soil profile and mislead the farmers on determining the optimum fertilizer application rate 5 . Therefore, understanding the role of tillage and N fertilization in the availability and distribution of micronutrients is crucial in formulating and developing cropping system strategies for sustainable agriculture.Micronutrient availability in cultivated plots is affected by tillage methods 6 . It has been reported that even a slight soil disturbance or tillage increases chemical and microbial activity that enhances nutrient release via mineralization of OM 7 . There have been inconsistencies in research reporting tillage effects on the concentration of extractable iron (Fe), zinc (Zn), copper (Cu), and manganese (Mn). Mahler 6 observed higher extractable Fe and Mn under conventional and reduced tillage than under no-tillage in the soils of northern Idaho, whereas other researchers reported the opposite 1,8,9 . Lavado et al. 5 and Hickman 10 reported concentrations of extractable soil Cu and Zn were unaffected by tillage. However, Shuman and Hargrove 11 observed lower Mn and Fe under no-tillage and reduced tillage than under conventional tillage due to a shift in exchangeable forms of Mn and Fe from inorganic to organic forms. Under reduced or no-tillage, the availability of some micronutrients increases and they appear in more readily available forms in the surface soils due to metal complexation by OM 12 . Additionally, OM increases microbial exudates which have been reported to enhance micronutrient availability to plants, especially Fe 13 , Cu, Mn, and Zn 14 . A similar effect of OM on B has been reported by Sarkar et al. 15 . Another key factor for micronutrient availability in plants and soils is N fertilization. The application of N fertilizers is largely based on crop demands while the input of micronutrients, which are significantly impacted by N fertilization rates, is less common. It has been reported that increasing N supply can enhance accumulation of Zn and Fe in wheat grain 16 . High N supply increases transporter proteins and nitrogenous chelators involved in the uptake, translocation, remobilization and grain allocation of Fe and Zn, and hence, increases the Zn and Fe in wheat grain 17 . However, Cakmak et al. 18 reported a decline in Zn and Fe in wheat grain with high N fertilization rates. Inconsistency in results from varying N application rates on soil micronutrient dynamics were also reported previously. A study by Wang et al. 17 indicated that N fertilization significantly increased the availability of Cu, Mn, and Fe and attributed it to the nitrification derived acidity. Contrastingly, Malhi et al. 19 reported that high N fertilization rate decreased the concentrations of extractable Cu and Zn and suggested further investigation is required to determine the cause. Information on the dynamics of plant essential micronutrients as a function of tillage and N application rates is limited, inconsistent, and region specific. Therefore, there is a need to examine the impact of N application rates and tillage methods on the concentration of micronutrients in soil and plant. This study was undertaken with the objective to investigate the long-term (75 years) effects of tillage and N fertilization rates on Mn, Cu, B, Fe, and Zn in soil, and wheat grain and straw under dryland WW-F rotation in the Pacific Northwest (PNW). The conceptual approach consisted of analyzing the soil (four depths: 0-10, 10-20, 20-30, and 30-60 cm) and wheat (grain and straw) samples of 1995, 2005 and 2015 for Mn, Cu, B, Fe, and Zn, and comparing the analyzed micronutrients among the treatments to determine the 20-years trend of micronutrient dynamics. Additionally, we compared the soil micronutrients of WW-F plots with that of nearby long-term (since 1931) undisturbed grass pasture (GP) plots to detect the tillage and N fertilization induced changes in soil micronutrients during 75 years. We hypothesized that: (i) The concentrations of extractable micronutrients are greater under conservation tillage [disk plow (DP) and sweep (SW)] than under conventional tillage [moldboard plow (MP)] over time. The basis of this hypothesis is that the greater amount of crop residue left at the surface due to reduced soil disturbance under SW and DP will allow accumulation of more SOM than under MP, noting soil OM is a source of plant essential micronutrients, and; (ii) The plots with high N application rates will have a greater concentration of micronutrients than in the zero or low N application rate plots over time. This hypothesis is based on the well-established fact that the N increases organic matter through increased crop root biomass and releases micronutrients in soil upon the decomposition of OM. Furthermore, N fertilization increases soil acidity and the availability of micronutrients in soil increase under acidic conditions. ## Results Only the significantly (p ≤ 0.05) affected micronutrients are reported in the text below. Tillage effect on soil micronutrients. The concentration of Mehlich III extractable Mn was significantly (P < 0.01) affected by the tillage method × soil depth interaction (Table 1). The DP had greater extractable Mn (131 mg kg −1 ) than under MP (111 mg kg −1 ) while no significant differences in extractable Mn were found between SW (121 mg kg −1 ), DP or MP at the 0-10 cm soil depth (Fig. 1). In the 20-30 cm depth, the MP had similar extractable Mn (84 mg kg −1 ) to that under DP (73 mg kg −1 ) but had greater extractable Mn than under SW (68 mg kg −1 ) (Fig. 1). Extractable Mn under DP and SW significantly declined with soil depth, while no significant decline in extractable Mn was observed under MP beyond 10-20 cm (Fig. 1). The MP had greater extractable Cu than under DP (1.13 mg kg −1 vs. 0.79 mg kg −1 ) in the 0-10 cm soil depth, and extractable Cu increased with soil depth under all tillage systems (Fig. 1). Concentrations of extractable Cu under DP were 0.79, 1.63, 2.06, and 2.35 mg kg −1 at the 0-10, 10-20, 20-30, and 30-60 cm soil depths, respectively (Fig. 1). The concentration of extractable Zn was also affected by the tillage methods; Zn was significantly greater under DP (1.92 mg kg −1 ) than under SW (1.38 mg kg −1 ) while it was comparable to Zn under MP (1.56 mg kg −1 ) (Fig. 2). Nitrogen fertilization effect on soil micronutrients. In this study, only extractable Cu was found to be affected by N fertilization rates. Extractable Cu was significantly greater without N application and declined with the application of N fertilizer (Fig. 2). A three-way interaction of N rate, tillage and year were observed for extractable Zn, Fe, and B in this study (Table 2). However, we did not observe any consistent trend of micronutrients change over the time. www.nature.com/scientificreports www.nature.com/scientificreports/ Soil micronutrients after 75 years of N fertilization and tillage versus grass pasture (GP). Since DP was the best tillage in maintaining micronutrients compared to the other tillage treatments in this study, we compared the soil micronutrients of DP plots with that of GP plots, the reference/baseline of this study, to detect the treatment's effect over 75 years. Compared with the nutrients in the GP plots, extractable Zn and Cu concentrations in the cultivated plots declined more than other nutrients. Extractable Zn decreased by at least 43%, and Cu decreased by 53% in the top 10 cm soil depth compared with their respective concentrations in the GP plots (Table 3). The concentrations of Mn and Zn at the 20-30 cm and 30-60 cm soil depths were comparable to those of GP. At the 20-30 cm soil depth, N fertilization contributed to a decrease in extractable Cu when compared with GP. Similarly, a pronounced decrease in extractable Cu was found at the 30-60 cm soil depth of DP plots that received N fertilizer above 45 kg ha −1 . Iron was not tested in the soils of GP, and B was detected only at the 20-30 cm depth in GP. Effect of tillage and N fertilization on micronutrients in wheat grain and straw. The total Mn concentration in wheat straw was largely influenced by N application rate, whereas Cu, Fe, and B were affected by the interaction of tillage systems and year (Table 4). The Mn in straw increased linearly with increasing N application rates (Fig. 3). The concentrations of Mn in straw were 27, 32, 44, 47, and 50 mg kg −1 at the 0, 45, 90, 135, and 180 kg N ha −1 , respectively. However, the concentrations of Cu, Fe, and B in the straw declined over the 20-year period (1995-2015) under all the tillage systems (Table 5). Except for the concentration of Mn, none of the micronutrients in wheat grain were affected by the treatments (Table 4). The Mn in grain increased with increasing N fertilizer rate up to 135 kg N ha −1 (Fig. 3). The concentrations of Mn in grain were 41, 44, 48, 54, and 52 mg kg −1 at the 0, 45, 90, 135, and 180 kg N ha −1 fertilization rates, respectively. ## Discussion Tillage and N fertilization effect on soil micronutrients. Extractable Mn declined with depth. It is well documented that the availability of soil micronutrients is associated with SOM 1,6 . Our results agree with previous studies, which also reported increased concentrations of extractable Mn in the upper 10 cm soil depth under tillage that promoted accumulation of plant residues at the soil surface and resulted in poor soil mixing 1,20 . In this study, the DP and SW had lower volumes of soil mixing and left more residue on the soil surface compared to MP. Depth of soil disturbance was low under DP (10 cm), and SW (15 cm) compared to MP (23 cm) and consequently differed in the percentage of residue cover (OM) left on the top 10 cm soil. The Mehlich III extractable Cu in the top 10 cm soil depth was higher under MP than under reduced tillage, which is in agreement with earlier studies 5,9 . Contrastingly, Mahler et al. 6 found lower concentration of extractable Cu under MP than under reduced tillage, while Edwards et al. 8 did not find a significant effect of tillage on soil Cu. Franzluebbers and Hons 9 also reported increased Cu concentration until 30 cm depth whereas, in this study, extractable Cu increased beyond the 30 cm soil depth. Soil profile distribution of Cu can be explained by its interaction with SOM. Copper is bound to SOM and migrates into subsoil with SOM acting as a carrier, forming soluble metal-organic complexes 21 3. Impact of 75 years of inorganic N application rate (N rate) on soil micronutrients and soil pH of dryland winter wheat-fallow cropping system under disk tillage management compared to nearby undisturbed grass pasture (GP). † Means sharing the same letter within the rows are not significantly different at 5% level of significance. ‡ Percentage calculated from the difference in the value of grass pasture (GP) and the highest value (if GP is greater) or the lowest value (if GP is lower) for the treatments within each soil depths, so that minimum deviation from the GP is calculated in either case. The downward and upward arrow indicates decline or incline from the soils of GP after cultivation, respectively. The column with both upward and downward in the same cell indicates that respective soil depth has some treatments that have greater value than GP and some treatments with lesser value than GP. www.nature.com/scientificreports www.nature.com/scientificreports/ and Fe levels in soils 22 . The Fe and Al oxides and oxyhydroxides adsorb Cu tightly and consequently reduce the mobility of Cu in fertilized soils 22 . ## Soil micronutrients after 75 years of N fertilization and tillage versus grass pasture (GP). Extractable Mn, Zn, and Cu and soil pH declined significantly after 75 years of N fertilization in the upper 10 cm soil depth at all tested N rates (Table 3). It is well-documented that N fertilization lowers soil pH, enhancing the availability of micronutrients 23,24 . It was evident in our study that the soil pH had decreased after 75 years of cultivation (Table 3) and significantly decreased in upper 20 cm soil surface (reported in another manuscript from the same experiment but with macronutrients 25 ). However, this acidification did not increase micronutrient availability over the study period, suggesting that continuous removal through crop harvest and meager contributions from crop residue had depleted micronutrients in the soil. The other likely reason for the significant decline of extractable Mn, Cu, and Fe in the upper 10 cm soil would be due to the presence of a higher percentage of OM (crop residue) in the upper 10 cm soil than deeper in the soil profile. The availability of these nutrients in the soil solution decreases with higher OM, as these elements have a high affinity for OM resulting in stable bonding 24 . Effect of tillage and N fertilization on micronutrients in wheat grain and straw. Inorganic N fertilization increased the concentration of total Mn in wheat grain up to 135 kg N ha −1 application rate (Fig. 3). In contrast with these results, Hamnér et al. 26 reported that N fertilization did not influence grain Mn in their study; however, they found increased concentrations of Fe, Zn, and Cu in the wheat grain as a function N fertilization. Table 5. Interaction effect of tillage system and year on total concentrations of copper, iron, and boron accumulation in wheat straw. † Means followed by the same uppercase letter in a row indicate no significant differences between years for each tillage system and means followed by same lowercase letters in a column indicates no significant differences between tillage system in each year at 0.05 probability level. (2019) 9:12020 | https://doi.org/10.1038/s41598-019-48408-6 www.nature.com/scientificreports www.nature.com/scientificreports/ The relationship between N fertilization and micronutrients is unclear, but previous studies have indicated a correlation of N to the movement of micronutrients within plants 27,28 . ## conclusion The findings of this study are significant for a sustainable dryland winter wheat-fallow cropping system. The results provide important insight into the impact of long-term tillage and inorganic N fertilization (75-years) on the distribution of micronutrients (Mn, Cu, Fe, Zn, and B) in soil and wheat. The study demonstrated the declining trend in the concentrations of extractable Mn, Cu, and Zn in cultivated soil (cultivation effect) when compared to the undisturbed grass pasture plot. It is evident that continuous cultivation with N fertilization and tillage may significantly reduce concentrations of plant essential nutrients over time. We found that disk plow tillage and high N application rates were better than other treatments studied. However, nitrification derived acidity must be considered and should be regularly monitored. Integration of organic amendments and inorganic nitrogen fertilizer application in nutrient management strategy may help to increase micronutrients in soil and wheat in a long-term as organic amendments are known to enhance nitrogen and micronutrients availability without acidifying the soil. A long-term study is needed to warrant the benefits of integrating organic amendments and inorganic N on micronutrients availability over time in the drylands of the PNW. ## Materials and Methods Study sites and experimental design. The study was conducted at one of the ongoing long-term experiments (LTE) of the Columbia Basin Agriculture Research Center (CBARC), near Pendleton, OR (45°42′N, 118°36′W, elev. 438 m.a.s.l.). This LTE was established in 1940 on a a well-drained Walla Walla silt loam soil (coarse-silty, mixed, superactive, mesic Typic Haploxeroll) with a 2-4% slope. The mean annual temperature is 10 °C, and ranges from −1 °C in January to 21 °C in July. Mean annual precipitation is 437 mm. The top 30 cm soil depth contains 20% clay, 68% silt, and 1.1% organic C, and has 16 cmolc kg −1 cation exchange capacity (CEC). The experimental plot is a randomized block, split-plot tillage and fertility experiment with three replications under dryland winter wheat-14 months fallow (WW-F) cropping system. Each block was divided into three main plots as tillage treatments and each main plot was divided into five subplots as N fertilization treatments. The three tillage treatments were moldboard plow (MP), disk plow (DP) and sweep (SW) with the size of 35 by 40 m each. Subplots comprised of five N fertilization rates (0, 45, 90, 135, and 180 kg N ha −1 ) and were 5.8 by 40 m in size. During late March to early April, primary tillage was performed in the fallow plots on the stubble left undisturbed since wheat harvest. The three tillage treatments differed in tillage equipment, surface residue cover at the time of seeding, and tillage depth. The percentage of residue cover left by MP, SW, and DP were 7%, 43%, and 34% respectively, and the tillage depths were 23 cm, 15 cm, and 10 cm, respectively. The MP is a soil inversive tillage whereas DP and SW are non-soil inversive tillage. Therefore, the MP is considered conventional tillage, and the SW and DP are considered reduced/conservation tillage in this study. A nearby grass pasture (GP) plot, undisturbed since 1931, was used as reference/baseline for this study to compare changes in treatments over time. The dominant grasses in this pasture are blue-bunch wheatgrass (Agropyron spicatum L. Pursh) and Idaho fescue (Festuca idahoensis L. Elmer). Field operations and soil sampling. After wheat harvesting in late July, the stubble was left undisturbed until primary tillage operations in late March. Plots were rod weeded two to four times between April and October to control weeds. During the first week of October, urea ammonium nitrate fertilizer was added to the top 10 cm soil using Viper Coulter (Yetter Manufacturing Inc. Colchester, IL). A week after N fertilization, wheat was seeded at the rate of 72 ± 5 kg seed ha −1 in 25 cm rows spacing. A JD8300 drill (Deere and Company, Moline, IL) was used for wheat seeding before 2002, and thereafter a Case IH 5300 disk drill (Klamath Basin Eq. Inc. Klamath Falls, OR) was used. The seed variety was Malcolm during the 1995-2005 period, and Stephens after that. Both were semi-dwarf varieties of winter wheat. Weeds were controlled using herbicides during the growing season. The soils were sampled by compositing the cores of north-central and south-central of each plot. Wheat grain and straw samples were collected from the center of the plot after the wheat harvest. The soils were sampled from four depths (0-10, 10-20, 20-30, and 30-60 cm) using a truck-mounted Giddings Hydraulic Probe (Giddings Machine Company, Inc., Windsor, CO) and a steel sampling tube (internal diameter 3.6 cm). In this study, the soil and plant samples from 1995 (archived samples), 2005 (archived samples) and 2015 cropping season were used. The ground soil samples were processed and analyzed at the Central Analytical Laboratory (CAL, Oregon State University). The Mehlich III method 29 was used to extract available Mn, Cu, Fe, B, and Zn from the soil samples, and a dry ash method 30 was used to extract the total concentration of these nutrients from the grain and straw samples. An inductively coupled plasma-optical emissions spectroscopy (ICP-OES, Model #2100 DV, Waltham, Massachusetts, USA) was used to determine the nutrients in soil and plant tissue extracts. Soil pH data were provided by the CBARC, and were determined with a pH electrode using 10 g samples in a 1:2 soil to 0.01 M CaCl 2 solution. ## Statistical analysis. A split-plot design analysis was used to test the effect of the treatments on the concentration of Mn, Cu, Fe, B, and Zn using the mixed model procedure in JMP © version 13 31 . Tillage system, N rates, and soil depths were considered the fixed effects while analyzing soil micronutrients. We didn't observe significant differences in soil micronutrients as a function of year and its interaction, therefore the analysis was done using the 2015 data only. Tillage system, N rates, and year were considered the fixed effects for tissue analysis. Replications and their interactions were considered the random effects in both the soil and tissue analysis. Multiple comparisons with Tukey methods were performed to determine differences in nutrients and letter groupings were generated using a 5% level of significance. (2019) 9:12020 | https://doi.org/10.1038/s41598-019-48408-6 www.nature.com/scientificreports www.nature.com/scientificreports/ Soil pH data were converted to H + concentration (µmol L -1 ) before ANOVA was performed. The pH scale is a logarithmic and small differences in pH represent large differences. However, the mean comparisons of soil pH represent the original pH data.

chemsum

{"title": "Micronutrients decline under long-term tillage and nitrogen fertilization", "journal": "Scientific Reports - Nature"}

discrete_cu(<scp>i</scp>)_complexes_for_azide–alkyne_annulations_of_small_molecules_inside_mammalian

## Abstract: The archetype reaction of "click" chemistry, namely, the copper-promoted azide-alkyne cycloaddition (CuAAC), has found an impressive number of applications in biological chemistry. However, methods for promoting intermolecular annulations of exogenous, small azides and alkynes in the complex interior of mammalian cells, are essentially unknown. Herein we demonstrate that isolated, well-defined copper(I)tris(triazolyl) complexes featuring designed ligands can readily enter mammalian cells and promote intracellular CuAAC annulations of small, freely diffusible molecules. In addition to simplifying protocols and avoiding the addition of "non-innocent" reductants, the use of these premade copper complexes leads to more efficient processes than with the alternative, in situ made copper species prepared from Cu(II) sources, tris(triazole) ligands and sodium ascorbate. Under the reaction conditions, the well-defined copper complexes exhibit very good cell penetration properties, and do not present significant toxicities. ## Introduction Organometallic catalysis has changed the feld of organic synthesis in the last half century, and has found important applications in other areas such as materials, energy or environmental sciences. In spite of such wide impact, the use of transition metal catalysis in biological contexts remains underdeveloped, probably due to the general belief that metalpromoted reactions are incompatible with the air atmospheres and aqueous environments of biological habitats, and that the metal complexes can be highly cytotoxic. 1 Only recently, a few examples demonstrating the viability of achieving transition metal promoted transformations in biological contexts, 2 and even in intracellular environments, 3 have been disclosed. Most of these reports deal with palladium or ruthenium-catalyzed uncaging of designed substrates equipped with inactivating handles. 4 More challenging intracellular metalpromoted coupling reactions involving two different abiotic precursors are much scarcer. This is not surprising, as these reactions require the cell entrance and "meeting" of up to three different partners, namely the metal complex and two exogenous reactants (Fig. 1a). Thus, while several groups have demonstrated the viability of achieving Suzuki or Sonogashira couplings on appropriately modifed proteins in E. coli, 5 the only two examples described so far in mammalian cells involve the use of palladium nanoparticles, and fxed cells. Curiously, the well-known copper(I) catalyzed azide-alkyne cycloaddition (CuAAC) reaction, 7 has been very scarcely explored inside the complex environment of living mammalian cells. Thus, while early work on the use of this reaction for biological purposes was restricted to bacteria, 8 most of the other "in vivo" applications have been limited to the modifcation of cell surface labelled glycans. 9 Probably, the established notion that copper is highly cytotoxic, and the requirement of excess of several additives, including "non-innocent" ascorbate, have precluded further research to implement the reaction in the challenging and crowded atmosphere of mammalian cells. 9d,10 In recent years, several groups have developed water soluble ligands for Cu(I) that accelerate the reaction and also act as sacrifcial scavengers/reductants of the reactive oxygen species (ROS) generated by copper, decreasing its cellular toxicity. 11 The accelerating effect can be further improved if the copper chelated ligand is covalently linked to the azide, a tactic that has even been used for the intracellular labelling of proteins. However, with this strategy, the copper complex is likely sequestered by the products, owing to the presence of the triazole and the copper chelating moiety. 12 Water soluble copper ligands linked to a cell penetrating peptide have been recently used to promote click reactions inside cells, but with low efficiency, and only for alkyne-modifed proteins. 10b If the modifcation of intracellular proteins is a highly relevant, and far from trivial, goal, achieving intracellular copper promoted reactions between two "freely diffusing small molecules" is even more challenging. Acceding to this type of reactivity can open new, exciting opportunities for biological or metabolic intervention, and for a metal-dependent generation of active drugs or optical signals. To the best of our knowledge, the only example of such type of intracellular CuAAC reaction relies on the use of cross-linked copper containing polymers termed metalorganic nanoparticles (MONPs), and requires high concentration of sodium ascorbate. 13 Another class of copper nanostructures that can also promote the reaction in water has been recently reported, however their activity is confned to the extracellular milieu. 14 Herein we report the frst examples of an intracellular CuAAC transformation involving two exogenous, freely spreading substrates (small molecule azide and alkyne), promoted by discrete Cu(I) complexes (Fig. 1b). We also present data on the compared reactivity, redox stability, cell uptake and toxicity of in situ made copper species versus Cu(I) predefned complexes. These studies allow for the discovery of an independently isolated, well-defned Cu(I) complex equipped with the BTTE ligand (3-4-{{bis{[1-(1,1-dimethylethyl)-1H-1,2,3-triazol-4-yl] methyl}amino}methyl}-1H-1,2,3-triazole-1-ethanol, L3), which performs much better than the in situ mixture obtained from the ligand, a Cu(II) source and sodium ascorbate. ## Results and discussion Our work was conceived on the hypothesis that designed, well-defned Cu(I) complexes might cross cell membranes and keep their oxidation +1 state under the reductive atmosphere of the cell. Thus we proposed to study the intracellular reactivity of tris-triazolyl-Cu(I) complexes generated in situ by reduction of Cu(II) precursors with ascorbate, as well as of isolated, well-defned Cu(I) complexes (Chart 1). 15 As substrates we chose fluorogenic azides that undergo an increase in fluorescent emission upon annulation with the corresponding alkynes. 16 The most habitual azide substrate for these purposes is 3-azido-7-hydroxycoumarin, however, in our hands, preliminary control tests with HeLa cells indicated that this azide presents a substantial background signal. We therefore moved to the 9-(azidomethyl)anthracene (1, Fig. 1b) which is almost non-fluorescent, but undergoes a ca. 150-fold increase in fluorescence upon its annulation with alkynes (Fig. S2 †). This increment can be explained in terms of suppression of the internal PET (photoinduced electron transfer) quenching on moving from the azide to the triazole structure. 16,17 As tris(triazolylmethyl)amine ligands we selected BTTAA (3- ), which has been shown to be rather effective in CuAAC in aqueous media, and even in E. coli. 9c We also prepared the analogues BBTE (L2) and BTTE (L3), which feature a hydroxyl group susceptible of conjugation to different units. Indeed, we synthesized the derivative L4 which contains a triphenylphosphonium moiety designed to favor cellular internalizations and, eventually, mitochondrial localizations (Chart 1a). As predefned Cu(I) catalysts, we initially aimed to explore several previously characterized species such as pyrazolyl, NHC (N-heterocyclic carbene) phosphite or phosphinite copper complexes (C1-C4, Chart 1b). It is surprising that the catalytic activity of this type of well-defned Cu(I) complexes had never been explored in biorelevant settings. Before moving to cellular environments, we investigated the performance of the above complexes in aqueous media. With ligands L1-L4, the catalytic reactions were carried out using 75 mol% of copper, by mixing CuSO 4 with 2 equiv. of the ligand in water (with 2% DMSO) at room temperature for 10 min, and adding the solution to either water or PBS (phosphate buffered solution) mixtures of anthracenyl azide 1 (100 mM) and propargyl alcohol 2 (200 mM), followed by sodium ascorbate (NaAsc, over 30 equiv.). For comparison purposes, we analyzed the conversion after 10 and 20 min, by using calibration curves (see Fig. 2 and Section S4 in the ESI †). In the absence of ligands, i.e., when only CuSO 4 and sodium ascorbate are employed, the reaction proceeds with poor yields (<10% even after 24 h, Fig. 2a, dark blue bars). However, with ligand L3 the product was obtained in 32% yield in water and 22% in PBS, after 10 min, while L2 was less effective. Notably, using the phosphonium containing ligand L4 we observed 50% of the triazole after 10 min, in both water and PBS (phosphatebuffered saline, Fig. 2a, purple bars), and a very good 70% yield after 20 min. As expected, if we skip the pre-treatment of the Cu(II) complexes with sodium ascorbate, there is no reaction. UV-Vis and 1 H-NMR analysis confrmed that mixing CuSO 4 , the ligand and sodium ascorbate generates a tris(triazole) Cu(I) species (Fig. S9 and S10 †). The performance of the predefned, isolated Cu(I) complexes C1-C4 (Chart 1b) was also assessed in the absence or presence of ascorbate, at 37 C (20 min, Fig. 2b). The carbene complex C3 is almost inactive, and the phosphite and phosphinite complexes C2 and C4 also led to very poor conversions (less than 2% of the product). With the complex [Cu(NCMe)(Tpa*)] [PF 6 ] (C1) 15a,b the reaction was slightly more efficient (13% in water and 5% in PBS). Importantly, addition of sodium ascorbate allowed much better conversions, specially, with C2 and C3. These results suggest that under the reaction conditions (open air flask), the Cu(I) species are readily oxidized, something that was further confrmed by EPR. Therefore, while C1 and C2 are stable in solid state, in DMSO they are very rapidly oxidized under air to give paramagnetic Cu(II) species (Fig. S11 and S12 †). Overall, the best conversions were achieved with the in situ made copper complexes in presence of ligand L4. Indeed, using this ligand it was possible to obtain the product in a satisfactory 46% yield, after 20 min, using just 25 mM of the copper source (Fig. 2c). 18 With the above information in hand, we moved to living mammalian cells using two different cell lines: HeLa and A549 (living human cervical cancer cells and adenocarcinomic human alveolar basal epithelial cells, respectively). In the experiments with sodium ascorbate, the copper containing mixture added to the cells was prepared by mixing CuSO 4 and the ligand (L) in a 1 : 2 ratio in water for 1 h, followed by treatment with an aqueous solution of sodium ascorbate (6 equiv.) for 30 min. 19 With the defned, discrete Cu(I) species C1-C4, cells were directly incubated with a freshly made DMSO solution of the complexes. The experiments were carried out by mixing cultured cells with the copper solutions (75 mM for in situ made complexes and 50 mM for discrete Cu(I) species) for 30 min in fresh DMEM (Dulbecco's modifed Eagle's medium), followed by two washing steps with DMEM prior to the addition of the reactants. The resulting cells were incubated with the azide 1 (100 mM) and the alkyne 2 (200 mM) in fresh DMEM for 60 min and washed twice with DMEM, before observation under the fluorescence microscope. It is important to note that we do not use cell fxation techniques, which allows for the preservation of the native living environment, and avoids artefacts or over-interpretations. In the experiments with in situ made copper species, in absence of ligands or with L2, we did not detect any intracellular fluorescence, while with L1 and L3 the fluorescent intensity was weak (Fig. S18 †). However, we were glad to observe that when using L4 as ligand, there was a clear blue intracellular fluorescence across the cytoplasm and in vesicles, with the cells showing an unaltered morphology (Fig. 3, panel C, D and E). Control experiments in absence of the copper species (Fig. 3, panel A and B), using the same threshold observation parameters, confrm that the signal must necessarily come from the expected reaction. 20,21 Remarkably, despite their low in vitro activity, the predefned Cu(I) complex C1, the phosphite complex C2 and the phosphinite complex C4 were able to raise some intracellular fluorescence in experiments carried out in the absence of ascorbate, while C3 failed to elicit any fluorescence (Fig. S19 †). Using MTT cytotoxicity assays we observed that more than 90% of the cells survived after 2 h of treatment with the standard Cu(II)/L4/ascorbate mixture, using 75 mM of the copper Chem. Sci., 2018, 9, 1947-1952 | 1949 source (80% survival after 12 h). With C1 the cell survival was slightly lower, reaching values of approx. 80% after 2 h (Fig. S22 †). The reactivity observed with the tris(pyrazolyl) copper species C1, prompted us to pursue the specifc preparation of a well-defned Cu(I) complex equipped with a tris(triazole) ligand. Thus, we focused on the isolation of a complex similar to C1 but containing the ligand L3 or L4. While with L4 we have not yet been successful, we could isolate a Cu(I) complex (C5) by mixing [Cu(NCMe) 4 ][PF 6 ] with equimolar amounts of L3 in methanol, and subsequent precipitation (Section S2 †). EPR monitoring of fresh DMSO solutions of this complex (C5) demonstrated a higher redox stability than C1. Therefore, while in the case of C1, 80% of Cu(I) is oxidized to Cu(II) after 20 min, under the same conditions, less than 30% of C5 was oxidized (Fig. S12 and S13 †). The in vitro performance of complex C5 was quite similar to that of tris(pyrazolylmethane)containing complex C1, however we were pleased to observe that this complex presents an excellent performance in native cellular settings, in the absence of sodium ascorbate (Fig. 4a, panel C and F); much better than that observed when the cells are incubated with the standard pre-made mixture containing Cu(II)/L3/ascorbate (Fig. 4a, panel B and E). To better appreciate the differences in efficiency, we have established a protocol to calculate reaction yields of the intracellular transformations, based on fluorescence measurements using a microplate reader (see Section S13 in the ESI †). The data were normalized with respect to the amount of anthracenyl azide (1, limiting reactant) uptaken by cells. Gratifyingly, when complex C5 was used, the product was obtained in approx. 18% yield, which is over 7 times greater than that obtained using the in situ prepared complex with ligand L3. The intracellular reactivity was also analyzed by flow cytometry, which confrmed that cells treated with C5 presented higher levels of fluorescence when compared with that resulting from the in situ made L3/copper complex. Indeed, C5 performed the best among all the copper species so far studied (Fig. 4b). The use of an extensive washing protocol to remove extracellular copper should assure that the reactions are taking place inside the cells. However, we further confrmed this by observing a total lack of reactivity in control experiments using extracellular media. Furthermore, we also observed that adding copper chelators like EDTA to the extracellular solution, in experiments carried out with living cells, has no effect on the results (see Fig. S20 †). Interestingly, there is a clear correlation between the copper uptake and the observed activity. Therefore, the phosphonium containing ligand L4 promoted a relatively high intracellular accumulation of copper. The ICP-MS analysis also indicates that the well-defned Cu(I) complexes C1 and C2 are very well internalized, which explains why we do observe some intracellular reactivity despite their poor in vitro activity. More important, the copper complex C5 is also very well internalized, leading to almost three times more internal copper than that from the corresponding in situ made copper complex with the ligand L3. Therefore, the better intracellular performance of C5 versus the in situ made mixture of CuSO 4 /L3/NaAsc appears to be, at least in part, associated to an improved internalization. Cell toxicity studies using different concentrations of C5, indicated over 70% viability after 2 h, which is raised to 82% using 25 instead of 50 mM (Fig. 5b and S22 †). If we normalize these values with respect to the amount of copper internalized by cells (ICP data) we can conclude that the toxicity of complex with L3, and complex C5 is similar. An additional control experiment indicated that the intracellular reaction is also feasible with 25 mM of C5, albeit the efficiency is slightly lower (over 9-10% yield, page S31). The above information confrms that the effectivity of an intracellular CuACC with C5 is associated to a good balance between redox stability and catalytic reactivity, and to its improved cell uptake properties, and furthermore confrms the viability of obtaining efficient, well-defned Cu(I) catalysts to be used in complex environments. ## Conclusions We have demonstrated that water soluble copper(I) complexes featuring designed ligands can readily enter mammalian cells and promote intracellular CuAAC annulations of small, abiotic and freely diffusible molecules. Our results indicate that using appropriate ligands, it is possible to tune the cell uptake and reactivity of Cu(I) complexes, and importantly, confrm the viability of using discrete copper species to promote efficient CuAAC annulations in the challenging interior of mammalian cells. Indeed, an independently isolated Cu(I)-tris(triazolylmethyl)amine complex, C5, that can be stored without degradation when kept under nitrogen, is capable of promoting the intracellular transformation even in the absence of ascorbate. This new complex displays better cellular uptake and better intracellular reactivity than that observed for the in situ made Cu(I)/L3 complex. This complex is therefore working as an "off-the-shelve" catalyst to promote challenging intermolecular annulations inside mammalian cells. The copper complex C5 circumvents some of the actual limitations of the "in vivo" CuAAC chemistry, since it avoids the use of excess of ligands or the use of reductants such as ascorbate. Current studies are focused on further improving the ligands to obtain even more effective catalysts that demonstrate negligible toxicity, on the development of copper complexes that can target different cellular organelles and on the use of the complexes for achieving designed biological alterations. ## Conflicts of interest There are no conflicts to declare. This journal is © The Royal Society of Chemistry 2018 Chem. Sci., 2018, 9, 1947-1952 | 1951

chemsum

{"title": "Discrete Cu(<scp>i</scp>) complexes for azide\u2013alkyne annulations of small molecules inside mammalian cells", "journal": "Royal Society of Chemistry (RSC)"}

synthesis_of_structurally-defined_polymeric_glycosylated_phosphoprenols_as_potential_lipopolysacchar

## Abstract: The biosynthesis of lipopolysaccharide (LPS), a key immunomodulatory molecule produced by gramnegative bacteria, has been a topic of long-term interest. To date, the chemical probes used as tools to study LPS biosynthetic pathways have consisted primarily of small fragments of the larger structure (e.g., the O-chain repeating unit). While such compounds have helped to provide significant insight into many aspects of LPS assembly, understanding other aspects will require larger, more complex probes. For example, the molecular interactions between polymeric LPS biosynthetic intermediates and the proteins that transfer them across the inner and outer membrane remain largely unknown. We describe the synthesis of two lipid-linked polysaccharides, containing 11 and 27 monosaccharide residues, that are related to LPS O-chain biosynthesis in Escherichia coli O9a. This work has led not only to multimilligram quantities of two biosynthetic probes, but also provided insights into challenges that must be overcome in the chemical synthesis of structurally-defined polysaccharides. ## Introduction Lipopolysaccharide (LPS), an essential component of the gramnegative bacterial outer membrane, is an important mediator of host-pathogen interactions. 1 LPS has a tripartite structure consisting of lipid A, the core oligosaccharide, and the Opolysaccharide (O-PS). 2 The frst two of these components are semi-conserved across all species while the O-PS is highly variable from organism to organism. For example, more than 180 different Escherichia coli O-antigens are known. 3,4 Structural differences in repeating units, chain length, and noncarbohydrate substituents (e.g., acetylation) make the O-PS one of the most diverse classes of naturally-occurring glycans. 1 Given its biological importance and structural diversity, understanding how LPS is biosynthesized has attracted significant attention. 2,5,6 A common model organism to study this process is E. coli O9a. The structure of the LPS in this organism, with a focus on the O-PS, is shown in Fig. 1A. Key features are a 'primer-adaptor' trisaccharide containing two mannopyranose (Manp) and one N-acetylglucopyranosylamine (GlcpNAc) residues, which links the lipid A-core domain to the O-PS. 10 Four Manp residues, in a mixture of a-(1/2) and a-(1/ 3)-linkages, comprise the repeating unit and each O-PS chain has 9-17 repeating units. 10,11 The structure is terminated with Fig. 1 Structures of (A) E. coli O9a LPS; (B) its phosphopolyprenol biosynthetic precursor; (C) biosynthetic probes synthesized in this paper. The structures in A and B are drawn using the Consortium for Functional Glycomics symbolic nomenclature (green circles ¼ Manp; blue squares ¼ GlcpNAc). 13 . a methyl phosphate group on O-3 of the non-reducing end Manp residue. 7,12 E. coli O9a LPS biosynthesis employs the ABC transporterdependent pathway. 2,10 This process involves the assembly of the full-length O-PS on an undecaprenol (C 55 ) pyrophosphate carrier (Fig. 1B) on the cytoplasmic side of the inner membrane and then its transfer across the inner membrane to the periplasm by an ABC transporter. Ligation of the O-PS to the Lipid A-core domain occurs in the periplasm and then the entire structure is exported across the outer membrane. The assembly of the undecaprenol pyrophosphate polysaccharide intermediate (Fig. 1B) is achieved by the integrated action of four glycosyltransferase (GTs) that install the primer-adaptor and O-PS repeating units, followed by the action of a bifunctional kinase/ methyltransferase (a protein named WbdD) 12 that caps the reactive hydroxyl group on the terminal Manp. This capping process also signals, through an unknown mechanism, the transfer of the intermediate across the inner membrane by the ABC transporter. Like other LPS biosynthetic investigations deciphering the assembly of the O9a O-PS has benefted greatly from the availability of small synthetic fragments of the larger molecule. These compounds have been used to characterise the activities of not only the GTs but also WbdD. However, questions remain about the specifcity of WbdD and proteinsubstrate interactions in the ABC transporter that recognizes and transfers the large lipid-linked glycan across the inner membrane remain obscure. In general, despite impressive recent advances, the 'flipping' of large glycans across membranes remains poorly understood at the molecular level and it is likely that short O-PS fragments will not be effective probes of this process. In addition, the impact of chain length on the action of the biosynthetic GTs is unknown. Answering questions of this type, not only for O9a LPS, but also for other polysaccharides, will require access to larger glycan probes and the development of efficient strategies to assemble them. In this paper, we describe an approach to synthesize two such probe molecules (Fig. 1C), containing either two (1) or six (2) tetrasaccharide E. coli O9a O-PS repeating units connected via the primer-adaptor trisaccharide to farnesyl pyrophosphate. These molecules thus contain 11 and 27 monosaccharide residues, respectively. Over the past several years, outstanding achievements have been made in the synthesis of structurally-defned polysaccharides. However, with some exceptions, most of the structures reported have either been homopolymers containing single glycosidic linkages or polymers with a disaccharide repeating unit. Less common have been syntheses of targets such as 1 and 2, the preparation of which are complicated both by the structure of a tetrasaccharide repeating unit and by the presence of the pyrophosphate and lipid moieties. ## Retrosynthetic analysis and strategy Two possible synthetic routes to 1 and 2 are shown in Scheme 1A. One approach (Route 1) includes a trisaccharide primer-adaptor and a tetrasaccharide repeating unit, which could be obtained from fve different protected Manp building blocks (3-7) and GlcpNAc derivative 8. The second approach (Route 2) involves a tetrasaccharide primer-adaptor, a tetrasaccharide repeating unit, and a trisaccharide cap, which can be assembled using one fewer building block than Route 1: compounds 4-8. Regardless of the route employed, it was necessary to develop an approach to appropriate tetrasaccharide repeating unit building blocks and be able to prepare them efficiently in multigram scale. We selected two targets, p-methoxyphenyl (PMP) glycosides 9 and 10 (Scheme 1B), that could be used in Route 1 or Route 2, respectively. Protecting groups greatly influence glycosylation reactivity, glycosylation stereoselectivity and deprotection efficiency. 27,28 These factors must be balanced with the complexity of the routes needed to assemble (usually monosaccharide) building blocks. Thus, great care was taken in choosing them in this investigation. In particular, an overall goal was to limit the number of benzyl (Bn) ethers as we envisioned their removal on the large target compounds could be challenging. 22 That said, we did select benzyl ethers to protect the majority of the hydroxyl groups on mannose residues A and D in both 9 and 10 to increase reactivities in glycosylation reactions. A benzoyl (Bz) ester was used to protect the C-2 hydroxyl group on residue A to control the selectivity of the a-glycosylation. As the temporary protecting group in residue D, a levulinate (Lev) ester was selected, given the orthogonality of this group to others used in the synthesis. Finally, we chose acetate (Ac) esters to protect the majority of the hydroxyl groups on residues B and C as we expected that they would be easier to remove than Bn ethers. In addition, given their location on the internal portion of the building block, they would not be expected to greatly affect glycosylation reactivity. In cases where hydroxyl groups on residues B and C are benzylated, this was done to simplify the preparation of the monosaccharide precursors. The synthesis of 9 and 10 is discussed below. The preparation of the monosaccharide building blocks needed to synthesize these tetrasaccharides is described in the ESI (Scheme S2 †). ## Synthesis of two repeating units and comparison of Routes 1 and 2 To compare the two routes shown in Scheme 1A, we prepared tetrasaccharide 16 and 21 (Scheme 2), which can be converted to the key building blocks 9 and 10, respectively, by manipulation of the protecting groups. The synthesis of tetrasaccharide 16 (Scheme 2A), started with the coupling of glycosyl acceptor 6 with thioglycoside donor 3 mediated by NIS (1.3 equiv.) and AgOTf (0.1 equiv.). The product was produced in poor (45%) yield, mainly due to the formation of the 1,2-othoester 12 (25%). To improve the yield, a larger amount of AgOTf (0.35 equiv.) was used in the reaction. Under these more acidic conditions, disaccharide 11 was isolated in 65% yield. Chemoselective removal of the levulinate group using hydrazine acetate gave acceptor 13 in 92% yield. Subsequent NIS/AgOTf-promoted glycosylation of 13 with 5 gave a 67% yield of trisaccharide 14. Removal of the levulinate ester provided 15, which was then glycosylated with 4 leading to the formation of tetrasaccharide 16 in 62% yield over the two steps. The synthesis of tetrasaccharide 21, which was needed for Route 2, also involved an alternating series of NIS/AgOTf-promoted glycosylations and levulinate protecting group removals with hydrazine acetate (Scheme 2B). In all glycosylations leading to mannosidic linkages described in this paper, the stereochemistry of the newly formed glycosidic linkage was determined by measuring its 1 J C-1,H-1 . These values were in the range of 171-177 Hz, as expected for an a-linkage. 29 A comparison of the synthesis of 16 and 21 is provided in Scheme 2C. Four monosaccharide building blocks (3-6) are needed for the synthesis of 16. In contrast, only three building blocks (5-7) are needed to prepare 21. In addition, compared with the overall yield for the synthesis of 16 (25%), the overall yield for the synthesis of repeating unit 21 was much higher: 43%. We then considered the relative glycosylation reactivities that could be expected in both routes in the reactions leading to longer oligomers of the repeating units. In Route 1, chain extension would involve the formation of an a-(1/2) glycosidic linkage. In Route 2, this process would require generation of an a-(1/3)-linkage. In a previous study 30 the relative reactivity between the C-2 and C-3 hydroxyl groups on mannose was investigated. Using a mannose acceptor with unprotected C-2 and C-3 hydroxyl groups, the a-(1/3)-disaccharide was isolated in 80% yield and no a-(1/2)-disaccharide was observed. This suggests that the equatorial C-3 hydroxyl group is more reactive than the axial C-2 hydroxyl group. After these considerations, we decided to focus on Route 2 to assemble the targets. Although this 'frame-shift' approach may appear more complicated, requiring three oligosaccharide building blocks compared to only two for Route 1, the former has three advantages: (1) fewer monosaccharide building blocks are needed for the preparation of 21 compared to 16; (2) there is a higher yield in the synthesis of 21 compared to 16 and (3) the key chain extension process will involve reactions at the more reactive C-3 hydroxyl group. ## Synthesis of tetrasaccharide primer-adaptor and trisaccharide cap domains needed for Route 2 To synthesize the targets via Route 2, a protected primeradaptor tetrasaccharide intermediate was needed (Scheme 3A). The preparation of this compound started by the coupling of glycosyl acceptor 8 (Scheme S1, ESI †) to mannose thioglycoside donor 7 mediated by NIS and AgOTf, which gave disaccharide 22 in 85% yield. The benzylidene acetal in 22 was then hydrolyzed and the resulting diol was acetylated leading to, in 94% overall yield, the formation of 23. From this disaccharide, the levulinate ester was removed using hydrazine acetate, providing 24 (98% yield), which was then subjected to glycosylation with 7 affording a 91% yield of trisaccharide 25. Another cycle of levulinate deprotection and NIS/AgOTf promoted glycosylation with thioglycoside 7 provided the tetrasaccharide 27, which was then treated with hydrazine acetate to afford the primeradaptor tetrasaccharide intermediate 28 in 73% yield over the three steps. The fnal intermediate needed for the targets was the trisaccharide cap at the nonreducing end (Scheme 3B). This building block was accessed by coupling of glycosyl acceptor 6 and glycosyl donor 4 using NIS/AgOTf-promoted glycosylation and subsequent levulinate deprotection to produce disaccharide alcohol 30 in 87% yield over the two steps. Subsequent glycosylation of 30 with thioglycoside 31 (see Scheme S2 in the ESI † for its synthesis) proceeded in 86% yield producing the target trisaccharide 32. Finally, a two-step functional group transformation sequence, from PMP glycoside to glycosyl trichloroacetimidate (TCA), led to the activated donor 33 in 65% yield over the two steps. This was achieved by ceric ammonium nitrate-mediated cleavage of the PMP group and then treatment of the resulting alcohol with tricholoracetonitrile and DBU. ## Exploration of 4 + 4 glycosylation conditions and synthesis of undecasaccharide 1 Having obtained the primer-adaptor and cap domains, we moved forward on assembling the frst target, undecasaccharide 1. Before doing this, we chose to exchange the benzylidene acetals on residues B and C in 21 with acetate esters to simplify the fnal deprotection steps (Scheme 4). We initially used 4 : 1 acetic acid-water at 80 C to hydrolyze the two acetals. Unfortunately, the yield for this reaction was only $60%. Successful acetal cleavage could, however, be achieved by reaction with iodine in methanol at reflux. 31 The 1 H NMR spectrum and mass spectrometric data of the crude reaction mixture showed that, in addition to the desired product 34, there was $15% of 35, formed by conversion of the ketone in the levulinate ester to a ketal. Tetrasaccharides 34 and 35 were inseparable; therefore, the mixture was treated with acetic anhydride and pyridine to give the corresponding acetylated products, which were then dissolved in a 2% solution of HCl in acetone. This led to hydrolysis of the ketal affording the levulinoyl-protected tetrasaccharide 10 in 88% yield over three steps. Conversion of 10 into O-trichloroacetimidate 36 was carried out under the standard conditions described above giving the activated donor in 65% yield. With the primer-adaptor tetrasaccharide acceptor 28 and repeating unit donor 36 in hand, their coupling was studied (Table 1). When 0.4 equiv. of TMSOTf was used as promoter (Entry 1), two side products -TMS ether 37 (Scheme 4, 35%) and glycal 39 (30%)were obtained and the yield for desired product, 38, was modest (52%). Side products 37 and 39 come from acceptor 28 and donor 36, respectively. We concluded that under these conditions, the hydroxyl group on 28 reacted with the TMSOTf promotor to form 37, which does not undergo glycosylation. As a result, some of trichloroacetimidate 36 has no substrate to glycosylate and it undergoes elimination to give 39. To minimize the formation of by-products, we reduced the amount of TMSOTf to 0.2 equiv. (Entry 2). The yield for 38 was improved to 65%; however, 37 and 39 were still produced in 30% and 15% yields, respectively. This suggests that the silylation of 28 is rapid and that using a more sterically-demanding Lewis acid might result in less of these two side products being produced. Indeed, when using 0.4 equiv. of TBSOTf (Entry 3), the yield of the desired product 38 was greatly improved, to 86%, and no 37 or 39 could be isolated. Although the newly formed H-1 resonance could not be identifed in the one-dimensional 1 H NMR spectrum due to overlap, all of the 1 J C-1, H-1 values could be measured from the 1 H-coupled HSQC spectrum. Given this fnding, we used TBSOTf as the promotor for all trichloroacetimidate glycosylations carried out during the synthesis of the targets (see below). With an approach for the construction of octasaccharide 38 in place, the focus shifted to the fnal glycosylation with the cap moiety and elaboration to the target (Scheme 5). Thus, cleavage of the levulinate protecting group in 38 with hydrazine acetate gave octasaccharide alcohol 40 (94% yield), which was subsequently reacted with trichloroacetimidate 33 to afford undecasaccharide 41. This 3 + 8 glycosylation provided 41 in 86% yield. After all of the monosaccharide residues were in place, the focus became changing the functional group on the nitrogen atom, introduction of the phosphate and lipid moiety and the fnal deprotection. The Troc group in 41 was removed in 86% yield via reductive elimination, which employed activated zinc in AcOH/THF, to afford a crude product with a free amine. A common problem of this reaction is the formation of the dichloroethoxy-carbamate by-product, 32 which was minimized by using freshly activated zinc dust. After N-acetylation using acetic anhydride and pyridine, the Troc protecting group was converted to an acetyl group to give undecasaccharide 42 in 90% yield over the two steps. It was next necessary to remove the benzyl ether protecting groups and replace them with acetate esters. This would simplify the deprotection at the end of the synthesis to a single ester cleavage step with an easy to remove by-product (methyl acetate). It can be difficult to remove large numbers of benzyl groups in large oligosaccharides using hydrogenolysis; 22 therefore, we chose to use Birch reduction. The Birch reduction is a strongly basic reaction and, as such, esters are readily cleaved under these conditions. It was then, in principle, possible to remove all of the protecting groups in a single step before replacing them with acetate esters, which was required before the pyrophosphate coupling reaction. However, after exploring this reaction, we found it more convenient to frst cleave the acyl groups, purify the molecule and execute the dissolving metal reduction. This approach has been used for the synthesis of other molecules. 33,34 and in our hands this strategy greatly simplifed the purifcation of the product after removal of the benzyl ethers. Thus, undecasaccharide 42 was treated with sodium methoxide to remove all of the acetyl and benzoyl groups. This intermediate was purifed and then subjected to Birch conditions giving a fully deprotected oligosaccharide, which was then acetylated to afford 43 in 65% yield over the three steps. The fnal steps in the synthesis involved the introduction of the lipid phosphate moiety. To do this, the anomeric TMSET protecting group was removed by treatment of 43 with 25% tri-fluoroacetic acid (TFA) in dichloromethane to give 44 in 83% yield. This hemiacetal was treated with dibenzyl N,N-diisopropyl phosphoramidite and tetrazole to afford a phosphite intermediate, which was oxidized with m-CPBA providing a 75% overall yield of glycosyl phosphate 45. Hydrogenolysis of the benzyl groups on the phosphate gave a glycosyl phosphate intermediate, 46, that was coupled to farnesyl phosphate (47) 35 using a carbonyldiimidazole (CDI)-mediated phosphoesterifcation. 36,37 Such coupling reactions are typically low yielding, 38 but through careful optimization of the conditions, including the use of a large excess of farnesyl phosphate and long (seven-day) reaction times, reasonable yields could be obtained. The product, 48, was then deacetylated with catalytic sodium methoxide in methanol affording the farnesyl pyrophosphate-linked oligosaccharide 1 in 56% yield from 45 over the three steps. The high-resolution electrospray mass spectrum of undecasaccharide 1 showed a molecular ion with two negative charges (M-2H) 2 at m/z ¼ 1101.8628, which agrees with calculated exact mass. The identity 1 was further confrmed by NMR spectroscopy (Fig. S1 and S2 †). ## Assembly of eicosaheptasaccharide 2 After the synthesis of the undecasaccharide 1 was secure, we moved to the preparation of 2, an eicosaheptasaccharide containing 27 sugar residues. Although the approach detailed above could be used to make larger oligosaccharides, we considered that using tetrasaccharide imidate 36 would be inefficient as it would allow chain extension only by one repeating unit in each glycosylation. We therefore decided to synthesize an octasaccharide donor (a dimer of 36) to facilitate chain extension. To do this (Scheme 6), a tetrasaccharide acceptor (49) was obtained in 93% yield by removal of the levulinate group in 10. Next, trichloroacetimidate donor 36 was used to glycosylate 49 promoted by TBSOTf to afford the desired octasaccharide 50 in 88% yield. As was observed in the synthesis of 38, spectral overlap prevented the identifcation of 1 H signal arising from the nascent glycosidic bond in the 1 H NMR spectrum of 50. However, all of the 1 J C-1,H-1 values could be measured from the 1 H-coupled HSQC spectrum, which confrmed the a-stereochemistry of the eight mannose residues in 50. Conversion of 50 to the trichloroacetimidate donor 51 was achieved by selective cleavage of the PMP group and subsequent reaction of the resulting hemiacetal with tricholoracetonitrile in the presence of DBU (65% yield over two steps). Starting from primer-adaptor tetrasaccharide 28 and octasaccharide donor 51, the 20-mer could be synthesized following a 4 + 8 + 8 glycosylation sequence. To obtain the desired 27-mer, it was necessary to synthesize a heptasaccharide donor (53, Scheme 6), analogous to the trisaccharide 'cap' used for the synthesis of 1. Glycosylation of 49 with imidate 33 in presence of TBSOTf led to the desired heptasaccharide 52 in 87% yield. Conversion of 52 into heptasaccharide donor 53 was achieved in 63% overall yield using the same method used to synthesize 51. A 4 + 8 + 8 + 7 reaction sequence was used to assemble the polysaccharide domain of the eicosaheptasaccharide (Scheme 7). The initial 8 + 4 glycosylation between octasaccharide 51 and tetrasaccharide 28 using TBSOTf proceeded in 86% yield. The subsequent deprotection of the levulinate ester on the product dodecasaccharide 54 using hydrazine hydrate at room temperature, conditions that had worked well on the smaller oligosaccharides, was surprisingly slow. Following this reaction by TLC was also complicated by the fact that the starting material and product were inseparable. In the 1 H NMR spectrum of dodecasaccharide 54, the resonance for H-3 of the terminal mannose residue (the hydrogen adjacent to the levulinate ester) appears at 5.37 ppm as a doublet of doublets ( 3 J ¼ 9.5, 3.5 Hz). This signal could be easily identifed in the crude 1 H NMR spectrum of the mixture and thus NMR spectrometry was used to follow the reaction. After four hours at room temperature, the spectrum showed that only 15% of the levulinate group was removed. To achieve full deprotection, a rotary evaporator was used to concentrate the reaction mixture and then the flask was kept rotating for 30 min at 40 C. Under these conditions, the crude 1 H NMR spectrum showed complete disappearance of the peak at 5.37 ppm, suggesting 100% conversion; the yield of 55 was 87% after purifcation. The reason for the low reactivity is unclear. We postulate that the molecule adopts a threedimensional structure that hinders either the formation of the hydrazone intermediate of the levulinolyl ketone moiety, or its subsequent intramolecular cyclization that releases the alcohol. While, in principle, this could be ascertained by TLC analysis of the reaction mixture, this was not possible given the size of the molecules, which rendered their chromatographic mobilities very similar. To further extend dodecasaccharide 55, a TBSOTf-mediated 8 + 12 glycosylation using octasaccharide donor 51 was executed, which provided eicosasaccharide 56 in 71% yield. The levulinate ester group on 56 was removed by using the same method for deprotection as that used on the dodecasaccharide to generate, in 87% yield, 57. Finally, TBSOTf-promoted glycosylation using heptasaccharide 53 as the donor and 57 as the acceptor, generated the desired protected eicosaheptasaccharide 58 in 73% yield. The identity of 53 was supported by its MALDI-TOF mass spectrum, which showed a molecular ion of the sodium adduct at m/z ¼ 10,578, consistent with the calculated mass for this molecule. With the polysaccharide core of the molecule assembled, the next step was the exchange of the Troc group on 58 for an acetate to produce 59. This was carried out in two steps and in 80% yield upon treatment with zinc in acetic acid and acetic anhydride and pyridine. Deprotection by Birch reduction to remove all of the benzyl ethers was then investigated. Initially, consistent with what had been done in the synthesis of 1, all of the acetate and benzoate groups in 59 were removed using sodium methoxide in methanol to afford a partially deprotected product. However, this deacylated product was insoluble in THF, the solvent used for the Birch reduction. Therefore, it was necessary to use the fully protected molecule 59 in this step. Application of the same conditions used for the undecasaccharide led to a mixture of two products: the desired product and one or more side products in which the 2-(trimethylsilyl)ethyl (TMSEt) group appeared to have been lost. Based on the 1 H NMR spectrum of the crude reaction mixture, the ratio between the desired and undesired product(s) was 1 : 1.5 (integration of the methyl signal of the NHAc group). The exact structure of the side product could not be determined by NMR spectroscopy and mass spectroscopy of the crude mixtures showed a number of species with molecular weights lower than that of the desired product. In addition to dramatically lowering the yield, the product and side products were impossible to separate. Fortunately, it was discovered that decreasing the reaction temperature to 80 C and shortening the reaction time to 1.5 h resulted in only a trace amount of the side products being produced. Under these conditions, the twostep yield for the Birch reduction and acetyl was improved to 47%. Removal of the TMSEt group using TFA in dichloromethane gave oligosaccharide 61 in 78% yield. Following the same phosphorylation reaction described in the undecasaccharide synthesis, phosphate 62 was obtained in 92% yield. After removal of the benzyl groups on phosphate 62 by hydrogenolysis, coupling between the resulting glycosyl 1-phosphate 63 and farnesyl phosphate, mediated by CDI, led to the formation of protected glycosyl phospholipid, which was then deacetylated using sodium methoxide in a mixture dichloromethane in methanol. The desired product 2 was obtained in 55% yield over three steps. The high-resolution electrospray mass spectrum of 2 showed a molecular ion with three negative charges (M-3H) 3 at m/z ¼ 1599.1826, consistent with the exact mass of the molecule. Considering the size of 1 and 2, we envisioned that their NMR spectra would not be very informative, but that was not the case. Both 1 H and 13 C NMR spectra (Fig. S1 and S2 †) provide strong support for the structure of the compounds. For example, in the 1 H NMR of 2 (Fig. S1 †), the peak at 5.50 ppm (dd, 1H, J ¼ 7.0, 3.0 Hz) is from the GlcpNAc H-1. The small 2 J P,H-1 (GlcpNAc) coupling constant (J ¼ 3.0 Hz) indicates the connection between pyrophosphate and GlcpNAc residue. Three peaks at 5.46, 5. ## Conclusion In summary, we report the frst chemical synthesis of large lipid pyrophosphate-linked LPS O-PS intermediates (1 and 2). An important design feature was a 'frame-shift' strategy, in which the molecule was assembled not by using building blocks corresponding to natural repeating unit, but instead one where disconnections were made between different residues. This allowed for a reduction in the number of monosaccharide building blocks required and better yields of the glycosylations throughout the synthesis. This non-conventional strategy should thus be considered in future when designing routes to large glycans. Other key features of the route were the preparation three building blocksrepeating unit donor 36, tetrasaccharide primer-adapter acceptor 28 and trisaccharide cap donor 33via an iterative cycle of NIS/AgOTf-promoted glycosylations and hydrazine acetate-mediated levulinate ester cleavages. In addition, a TBSOTf-promoted glycosylation method was developed for glycosylations between oligosaccharide acceptors and trichloroacetimidate donors generated from the oligosaccharide building blocks. Following a 4 + 4 + 3 strategy, the protected undecasaccharide 41 was assembled in glycosylation yields between 82% and 88%. A 4 + 8 + 8 + 7 strategy was employed to synthesize the protected eicosaheptasaccharide 58 in good to excellent yields (70-86%). After nine additional steps, including protecting group manipulation, phosphorylation, coupling with farnesyl phosphate and fnal deprotection, we produced 13 mg and 7 mg quantities of 1 and 2, respectively. While we employed farnesol as the polyprenol in these targets, the use of longer polyprenols should be straightforward. 41 In addition to providing access to valuable probe molecules that are currently being used in biosynthetic investigations, the strategy developed here can be extended to prepare even larger fragments of this O-PS. Moreover, the study provides insights into the challenges faced when assembling structurally-defned polysaccharides and solutions to circumvent them. In particular, it was necessary to overcome not only the formation of unproductive side products in glycosylation reactions (conversion of 36 into 38), but also difficulties in removing protecting groups either selectively (synthesis of 55), or in bulk (the Birch reduction of 59). Indeed, this work suggests that the efficiency of the glycosylations is not dramatically affected when carried out molecules of increasing size, something that has been previously reported for other couplings of large mannosecontaining oligosaccharides 23 and furanose-containing oligosaccharides. 24,25 On the other hand, issues such as low reactivity and more mundane problems such as poor solubility, similar chromatographic mobilities, and spectral overlap, complicated the analysis and deprotection of large intermediates. These latter issues underscore the importance of considering these factors when designing synthetic routes to structurally-defned polysaccharides. Such problems have been previously encountered 22,42 and suggest that the development of new protecting groups that can be removed in quantitative yield, and creative methods for reaction monitoring and execution, are needed additions to the arsenal of methods for synthetic polysaccharide chemistry. Such advances, in addition to improved methods for glycoside bond synthesis, would allow more straightforward and efficient access to structurally-defned complex polysaccharides.

chemsum

{"title": "Synthesis of structurally-defined polymeric glycosylated phosphoprenols as potential lipopolysaccharide biosynthetic probes", "journal": "Royal Society of Chemistry (RSC)"}

the_application_of_poorly_crystalline_silicotitanate_in_production_of_225ac

## Abstract: Actinium-225 ( 225 Ac) can be produced from a Thorium-229/Radium-225 ( 229 Th/ 225 Ra) generator, from high/low energy proton irradiated natural Thorium or Radium-226 target. Titanium based ion exchanger were evaluated for purification of 225 Ac. Poorly crystalline silicotitanate (PCST) ion exchanger had high selectivity for Ba, Ag and th. 225 Ac was received with trace amounts of 227 Ac, 227 th and 223 Ra, and the solution was used to evaluate the retention of the isotopes on PCST ion exchanger. Over 90% of the 225 Ac was recovered from PCST, and the radiopurity was >99% (calculated based on 225 Ac, 227 th, and 223 Ra). The capacity of the PCST inorganic ion exchange for Barium and 232 Th was determined to be 24.19 mg/mL for Barium and 5.05 mg/mL for Thorium. PCST ion exchanger could separate 225 Ac from isotopes of Ra and th, and the process represents an interesting one step separation that could be used in an 225 Ac generator from 225 Ra and/or 229 th. capacity studies indicated pcSt could be used to separate 225 Ac produced on small 226 Ra targets (0.3-1 g), but PCST did not have a high enough capacity for production scale Th targets (50-100 g).Ion-exchange chromatography has been successfully used to separate radioisotopes for medical applications, nuclear fuel reprocessing and other applications 1,2 . However in many instances the ion-exchange material lacks desired selectivity. Current methods of separation rely on commercially available ion-exchange resins that preferentially bind the element based on charge 3 . Extraction chromatography methods have been developed for some separations 4 , but the extraction chromatography resins sometime have slow flow rates, and the extractant can be eluted. Often the process to purify a radioisotope requires the use of multiple columns and result in consuming more time and labor. Chemical separations used in accelerator isotope production process at Brookhaven National Lab (BNL) present interesting challenges. The target masses for production scale targets irradiated at BNL are over 50 grams, and the mass of the radioisotopes produced is less than 0.0001 grams 5,6 . The separation presents challenges if no ion exchange resins are available that have more selective for the isotope of interest rather than the target material.Production of 225 Ac has been of particular interest recently since efficacy of this material has been demonstrated in a treatment of certain types of cancer. 225 Ac can be made available by several routes: separation from 229 Th/ 225 Ra generator, by high energy (100-200 MeV) proton irradiation of natural Thorium target, or by irradiation 226 Ra target with low energy proton (10-24 MeV) 7,8 . Chemical separation of 225 Ac from thorium irradiated with high energy protons is especially challenging. The irradiation results in fission of the thorium in the target and, in addition to 225 Ac, produces a variety of potentially useful radionuclides such as 111 Ag, 105 Rh, 140 La, Ra isotopes, and 140 Ba. As previously mentioned, 225 Ac (t 1/2 = 9.9 days) and its daughter; 213 Bi (t 1/2 = 45 min) are emerging as important isotope for targeted alpha therapy [9][10][11] . Other isotopes in the list can be used for beta therapy ( 111 Ag, 105 Rh, and 140 La), or as a parent in medical isotope generators, for example Radium isotopes and 140 Ba.The use of inorganic ion-exchange materials which selectivity stems from the crystal structure of the ion-exchanger could provide a more attractive mode of separation. Crystalline silicotitanate (CST) and the poorly crystalline CST (PCST) have been synthesized hydrothermally in alkaline media 12 . The reduced crystallinity was obtained by either shortening the reaction time at the same synthesis temperature for CST (200 °C) or reducing the temperature to 170 °C and changing the hydrothermal reaction time. The CST inorganic ion exchangers were initially developed in the 1960s and have been evaluated for nuclear waste treatment due to the materials remarkable selectivity toward Cs and Sr [13][14][15] . Studies showed that Cs and Sr 2+ cations demonstrate higher rate of uptake by poorly crystalline CST compared to the crystalline form. This was attributed to the higher surface area and smaller particle size, which was highlighted to account for the increased rate even though there would be otherwise slow diffusion through the channels. The high selectivity of CST ion exchanger for Cs and Sr was used to decontaminate the Fukushima site 16 . Titanium-based ion exchange materials have been demonstrated to be selective for strontium and actinides in highly alkaline environments 17 . The studies reported herein seek to synthesize, characterize, and evaluate PCST inorganic ion exchanger in the separation process of 225 Ac from irradiated Th target. During the accelerator production of 225 Ac from a 232 Th target 227 Ac and the daughters of 227 Ac are coproduced 18 . The daughters of 227 Ac ( 227 Th and 223 Ra) can be separated from 225 Ac during the purification process 19 . However after purification the 227 Ac would grow in and reduce the purity of 225 Ac 18 . Various literature studies to purify 225 Ac from Thorium and/ or radium have focused on organic cation or anion ion exchange resins 1,2 . The PCST ion exchanger has been synthesized and evaluated for the removal of Sr, Cs and other isotopes from waste streams 20 . In this manuscript the purification of 225 Ac from 227 Th and 223 Ra with PCST ion exchanger was developed. The utility of the purification process was evaluated in the following applications: 225 Ac production from proton irradiated Th or Ra, 225 Ac produced from the 229 Th/ 225 Ra generator, and to purify 225 Ac from the daughters of 227 Ac. ## Results Inorganic ion-exchangers were synthesized by hydrothermal synthesis, purified and the phases were confirmed by comparing XRD patterns to published results 20 . Figure 1 outlines the study design to evaluate the inorganic ion exchangers. Initially, ion-exchange properties of the synthesized materials were evaluated by the batch method, and distribution coefficients (K d ) were determined for 225 Ac and Th. Subsequent studies determined the K d of several elements (Th, La, Ce, Rh, Ag, and Ba) on PCST ion exchanger. Optimal conditions for the separation on PCST inorganic ion-exchanger were evaluated with a representative sample containing 225 Ac, 227 Th, and 223 Ra. Capacity studies were performed with barium and thorium on PCST to determine if the ion exchanger could be used for radium (0.3 g) and/or thorium (50-100 g) production targets. The stability of the PCST ion exchanger was determined in: ammonium acetate buffer from pH 5 to 1, hydrochloric or nitric acid at 0.1,1, 2 and 3 M. ## Evaluation of inorganic Ion-Exchanger Selectivity. The evaluation of the selectivity of poorly crystalline silicotitanate (PCST) for Rh, Ba, La, Ce, Th, and Ag were performed with batch studies, and the data for Th, Ag, and Ba are presented in Fig. 2. The results of the PCST ion exchanger show an insignificant selectively at low pH for Ba, La, Ce, Th and Rh, with a higher selectivity for Ag (3465 ml/g) at a pH of 1. As the pH goes from 1 to 5 the PCST ion exchanger increases selectivity for Ba, and Ag (46305 ml/g, 6796 ml/g respectively). The PCST ion exchanger had low selectivity for Ce, La and Rh (100 ml/g, 71 ml/g and 85 ml/g respectively). The distribution coefficients for the trivalent cations Ce and La increased as the pH went 1 to 5. column studies. To increase the flow rates buffer was added to the 100-200 mesh PCST material and the solution was decanted. This was able to remove fine particles of PCST material, and the flow rates increased to 0.25-1 ml/min. PCST breakdown study. The quantification limit for Ti on the ICP-OES was defined by analysis of diluted standards and the acceptance criteria for the true concentration value and the % RSD was within 10%. The quantification limit for Ti by ICP-OES was determined to be 0.005 ppm. ICP-OES analysis of all samples indicated Ti breakthrough (Fig. 3). The Ti breakthrough was lowest in 0.5 M ammonium acetate pH 5 with 0.05-6 µg of Ti present in the load, 0.5 M ammonium acetate pH 5 and 3 rinses. Rinsing the PCST material with ammonium acetate at pH 1 resulted in 352-405 µg of Ti. In subsequent rinses with 0.1 M HCl or nitric acid the PCST material showed a slight higher amount of Ti breakthrough in HCl (91-110 µg) versus nitric acid (63-86 µg). Higher concentrations of acid resulted in more breakthrough of the Ti from the PCST with 1 M (480-700 µg Ti per fraction) 2 M (750-1022 µg Ti per fraction) and 3 M (703-1135 µg Ti per fraction). In all elutions with HCl or nitric acid breakthrough of the Ti from the PCST material was higher for HCl. Separations of Th, 225 Ac and other metals. In summary the Kd data indicated: the PCST ion exchanged has favorable properties to capture Thorium at pH values of 1 to 2 while 225 Ac would be less favorable; the resin has high affinity for Ba at pH values of 3 to 5, and is more selective for Ag at pH values of 1, 4 or 5 than Th, La, Ce, and Rh. A column of PCST ion exchanger was used to capture Thorium and barium from a solution containing Th, Ba, 225 Ac, Rh, La, Ag, and Ce in 0.5 M NaOAc at pH 2 (see Supplementary Fig. S4). The eluted solution contained 95% of Ac, 90% of Ce, 73% of La and 77% of Rh. The data indicates 15% of La, 5.8% of Rh, and Th, Ag and Ba were totally absorbed and retained on the PCST column. The absorbed metals except Ag were recovered from the PCST column using 3 M nitric acid solution. The capacity of the PCST inorganic ion exchange materials for Barium and 232 Th was determined to be 24.19 mg/mL for Barium and 5.05 mg/mL for Thorium. Ba and La column studies. The PCST ion exchanger was able to retain Ba in 0.5 M ammonium acetate at pH 5 while eluting La (see Supplementary Fig. S2). Combining the load and rinse 1-3 provided 85% recovery of La while only 0.15% of the Ba was present. Rinse solutions 4-5 and elution solutions 1-2 recovered 99.8% of the barium with 92.8% of the barium eluting in 0.5 M ammonium acetate at pH 1. 223 Ra, 225 Ac, 227 Th studies. pH study & PCST breakdown: A study was performed with PCST material with 225 Ac, 223 Ra and 227 Th in 0.5 M ammonium acetate at pH 5 and the column was rinsed with the buffer at 4.5, 4, 3.5, 3, 2.5, 2.0, 2.5, 1.5, 1 (Fig. 4). The 225 Ac eluted with two peaks at pH 5 and pH 3 with 96% eluting in the load and fractions with a pH from 5-3.5. The 223 Ra and 227 Th were both retained on the column and began to elute from the column at pH 1.5 with 98-100 percent of the isotopes eluting in the buffer at pH 3 to 1. The activity retained on the column was not measured. The breakdown of the PCST was evaluated and Ti was below quantification limits in the eluted load and pH 5 solutions, but Ti was quantified in all other fractions. The amount of Ti breakthrough in buffer at pH values from 4.5-2.5 was less than 1 μg per fraction. Eluting the PCST column with 0.5 M ammonium acetate buffer at pH 2, 1.5 and 1 resulted in higher breakthrough of Ti (11.7, 80, and 89 μg). Initial optimization of the separation with PCST: Initial optimization of the separation examined using a rinse sequences with 6 bed volumes (BV) or column volume of 0.5 M ammonium acetate at both pH 5 and 3, and 12 www.nature.com/scientificreports www.nature.com/scientificreports/ BV at pH 1 (see Supplementary Fig. S3). Combining the load, pH 5 and 3 rinses resulted in 91% of the 225 Ac, and approximately 39-41% of the 225 Ac was eluted in pH 5 and 34-42% of the 225 Ac was eluted in the pH 3 rinse. The 0.5 M ammonium acetate pH 1 rinse step eluted 44-81% of the Ra and 95% of the Thorium. The column retained 0.1-1.4% of the 225 Ac and 19-56% of the 223 Ra. Optimized purification and PCST breakthrough: The rinse sequence used: 3X3BV of 0.5 M ammonium acetate at pH 5, 2X3BV of the buffer at pH 3, 2X3 BV of the buffer at pH 1 (Fig. 5). Combining the load, pH 5 and the first pH 3 rinse (heavy black line) resulted in the elution of 91% of the 225 Ac. Based on the activity of 225 Ac, 223 Ra and 227 Th the initially radionuclidic purity of the 225 Ac was 78.4%, and the radionuclidic purity of the purified 225 Ac in the combined fractions was 99.3%. In the pH 5 rinse step 62.9% of the 225 Ac was eluted and in the pH 3 rinses 21.3% of the 225 Ac was eluted with 18.9% eluting in the first pH 3 rinse and only 2.3% eluted in the second pH 3 rinse step. The second pH 3 rinse and the pH 1 rinses eluted 95% of 227 Th. The pH 1 rinse contained 42% of 223 Ra and 53.7% was retained on the column. Ti breakthrough was checked in the second pH 5, first pH 3 and both pH 1 fractions, and the Ti breakthrough was 0.025, 0.29, 120.7 and 142.5 µg per fraction consistent with previous studies. ## Discussion Different separation methods and materials are being evaluated to purify 225 Ac from Thorium, so that the US Department of Energy can supply 225 Ac to researchers and clinicians on clinical scales 19, . In this manuscript PCST ion-exchanger was evaluated to determine if the material can be utilized for the purification of 225 Ac from Thorium or radium targets or in a 229 Th/ 225 Ra generator. www.nature.com/scientificreports www.nature.com/scientificreports/ column studies. Column studies with PCST ion exchanger were plagued with poor (1 ml/15 min) or no flow rates. Sieving the PCST material through various mesh filters did little to increase the flow rates. In all column studies 100-200 mesh PCST was used and reasonable flow rates were achieved by soaking the PCST in buffer and decanting the buffer. Acid study of PCST. Ammonium acetate buffer at pH 1 and both hydrochloric acid and nitric acid at a concentration of 0.1 M lead to breakdown of PCST material. The breakdown of the PCST at pH 5-3 is far less then at more acidic pH values. Using the PCST material at pH 1 or in 1, 2, and/or 3 M acid may require a cleanup column to remove Ti breakthrough. Separation of Ba from La. The Ba-La separation was conducted to assess the potential of a separation of Ra and Ac radioisotopes with a PCST column, with the stable isotopes serving as surrogates for the radioisotopes. The elution of both metals at separate pH values was clear, and La eluted at pH 5 while Ba eluted at pH 1 (see Supplementary Fig. S2). Application: This purification approach could be used to separate 140 La from 140 Ba in a generator system. The study indicates all trivalent lanthanides will be eluted at pH 5 with 140 La and 225 Ac. Evaluating PCST ion exchangers for purification of 225 Ac. The elution of Ba and La on PCST was repeated with 223 Ra, 227 Th and 225 Ac, and the separation was optimized to purify 225 Ac. The separation of 225 Ac from 223 Ra and 227 Th had good reproducibility. In five studies (the pH study, three optimization studies with 225 Ac, 223 Ra and 227 Th, and the column studies 225 Ac, Th, Ag, Ba, Rh, Ce, La) greater than 90% of the 225 Ac eluted in the load, pH 5 and/or pH 3 solutions. The optimized elution method to separate 225 Ac was selected based on the elution of the highest percentage of 225 Ac with the lowest percentage of impurities, in this case, 223 Ra and 227 Th. The optimized method rinsed the PCST column with more 0.5 M ammonium Acetate at pH 5 and 3 resulted in a shift in the 225 Ac elution peak. The result was more 225 Ac was eluted earlier with 92% eluted in the pH 5 and first pH 3 rinse step. The process produced very pure 225 Ac with 223 Ra and 227 Th eluting at low pH, and this data indicates the optimized method with PCST material could be used in several different production approaches to purify 225 Ac from Thorium and/or Radium radioisotopes. 225 Ac from a 225 Ra/ 229 Th generator: 225 Ac has been produced at Oak Ridge National Lab (ORNL) from a 225 Ra/ 229 Th generator and they produce 5.5 × 10 10 Bq (~1.5 Ci) per year 24 . The one column separation of 225 Ac from 223 Ra and 227 Th with PCST could simplify the multicolumn approach used at ORNL to purify 225 Ac in the 225 Ra/ 229 Th generator. The ORNL process is a four step chemical process with two MP1/NO 3 columns to separate 225 Ac and 225 Ra from 229 Th. Then the 225 Ac is purified from 225 Ra with two AG50X4/1.2 HNO 3 purification steps. The high selectivity of PCST for both thorium and radium would simplify the purification of 225 Ac, and the process would be one column, making the purification shorter than the ORNL process. Accelerator produced 225 Ac: Accelerator produced 225 Ac can be produced at high energy (>100 MeV) with a natural thorium target or at low energy (10-25 MeV) from a 226 Ra target. PCST for the purification of 225 Ac from 232 Th targets: Although the PCST column worked to separate 225 Ac from Thorium, Ra and some fission products the approach is similar to published separations that capture Thorium on an MP1 column and let 225 Ac pass through the column 24 . This separation strategy would work for smaller thorium target. For larger scale clinical production of 225 Ac with 50 to 100 g Thorium targets and potential Thorium stack targets could be required resulting in 100-600 grams of Thorium in the separation. The small capacity of PCST for Thorium indicates a large mass of PCST ion exchanger would be needed to capture all the Thorium. To process one 50 gr Thorium target it was estimated 10 L PCST column would be needed, so the material does not have a reasonable capacity to purify 225 Ac from Thorium targets. PCST for the purification of 225 Ac from 226 Ra targets: Low energy protons irradiating a 0.3 g 226 Ra target can produce clinical scales of 225 Ac (~ 1 Ci/target) 7,25 . The amount of PCST ion exchange material to process a 226 Ra target was estimated from capacity studies with barium, and the data indicates a minimum of a 12.5 mL PCST column would be needed to retain 226 Ra in the target. The PCST separation method to purify 225 Ac could be used to purify 225 Ac from a 223 Ra targets. 226 Ra could be eluted from the PCST ion exchanger in pH 1 buffer, and the process could be used to recycle the 226 Ra. ## Applications of separation. Removal of radio impurities ( 223 Ra and 227 Th) in accelerator produced 225 Ac from Th targets: Accelerator produced 225 Ac from a proton irradiated thorium targets has 0.1% abundance of 227 Ac (t 1/2 = 21.8 years) at end of bombardment. 227 Ac decays to 223 Ra (t 1/2 = 11.4 days) and 227 Th (t 1/2 = 18.5 days), and 227 Ac and the daughters represent the major radio-impurities for 225 Ac. The expiration of a batch of accelerator 225 Ac is defined by the purified 225 Ac failing one of its specifications, which are currently being determined by the Trilab 225 Ac team. The radiopurity of 225 Ac will likely be the first specification that fails. The Trilab team has estimated the radiopurity post purification in the presence of 227 Ac and 227 Ac and daughters for BLIP produced 225 Ac from Th targets 18 . Removal of 223 Ra and 227 Th would increase the radiopurity of 225 Ac produced from a Th target at BLIP. During the development of the PCST separation the original radiopurity of 225 Ac was 78.4% (calculated from 225 Ac, 227 Th and 223 Ra), but after performing the purification with PCST column the radiopurity was 99.3% and the recovery of 225 Ac was 92.4%. This improvement in radiopurity indicates that this separation can be used to extend 225 Ac shelf life by removing 227 Ac daughters, 223 Ra and 227 Th. A PCST column run on the 225 Ac sample would remove 223 Ra and 227 Th; increasing the radiopurity to greater than 98% and this could be done out to 30.1 days. For 225 Ac produced from a Th target at BLIP the radiopurity falls below 98% after 14.9 days after the last separation step. A radiopurity greater than 98% can be achieved with a PCST column out to 27.8 days. In a clinical setting, the collection of the 18 BV of the pH 5 rinse and first 6 BV of the pH 3 rinse would recover 92.4% of eluted 225 Ac and a negligible percentage of impurities. This 12 mL sample of 225 Ac would be easy to evaporate. www.nature.com/scientificreports www.nature.com/scientificreports/ conclusion The impact of developing a material with specific isotope and/or metal selectivity would potentially be invaluable in assisting with efforts in medical isotope production. The studies herein evaluated titanium ion exchanger and examined if the material could be used for the purification of 225 Ac. Examination of the effects of acid rinses on PCST indicated that even 0.1 M acid, either nitric or hydrochloric, breaks down the material and resulted in the elution of the material (Titanium). This led to the conclusion that acid more dilute than 0.1 M is needed when working with PCST. For the separation of 225 Ac from radioactive Ra and Th, the optimized method used 18 BV of buffer at pH 5 and 6 BV of buffer at pH 3. This lead to a high recovery of eluted Ac (92.4%) and high radiopurity (99.3%). An 225 Ac 229 Th generator can be established based on this separation. Capacity studies of Barium and Thorium on PCST indicated that the material did not have a high enough capacity for a production scale thorium target (≥50 g) of PCST. However, PCST could be used to purify 225 Ac from smaller production scale 226 Ra targets (0.3 g). ## Materials and Method Reagents were used from manufacturer without additional purification: phosphoric acid, fumed silica, titanium isoproproxide and KOH Pellets were purchased from Sigma Aldrich. Sodium hydroxide (98%), nitric acids (70% optima) and trace metal grade hydrochloric acid were purchased from Fisher. La, Ce, Lu ICP standards were purchased form Fluka in 1000 mg/L concentrations. ICP single elemental certified standards of: Th, Ag, Ba, Rh, Ce, and La were purchased from SPEX Certiprep. All solutions were prepared using Milli-Q water and all experiments were conducted at room temperature. All the chemicals used were of analytical reagent grade. Buffers were prepared from previously prepared 0.5 N Sodium Acetate buffer and adjusted with 8 N HCl or 10 M NaOH. Initial and equilibrium pH readings were obtained using a Denver Instruments UB-10 pH/mV meter calibrated at pH 2.0, 4.0 and 7.0. Since Ba chemically behaves similarly to Ra and La is chemically similar to 225 Ac for some studies Ba and La can be used as surrogates. 225 Ac radiotracer was supplied by Oak Ridge National Laboratory as a dried sample, and the sample was dissolved in 0.1 M HCl solution prior to use. 223 Ra and 227 Th were present in some 225 Ac samples as a result of the decay of 227 Ac. Synthesis of PCST inorganic ion-exchanger. PCST were synthesized according to published methods 26,27 . Phase and purity was confirmed by powder X-Ray diffraction (patterns collected using a Rigaku MiniFlex II Desktop X-Ray diffractometer sampling at 0.040 degrees at a speed of 1degree/min, starting at 5 degrees and ending at 60 degrees. 225 Ac and Th. Experimental solutions, consisting of 20 mg L −1 of Th and 3.7 × 10 4 Bq 225 Ac in 0.5 M sodium acetate (NaOAc), were adjusted to a pH of 1, 2, 3, 4 or 5 with 10 M NaOH or 69% HNO 3 . Inorganic ion-exchangers HCST (100 ± 0.5 mg) were added to 10 mL of the metal containing solution. The tubes were shaken (using Thermo Scientific compact digital microplate shaker) for 12 hours at room temperature. The tubes were centrifuged (4000 rpm, 2598 RCF) for 4 minutes and the aqueous phases were separated using 0.2 µm micro syringe filter. An aliquot from the aqueous phases was diluted in 2% nitric acid and analysis performed. ## Determination of K d values for different inorganic ion-exchangers. Rh, Ag, Ba, La, Ce and Th on PCST inorganic ion-exchangers. A solution consisting of: 30 mg L −1 of each of Ag, Ba, Rh, Ce, and La; and 150 mg L −1 of Th in 0.5 M NaOAc. Ba was used as a surrogate of Ra, and La was used as a surrogate for 225 Ac as the chemistries are similar. The PCST inorganic ion-exchangers (30 ± 0.3 mg) were added to the 10 mls of the metal containing solution. The samples were processed and analyzed as described in the previous section. ## Analysis and calculations of Batch studies. Actinium activities in the initial, intermediate and final solutions were determined by using a gamma spectrometry (ORTEC) with a calibrated high purity germanium detector 28 . The separation fractions containing 225 Ac, 227 Th and 223 Ra were quantified after 24 hours by gamma spectroscopy at 236 ( 227 Th), and 269.6 keV ( 223 Ra). After 24 hours the original activity associated with 213 Bi and 221 Fr has decayed away (24 hours >10 half-lives of 213 Bi and 221 Fr). The presence of 213 Bi and 221 Fr in the samples is the result of in growth from 225 Ac, and activity associated with 213 Bi and 221 Fr would be at equilibrium with 225 Ac. At the time of analysis, 225 Ac and its daughters (specifically 221 Fr and 213 Bi) were at equilibrium with 225 Ac, and the gamma peak at 440 KeV for Bi-213 was used to quantify 225 Ac. The 218 KeV gamma peak for 221 Fr was used to quantify 225 Ac and similar results were obtained. The concentration of thorium and other metals were measured using ICP-OES (Perkin Elmer Optima 7300 DV spectrometer), and the instrument was calibrated according to published methods 28 . The wavelength (λ) used for thorium, barium, lanthanum, cerium, rhodium, silver and titanium analysis were 283.73, 233.527, 408.672, 413.764, 343.489, 338.289 and 334.94 nm respectively. K d values were calculated based on the following equation: ( ) Where C i = Stock concentration metal, C f = Final concentration metal, V = Volume of stock solution, and g = Measured weight of ion-exchanger. K d values were plotted versus pH, and the standard deviation of triplicate samples was calculated.

chemsum

{"title": "The application of poorly crystalline silicotitanate in production of 225Ac", "journal": "Scientific Reports - Nature"}

nanocluster_growth_and_coalescence_modulated_by_ligands

## Abstract: We describe a model of nanocluster formation that incorporates competition between ligand adsorption and nanocluster growth. Growth occurs through the addition of a metal-ligand complex and coalescence of nanoclusters. The competition between ligands for binding sites on the nanoclusters and growth of the nanoclusters through coalescence creates interesting growth pathways. The patterns are reminiscent of those observed in the synthesis of gold thiolate nanoclusters. For a particular set of rate coefficients, described herein, we observe the formation of a kinetically stable nanocluster that participates in coalescent growth. This determines the size interval of the resulting nanoclusters in the size distribution. The kinetically stable cluster can be tuned by modifying the functional form of the number of surface sites on the nanoclusters, thereby changing the growth pathway and the final sizes of the clusters.Nanoclusters | Ligands | Coalescence N anoclusters (NCs), defined as clusters of metal atoms less than 2 nm in diameter (1), contain unique properties with applications in catalysis (2-6), bio-imaging and sensing (7-10), and medical therapies (11). The small sizes of NCs present many synthetic challenges (1). The primary challenge is NCs are metastable since thermodynamic stability increases with particle size (12). Therefore, successful synthesis of NCs requires precise control over experimental conditions, often utilizing ligands to trap NCs in a metastable state (6), sometimes referred to as colloidal stability (13). Synthesis of NCs generally proceeds through two routes:"bottom-up" and "top-down." In the bottom-up approach, metal ions are reduced to zero-valent metal atoms, often in aqueous solution in the presence of ligands, which then proceed to grow to form NCs and nanoparticles. Using this approach, atomically precise NCs have been synthesized (1-3, 6). However, a detailed understanding of their formation is still lacking due to a dearth of experimental data at early times in the formation process (13). Current advances in microfluidic technology may help to improve this situation ( 14).Among the many approaches to modeling nanoparticle formation, kinetic rate equations (KREs) (15-19) have shown size focusing is possible in a purely reaction-limited regime (15). Further, KREs model the evolution of nanoparticle nucleation and growth beginning at the formation of zero-valent metal atoms (monomers), and have an advantage at describing the early stages of NC formation. However, the existing KRE models are limited for two reasons: 1) assumptions from classical nucleation theory (CNT) are incorporated (15,18) and 2) a coalescent growth mechanism is lacking (15,(17)(18)(19).Experiments indicate that CNT fails to correctly describe NC formation, (13,20,21) or at least requires modifications to do so (22). In many experiments, the failure of CNT is demonstrated by the observation of pre-nucleation clusters-clusters considered too small by CNT to be stable in solution 35 (23). For strongly associating systems (such as Ir, Au, Ag, 36 etc.), it has been argued the concept of a critical nucleus from 37 CNT does not apply (13,16). Therefore, the assumption of a 38 critical nucleus in theoretical models is questionable, especially 39 since experiments indicate the critical nucleus is just one or 40 two monomers in strongly associating systems (16, 24, 25). 41 Coalescent growth is an important growth mechanism for 42 nanoparticles (13, 26-31), but was not included in other ligand-43 mediated growth models (17-19), including our own (19). growth such that i + k must always be less than imax (the maximum size of the NCs in the model) and j + l must always be less than the Ns(i + k) (the available number of surface sites on a cluster with i + k metal atoms). ) ka i,j = ka(1.0 − j Ns(i) ) ) k d i,j = k d j Ns(i) ke i,j = ke j Ns(i) From the above reaction scheme, rate equations for each species can be derived. Next we provide the rate equation for Ci,j (Eq. ( 7)). The brackets indicate molarity and the factor 1/4 in the last term modeling coalescent growth avoids overcounting the possible combinations of clusters. The equations are solved using the differential equation package in Julia (33,34). The code is available as a Julia package at https: //github.com/dsuvlu/NanoclusterModeler.jl. Additional details about the methods can be found in the Materials and Methods section and in the Supporting Information. The indices w, x, y, and z in Eq. ( 7) are dummy indices. Table 1 provides a summary of the rate coefficients used in the model. For all of the following results, we set imax = 68 and the rate coefficients k ub = 10 −7 s −1 and k d = 10 ## Results and Discussion Utilizing the model, we investigated ligand-mediated nanocluster formation and growth by systematically varying the rate coefficients and initial concentrations of M + and L. For the following results, we set kp = 10 3 s −1 , k b = 10 5 M −1 s −1 . This was motivated by the observation that experiments demonstrating synthesis of NCs are often conducted with a strong reducing agent, such as NaBH4. Furthermore, the metal atoms are often bound to one or more ligands before the addition of the reducing agent to the solution. Therefore, our choices of kp and k b ensure fast reduction kinetics and that most of the growth will occur through the addition of a metal-ligand complex to the NCs. Figure 1 outlines four scenarios of the different growth pathways observed. In scenario (I) the ligand elimination rate ke from the surface of clusters Ci,j is larger than the ligand association rate ka. In this case, there are few ligands on the surface of the NCs to prevent growth and coalescence. Therefore, the clusters grow to large sizes. In scenario (II) the ligand association rate is about three orders of magnitude larger than the elimination rate, but approximately the same order of magnitude as the growth rate. In this case the ligands attach to the dimer faster than the dimers can grow. Consequently, growth stops at the dimer. In scenario (III), the ligand association rate is at least three orders of magnitude larger than the elimination rate, but the growth rate is at least four orders of magnitude larger than the ligand association rate. As a result, the NCs grow growth proceeds in multiples of the kinetically stable cluster. ## 99 We expand on these observations in the following discussion. displays the NC number density as a function of diameter for ke = 10 2 M −1 s −1 . In this case, the ligand elimination rate is large enough to allow NC growth beyond the kinetically stable cluster. Figures 3 and 4 compare the growth pathways of NCs without (kc = 0) and with coalescence (kc = 10 3 M −1 s −1 ), respectively, for a ligand association rate of ka = 10 −3 M −1 s −1 , which is a factor of 10 6 smaller than the ligand association rate in Fig. 2. Without coalescent growth, and while ka/ke ≥ 1, the NCs again form the kinetically stable C8,8 (Fig. 3 (a) and (b)). However, since ka is much smaller in this case than that of Fig. 2, we do not observe stabilization of the dimer in Figs. 3 and 4. Comparing Fig. 3 (b) to Fig. 4 (b), where the latter incorporates coalescence with kc = 10 3 M −1 s −1 , we observe NC growth in factors of the kinetically stable cluster. In this case, kc, kg ka and ka/ke ≈ 1, so that ligand elimination occurs at a rate that allows slow coalescence of the NCs. Under these conditions, the NCs display a particulate size distribution with the spacing determined by the size of the kinetically stable cluster. This is a demonstration of scenario (IV) as discussed earlier. Movie S1 in the Supporting Information demonstrates the time evolution of the number density and adsorbed ligands for scenarios (I) and (IV). The Supporting Information also contains an expanded collection of results from which we extracted the scenarios discussed here. For example, Figures S15 to S18 and Movie S2 display data where the concentration was scaled for different combinations of ka and ke. We also describe the time evolution of the NC size distribution for the rate coefficients used in Fig. 4(b), but with different expressions for Ns(i). Figure 5 illustrates the time evolution of the NC number density as a function of diameter for different expressions of Ns(i). The size of the kinetically stable cluster, as shown in the third plot of the number density in the time evolution, changes from C10,10 to C4,4 as the scaling factor in Ns(i) changes from the 2.20 to 1.70. This completely changes the results in the size distribution of NCs after coalescent growth has completed. Movie S3 in the Supporting Information illustrates the time evolution of the data displayed in Fig. 5. Figure 6 demonstrates that each of the kinetically stable clusters occur at i/Ns(i) = 1.0 for each expression of Ns(i). This observation can be rationalized in the following way. While i/Ns(i) < 1.0, the number of surface sites is greater ## 171 A strength of our model is its unique ability to produce the 172 interesting NC growth patterns observed in Fig. 5. These in-173 teresting growth pathways originate from competition between 174 NC growth and the ligand association/elimination kinetics modulated by coalescence. We believe this feature generally 176 describes NC growth when the metal atoms are attached to 177 ligands prior to reduction. However, the interesting growth 178 pathways just described only occur for a particular set of 179 rate coefficients. The rate coefficients are such that dimer-180 ization occurs relatively quickly, but not so quickly that all 181 the monomers are depleted from the solution. This allows 182 monomers to attach to the clusters so the clusters grow to 183 larger sizes, eventually forming the kinetically stable cluster. 184 Furthermore, the ligand association rate is sufficiently small 185 such that the ligands do not attach to the growing NCs and 186 prevent growth. These conditions allow the kinetically stable 187 cluster to form. Additionally, the ligand elimination rate is 188 sufficiently slow to allow coalescent growth to occur so that the 189 NCs grow in multiples of the kinetically stable cluster. While 190 this result is exciting because it demonstrates ligand-mediated 191 coalescent growth in a kinetic model for the first time, it also 192 reveals a deficiency in the model. The model does not ascribe 193 any thermodynamic stability to the NCs except in the sense 194 that the NCs prefer to grow to larger sizes. The model could 195 be improved by making a select combination of metal atoms 196 and ligands particularly stable, for example, by the use of 197 super-atom theory (35,36). Furthermore, we do not incorpo-198 rate dis-or comproportionation in the model which has been 199 shown to be important for NC formation (37). We also do 200 not describe further NC interactions between charged species, 201 which can be addressed, for example, using DLVO theory (13). 202 Lastly, the NCs in Fig. 5 sizes with an increase in imax. This could be remedied by 206 using a size-dependent ligand association rate, such that the ## Materials and Methods The full system of ordinary differential equations are provided in the Supporting Information. The number of ODEs neqns is approximately equal to neqns ≈ imax Ns(i). If imax = 68 and Ns(i) = [2.08i (2/3) ], then neqns ≈ 1500. In this case, a single calculation takes approximately five hours on a single Intel Skylake processor. If imax = 500, then neqns ≈ 40000. The system becomes prohibitively large as imax increases. To solve the equations we used the radau solver from the Julia (v1.1.1) differential equation package (v6.9.0) with the absolute and relative tolerances set to 1.0 × 10 −10 . The package ODEInterfaceDiffEq (v3.5.0) was used to interface to the radau solver. The equations were solved for a time span of 1.0 × 10 6 s. Equations for [C2], [Ctot], Davg, and number density ρ i are provided below. The diameters of the NCs are calculated according to a method developed elsewhere (18), where D i = D M (i/0.45) 1/3 and D M is the diameter of a monomeric unit which we set to 0.25 nm. The code is available as a Julia package at https://github.com/dsuvlu/NanoclusterModeler.jl. [Ci] = j [Ci,j ] [Ctot] =

chemsum

{"title": "Nanocluster Growth and Coalescence Modulated by Ligands", "journal": "ChemRxiv"}

the_effect_of_n-heterocyclic_carbene_units_on_the_absorption_spectra_of_fe(ii)_complexes:_a_challeng

## Abstract: The absorption spectra of five Fe(II) homoleptic and heteroleptic complexes containing strong sigma-donating N-heterocyclic carbene (NHC) and polypyridyl ligands have been theoretically characterized using a tuned range-separation functional. From a benchmark comparison of the obtained results against other functionals and a multiconfigurational reference, it is concluded that none of the methods is completely satisfactory to describe the absorption spectra. Using a compromised choice of 20% exact exchange, the electronic excited states underlying the absorption spectra are analyzed. The low-lying energy band of all the compounds shows predominant metal-to-ligand charge transfer (MLCT) character while the triplet excited states have metal-centered (MC) nature, which becomes more pronounced with increasing the number of NHC-donor groups. Excited MC states with partial charge transfer to the NHC-donor groups are higher in energy than comparable states without these contributions. The presence of the low-lying MC states prevents the formation of long-lived MLCT states. ## Introduction Transition metal organometallic complexes continue attracting attention for various catalytic applications, including converting and storing solar energy in a more sustainable form. The latter endeavour is highly fueled by the increasing growth of energy demands of human society and the lack of exhaustible resources. Successful examples of catalytic homogeneous and heterogeneous systems utilize noble-metal containing photosensitizers, see e.g. Refs. 4,5. However, from the economical and ecological points of view, the replacement of noble metals with earthabundant, inexpensive, and not-toxic metal is enticing. Iron is a good candidate for this purpose. Thus, iron complexes are heavily in the spotlight, not only as promising photocatalysts, but also as convenient alternatives for conventional processing and magnetic storage of information due to their capabilities for photoinduced ultrafast spin-flip. 6,7 The development of new photonic materials requires understanding of the underlying photophysical processes 9 as well as how chemical substitution influences targeted properties, such as the lifetime of charge-transfer (CT) states, the energetic ordering of the electronic excited states and the reaction yields of competing photophysical channels. In applications involving spin-flip, the interplay between local metal centered (MC) and metal-toligand charge-transfer (MLCT) states is also of primary importance. To this end, theoretical modeling has emerged as a powerful tool to guide and complement experimental techniques. Due to the size of most transitions metal complexes and their large number of accessible electronic excited states density func-tional theory (DFT) and its time-dependent (TD) extension in the linear-response formulation have positioned themselves as one of the most popular computational avenues to deal with such systems, see e.g. Refs. 10-18. The capabilities of DFT to obtain ground state properties for large systems are well established, 19,20 although results might strongly depend on the employed exchange-correlation functional. 21 Also the deficiencies of DFT and TDDFT are well-known. The erroneous description of long-range CT properties such as ionization potentials (IP), electronic polarisabilities, and energies of CT electronic states with conventional DFT stem from the approximate description of the exchange potential, leading to a self-interaction error and wrong decay of the electron density in the long-range limit. 22,23 Additionally, the so-called derivative discontinuity condition 24 is not exactly fulfilled in DFT, leading to different approaches, such as, e.g., scaled hybrids 25,26 and range-separated hybrid functionals. 27,28 In particular, optimally-tuned functionals have been successfully applied in a number of systems, yielding improved descriptions of various molecular properties related to fundamental and optical gaps such as IPs, 29,30 CT and Rydberg transition energies, optical rotation, hyperfine couplings, and others. The aim of this paper is to investigate the performance of various density functionals including tuned range-separated ones vs. a multiconfigurational reference 42 applied to the excited-state properties of a series of iron(II) homoleptic and heteroleptic complexes that were recently synthesized by Zimmer and coworkers 8 (see Fig. 1). These authors showed that the introduction of strong sigma-donating N-heterocyclic carbene (NHC) ligands in metal-organic complexes prompts to increase the photochemically relevant MLCT state lifetime by destabilizing e * g orbital energies. Here, we analyze the energetics and the nature of the electronic transitions, paying attention to the order of MLCT versus MC states, upon introducing different changes in the ligand substituents. The rest of the manuscript is organized as follows. After a brief introduction in the theory of long-range separated functionals as well as details of multiconfigurational calculations, we present our results on the lowest excited states and absorption spectra computed with different methods for the series of iron catalysts. We then proceed to describe the excited-state properties of the considered complexes and analyse the nature of the lowest-lying excited states, to conclude with the main findings. ## Computational Details The DFT functional optimization has been performed using two parameters, α and ω, for the generalized form of the short-/long-range partitioning of the Coulomb interaction 43 and an error function kernel Γ(ωr 12 ) = erf(ωr 12 ) within the LC-BLYP functional, i.e. 1 In order to tune the functional for the systems 1 to 5, the so-called ∆SCF method 31,47,48 has been applied. Here, the IP and EA are calculated as the differences between the ground state energies of systems with N and N ± 1 electrons, i.e. This yields separate tuning functions and In order to obtain a proper description of the fundamental gap, J 0 (α, ω) and J 1 (α, ω) for IP and EA should be minimized simultaneously leading to following tuning function: 38,49 In principle, this function provides a manifold of (α, ω)-pairs where J * is minimal. Selecting optimal values from those pairs requires an additional criterion. For the exact exchange-correlation density functional, the energy E(N) must vary linearly for fractional electron numbers between integer Ns. 50 This, however, does not hold true for many functionals. Accordingly, segments of E(N) have a certain curvature (could be both positive or negative) referred to as (de)localization error, 51 for a detailed discus-sion see Ref. 41. Tuned range-separated hybrid functionals usually result in small E(N) curvatures ensuring small delocalization error. However, to assist an unambiguous choice of optimal parameters, we have used the curvatures of the E(N)-dependence for fractional charges and have chosen, whenever possible, the (α, ω)-pair whose curvature is closest to zero. As a particular measure of the curvature we have chosen to indicate the deviation from the idealized linear dependence, see also Fig. 3b. All functional tuning calculations have been performed using the 6-31G(d) basis set for all atoms. The initial geometries have been optimized using the LC-BLYP functional with ω=0.17 bohr −1 which is typical for the complexes of this size. 32 We note that the optimized LC-BLYP geometries are nevertheless very similar (averaged RMSD for all five complexes is 0.21, see Cartesian Coordinates in the Supporting Information † ) to the geometries optimised with the TPSSh 52,53 functional employed in Ref. 8. As the inclusion of an implicit solvent in this tuning procedure has been shown to deliver erroneous results, 32,54 the optimal tuning process has been done in vacuum. The step size for varying α has been 0.05 and for ω 0.01 bohr −1 . In order to calculate the final electronic excited state energies and wavefunctions, the geometries of all investigated complexes have been reoptimized employing the chosen tuned LC-BLYP functionals with corresponding α and ω values and the larger basis set def2-TZVP. 55 On these geometries, TDDFT computations were carried out with the optimally tuned LC-BLYP functional. Comparison with multiconfigurational reference data (see below) suggested an adjustment of the obtained (α, ω)-pairs. A total of 50 singlet and 50 triplet TDDFT excited states have been calculated, expecting that this number of states is enough to capture the lowest energy region of the spectrum adequately. Solvent effects (acetonitrile) are included implicitly within the polarized continuum model approach. 56 In order to artificially add broadening to the spectra, each of the excitations have been convoluted with a Gaussian function (for the FWHM see Figure 4). All tuning calculations were done with the Q-Chem 5.1 package 57 whereas further geometry optimizations and absorption spectra computations were done with Gaussian 16. 58 Excited state analysis has been done using with the TheoDORE package, 59 which enables automatic quantitative wavefunction analysis and straightforward assignment of excitation localization at predefined molecular moieties. 60 For the two smallest complexes of Fig. 1 (4 and 5), complete active space self-consistent-field 61 (CASSCF) and CAS second order perturbation theory 62,63 (CASPT2) calculations with relativistic ANO-RCC-VTZP 64 basis set have performed as a reference using OpenMolcas. 65 The highest possible Abelian point symmetry group (C 2 for 4 and D 2 for 5) and Cholesky decomposition have been utilized. To account for both MCLT and MC states, the active space comprised 10 electrons in the following 12 orbitals (10e/12MO): three essentially non-bonding 3d-orbitals of the iron atom as well three corresponding Rydbergs-like 4d orbitals to account for the double-shell effect, two σ d -bonding and two σ * d -antibonding orbitals, and the two vacant π * -orbitals of ligands (see Fig. 2). State-averaging with equal weights have been done over the 10 lowest states of a given symmetry and multiplicity in CASSCF calculations; selected states have been computed with state-specific CASPT2. The frozen-core approximation have been utilized at the CASPT2 level. A default IPEA shift 66,67 of 0.25 a.u. and an additional imaginary shift of 0.2 a.u. to cope with intruder states problem have been applied. To avoid the comparison of different solvent models in different quantumchemistry packages, the calculations of the lowest transitions in CASPT2 and various DFT flavors (using def2-TZVP basis set) have been done in vacuum. Depending on where electrons and holes are localized within an electronic excitation, transition metal complexes can exhibit MC, MLCT, LC, ligand-to-metal charge transfer (LMCT) or ligandto-ligand charge transfer (LLCT) states. The nature of these states can be assigned by inspecting the orbitals involved in the electronic transition. However, this process is not only subjective but very tedious if many wave function configurations contribute to a single electronic state and on top if many electronic states are to be analysed. In this work, the assignment of the character of the states has been done using charge transfers numbers between groups of atoms defined as fragments. 60 Chemical intuition can be used to partition the molecule, but it is also possible to do a correlation analysis with subsequent hierarchical clustering 60 in order to get a comprehensive understanding of the effect a particular moiety has within a ligand. In such a procedure, charge transfer numbers are first obtained considering every atom as an independent fragment (with the exception of hydrogens that are added to the connected carbon atom and thus treated as one fragment) and then a correlation matrix of the fragments' contribution to the excited states is calculated. Through hierarchical clustering, fragments with high correlation are then combined to obtain an automatic fragmentation of each molecule. In this automatic fragmentation the threshold of the coefficient of determination R 2 can be varied, depending on the graining desired. To ease of analysis the amount of fragments should not be very high, but small enough to make chemical sense. ## Optimization of the Range-Separation Parameters The optimal values obtained for the range-separation parameters α and ω of the Fe(II) complexes shown in Fig. 1 are collected in Table 1. A typical example of a 2D plot of J * (α, ω) resulted from single-point calculations on the grid is presented in Fig. 3a for the complex 1. Only for 3, no minima have been found for the 1D curves constructed at the constant α values. In this case, we have chosen the minimal point on the J 0 (α, ω) 2D surface that correspond to the IP-only tuning. Importantly, the criterion of smallest piecewise curvature does not in general provide a tool to unambiguously select the best (α, ω)-pair, as this curvature is very small for such tuned functionals. An example is given in Fig. 3b for the complex 1, where the curvatures for all (α, ω)-pairs are being two orders of magnitude smaller than for standard density functionals. 50 Based on this criterion, no exact exchange in the short-range should be included in the functional (α = 0) for all five Fe(II) complexes (Table 1). Previously, the fraction of the exact exchange of 0.20 has been generally recommended for non- 1 Optimized range separation parameters, α and ω, for compounds shown in Fig. 1. While α = 0 results from optimal tuning, α = 0.2 has been chosen for the TDDFT calculations after comparison with the CASPT2 results. Compound ω 35,37,68 . In a recent publication 69 , a value of α of 0.10-0.15 has been argued to be optimal for a series of iron spin-crossover compounds after analyzing the adiabatic energy difference between high-and low-spin states based on the comparison with the OPBE reference data 70 implying that the errors inherent to the OPBE functional may also influence this conclusion. With the constant amount of global exact exchange, the values of ω for all compounds are very close to each other, what can be rationalized by the comparable size of the ligands. The inverse value of range-separation parameter ω −1 reflects a characteristic distance for switching between short-and long-range parts or, in other words, an effective electron screening (delocalization) length. Previously, optimal ω values were found to decrease with increasing system size and conjugation length. 29,36, For Ir(III) photosensitizers of comparable size, similar ω-values of 0.14-0.18 bohr −1 have been found. 32 ## Energetics of the lowest-lying electronic excited states Table 2 collects selected energies obtained with all the methods employed in this work for complexes 4 and 5. For this comparison, only few states -the lowest of dominantly MLCT and MC character (see systematic analysis in Section 3.4) -in both singlet and triplet manifolds have been selected; note that these states are not necessarily the two energetically lowest ones. For the tuned LC-BLYP case, three different optimal pairs of parameters (α, ω) have been considered: no constant exact exchange (α=0.0) as it was predicted by analysis of fractional charges, as well as 15 and 20 % of exchange because those parameters have been proposed in earlier publications. 69,70 Taken CASPT2 as a reference, one can clearly see that the purely local BLYP drastically underestimates the energies of MLCT states but overestimates the MC states. Importantly, the energies of MC states predicted with the tuned LC-BLYP (α=0.0) almost coincide with those of BLYP, supporting the underlying assumption that at short interelectron distances the local BLYP functional dominates in Eq. (1). At the same time, it is evident that the inclusion of certain portion of exact exchange in the optimally-tuned functional may have a crucial effect on the position of MC states. Indeed, if various α values in LC-BLYP are compared, one can see that the inclusion of increasing portions of exact exchange lowers the MC energies towards the CASPT2 reference. Simultaneously, the MLCT energies agree better with the reference. The popular B3LYP functional performs comparable to the LC-BLYP (α=0.20), as both functionals include 20% of exact exchange. The other popular long-range functional chosen, CAM-B3LYP, gives reasonable energies for MC states but overestimates the MCLT energies. Clearly, the test set for the comparison of the lowest energies is not large enough to deduce ultimate conclusions but the same trends have been observed for the test calculations for all studied complexes with the 6-31G(d) basis set (not shown for the sake of brevity). In addition, it is fair to keep in mind the possible inaccuracies of CASPT2 connected with the moderate size of active space and incomplete account for electron correlation. As such, the comparison above should be taken with a pinch of salt. ## Vertical excitation spectra The absorption spectra for all the complexes has been calculated using the LC-BLYP functional with α = 0.20. This α value was chosen in accordance with the above results to reasonably describe the important MC states. The corresponding ω values can be found in Table 1. The convoluted spectra (blue line) are compared against the experiment 8 (black line) in Fig. 4. The blue sticks indicate the most important absorbing states calculated at the equilibrium geometry, from which the convoluted spectra has been obtained. In general, all calculated spectra show decent agreement with the experiment but they are shifted to higher energies by approximately 0.5 eV. In what follows, we shall discuss this comparison with some detail. In compound 1, the shape of the first computed band at 2.8 eV is less broad than in the experiment and the separation between this and the next band, with an onset at 3.6 eV is larger than it should; this peak is also predicted to be closer than it should to the next one at 4.0 eV. We refrain from discussing the results at higher energies, as the number of electronic excited states included are not sufficient to map the full available experimental spectrum. In terms of shape, compound 2 shows the largest disagreement. The first experimental band of the absorption spectrum, peaking at 2.75 eV followed by a smaller peak at 3.30 eV, is not well described by theory, which shows a band from 3.25 to 4.00 eV with two peaks that experimentally are much closer together. Further, the intensities are also incorrect, with the strengths of the peaks being reverse. Noticeable is also a peak from 4.25 to 4.75 eV, which is not present in the experimental spectrum. The calculated spectrum for 3 shows two separated bands from 2.40 to 4.00 eV, which are broader in the experiment and with less separation. Part of these inaccuracies are due to the fact that our spectra are simulated at the limit of using only the equilibrium geometry, while nuclear motion affects the transition energies beyond the broadening that a Gaussian convolution would suggest. 78,79 The compound 4 shows the best agreement with the experiment. The largest discrepancy is an additional peak at 4.00 eV. This shoulder cannot be seen in the experiment. The calculated spectrum for 5 also agrees well with the experiment. The peak at 4.25 eV is closer to the first band than in the experiment, and includes a second smaller peak at 4.50 eV that cannot be seen in the experiment. ## Electronic state character and correlation analysis In order to identify the electronic character of the excited states, systematic wavefunction analysis was used, where the compounds were partitioned first as the sum of three fragments: the Fe metal center and the two ligands (the polypyridyl and/or Nheterocyclic carbene (NHC) units). Fig. 5 shows the resulting assignments in each of the 50 singlet and 50 triplet states (left panels) as well as their weight to the absorption spectra (right panels) for all complexes 1-5. In all complexes, the lowest twenty singlet transitions are dominated by MLCT (blue boxes) and a few MC (red) excited states. For 5 these are followed by more MLCT transitions and some LC (green) and LLCT (orange) states. Interestingly, in the complexes with NHC groups 1-4 more LC and LLCT states can be seen at higher energies. If one now looks into how these states contribute to the absorption spectrum, it is clear that the low energy band is of predominant MLCT character (blue line) while the higher has contributions of both MLCT and LLCT (orange) states. The MC and LC states are mostly dark and the LMCT are in a small amount. Inspection of the triplet states shows a similar pattern as the singlets. The lowest triplet states have strong MC character, followed by MLCT states and then mostly LC and few LLCT states at higher energies. The lowest MC states in 5 are clustered together and have almost identical, high coefficients for the MC state. They are followed by states with almost no MC contributions. This is not true for the complexes with NHC ligands. For complexes 1 and 4, both with two NHC donors, some differences can be appreciated. In 1 the singlet MC states are lower than in 4 but the low energy triplets of both complexes are of MC character, except for T5 in 1. This MLCT state breaks up the series of MC states. In 3, with 3 NHC donors, the cluster of low energy MC states is broken up by two MLCT states. Finally in the compound with 4 NHC donors, 2, the MC states cluster even less. Instead, several MLCT states are lower in energy than half of the MC states. This behaviour is in agreement with the results of Ref. 8 where an increased number of NHC donors led to a destabilisation of the high-spin MC states. This in turn leads to increased 3 MLCT ## lifetimes. Albeit informative, the previous fragmentation scheme with all equivalent ligands prevents one to know which ligand in particular is involved in the excitation or whether the charge prefers a particular region within the ligand. Moreover, this particular fragmentation scheme has been enforced based on chemical intuition, but not on the actual distribution of density within the molecule. If instead a correlation analysis with subsequent hierarchical clustering is done the fragmentation depicted in Fig. 6 is obtained. With two identical polypyridyl ligands, the homoleptic complex 5 is the simplest of all the complexes and as expected, the automatic fragmentation procedure provides three fragments, the Fe center and the two ligands as separate fragments, as chemical intuition would do. Despite also being homoleptic, this is not the case for 4. Here the automatic fragmentation separates the NHC-rings of the ligands and collects them in one fragment. The other two rings of each ligand are considered a single fragment. With the same thresholds, the results for 3 initially led to five fragments. One consisted of the 2,6-diisopropylphenyl (dipp) groups alone. However, this fragment does not play a big role in the excitations and thus has low correlation with the other fragments. Therefore, the dipp-groups were added to the fragment with their neighbouring NHC-rings. The other ligand is split into two parts, the NHC-ring and the polypyridyl rings, making the four fragments that are depicted in Fig. 6. The results and fragmentation scheme for 2 are similar to those of 4: The NHC-rings are separated and collected into one ligand while the two central pyridyl rings form the other ligand for the analysis. The fragments obtained in 1 can be compared to those for 3: The NHC ligands and the central pyridyl ring connecting them are collected in one fragment. The unsubstituted terperydine ligand is collected as another fragment. In contrast to 3 the clustering favours adding the dipp groups to the terpyridine ring of the other ligand, not the neighbouring NHC groups. In conclusion, the correlation analysis provides two important messages. The first is that the NHC groups have a clear effect in the photophysics of these complexes. The second is that this effect is different depending on how the carbenes are incorporated into the ligands -and this effect is difficult to predict a priori. Interestingly, the charge density is not necessarily fully delocalized over the whole ligand, but different NHC fragments induce localized excitations and should be considered as independent fragments. Note that, the central iron atom is always considered as one fragment for the analysis so that the MC, LMCT and MLCT transitions can still be identified as it is conventional in coordination chemistry. In contrast, LC and LLCT states in the following results do not correspond directly to conventional LC and LLCT states because the ligands are broken and clustered in a different way than intuition might dictate. However, it is always possible to consider their sum, reaching the same rough assignment as one would do in Fig. 5. In the following, we will analyse the particular roles of each of these fragments by inspecting the corresponding CT numbers among them, as depicted in Fig. 7. We start with 5 because with two symmetric and identical ligands is the simplest and can be used as a reference. Its two ligands are labeled as L 1 and L 2 . As expected, the ligands contribute symmetrically to most excitations, which for simplicity are indicated by different shades of the same color. States with asymmetrical contributions can usually be found as pairs of degenerate states. These states have almost identical energies (see Table 6 in the SI †) and contributions, but the L 1 and L 2 terms are switched. Example pairs are S2 and S3, S4 and S5. In 4, the polypyridyl rings are collected into L 1 and the NHC rings in L 2 . One can see that CT to the NHC rings plays a role in virtually all of the MLCT contributions, justifying this separation. In 3, it can be seen that the single unsubstituted NHC ring in L 2 does not play a role in most of the excited states. In particular, it contributes to almost no MLCTs. The MC triplet states higher in energy show no ML 1 CT or ML 2 CT contributions, indicating that the dipp-substituted NHC-donors of L 3 are more important here. The four NHC rings of 2, collected in L 2 , contribute to most MLCT transitions at least partially. In addition it can be seen that MC states with larger ML 2 CT than ML 1 CT contributions are higher in energy. This is similar to the results for 3, were the substituted NHC-donors seem important for the increasing MC energies. CT from L 2 occurs in the higher energy states to the pyridyl rings of L 1 . Compound 1 has the least contributions of the fragment with NHC donors, L 2 . It can be clearly seen that the highest MC states of both the singlets and triplets have ML 2 CT contributions. MLCT states characterised by contributions of L 2 are also generally higher in energy than those with charge transfer to L 1 . The automatic fragmentation clearly identified the NHC groups as separate contributors to the excitations. Those contributions were generally larger for higher energy excitations. Both lowenergy MLCT and MC states are generally higher in energy if they include contributions of charge transfer to a NHC fragment. Nevertheless, this effect seems to be stronger for the MC states. The reference molecule with no NHC donors 5 has clustered MC states as the lowest states of both singlet and triplet excitations. In contrast, with increasing number of NHC donors more MC states are shifted higher than some MLCT states in both singlets and triplets. ## Conclusions In this work, we have investigated five Fe(II) homoleptic and heteroleptic complexes that include strong sigma-donating Nheterocyclic carbene and polypyridyl ligands with the aim to characterize the effect of the carbene moiety on the position of MLCT and MC excited states. For this purpose, the absorption spectra of the complexes was calculated with an optimally-tuned range separation functional, LC-BLYP. To benchmark the LC-BLYP functional beyond the usual conditions for optimal tuning, selected low-lying singlets and triplet excited states of MLCT and MC character were computed in gas phase at different levels of theory. Methods besides the LC-BLYP functional where the influence of the α and ω tuning parameters was explored, include the BLYP, B3LYP and CAM-B3LYP functionals as well as CASPT2, which was taken as a reference. The results indicate that none of these DFT approaches are completely satisfactory for these complexes. Compared to the CASPT2 reference, the tuned LC-BLYP with a constant portion of exact exchange of 20% was taken as a compromise. This provided provided a reasonable description of the absorption spectra of the complexes with which the comparison was made and spectra of different quality in the rest of the com-plexes. The discrepancies between theory and experiment point to the fact that tuning towards particular transitions (two lowlying MC and MLCT CASPT2 states) does not necessarily provide a uniform improvement across the spectrum, hinting to strong differential correlation 80 in these complexes. The analysis of the electronic excited states underlying the absorption spectra with the chosen tuned LC-BLYP functional shows that excitation occurs to singlet states of predominant MLCT character. As expected for non-emissive iron complexes the triplet excited states have MC nature in the lowest part of the absorption spectrum, followed by MLCT states. Increasing the number of NHC-donor groups leads to a blue shift of only a part of these triplet MC states. But, some of them stay in the low-energy region thus preventing the presence of long-lived MLCT states. Overall, this paper also highlights the usefulness of quantitative wave function analysis and hierarchical clustering, as it reveals how the carbene rings play an important individual role in the excitations, by localizing part of the electron density in the excitation. We expect that the gained insight can be useful in the design of alternative Fe(II) complexes with long-lived MLCT states. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "The effect of N-heterocyclic carbene units on the absorption spectra of Fe(II) complexes: a challenge for theory \u2020", "journal": "ChemRxiv"}

photoanodic_and_photocathodic_behaviour_of_la₅ti₂cus₅o₇_

## Abstract: The particulate semiconductor La 5 Ti 2 CuS 5 O 7 (LTC) with a band gap energy of 1.9 eV functioned as either a photocathode or a photoanode when embedded onto Au or Ti metal layers, respectively. By applying an LTC/Au photocathode and LTC/Ti photoanode to, respectively, photoelectrochemical (PEC) water reduction and oxidation concurrently, zero-bias overall water splitting was accomplished under visible light irradiation. The band structures of LTC/Au and LTC/Ti calculated using a semiconductor device simulator (AFORS-HET) confirmed the critical role of the solid/solid junction of the metal back contact in the charge separation and PEC properties of LTC photoelectrodes. The prominently long lifetime of photoexcited charge carriers in LTC, confirmed by transient absorption spectroscopy, allowed the utilization of both photoexcited electrons and holes depending on the band structure at the solid/solid junction. ## Introduction The exploration and development of solar energy related technologies has attracted much interest in recent years because of the global concern for environmental problems and depletion of fossil energy resources. Photoelectrochemical (PEC) water splitting is one of the ideal approaches to converting solar energy into H 2 as a clean and renewable chemical resource. To fabricate a photoelectrode for PEC water splitting, semiconductor thin flms are deposited on a conductive layer. Because of the band bending caused by charge carrier diffusion between the semiconductor and the electrolyte at the solid/ liquid junction to establish thermal equilibrium, in general, ntype semiconducting materials function as photoanodes for water oxidation, while p-type semiconducting materials function as photocathodes for water reduction. 1,2 In addition, the backside conductive layers are chosen so as to form an ohmic contact for the majority carriers of semiconductors at the backside solid/solid junction in order to avoid undesirable rectifcation of carrier transport. 2,3 To achieve unassisted PEC water splitting, also referred to as artifcial photosynthesis, combinations of photoelectrodes and photovoltaic cells have also been studied recently. 4 La 5 Ti 2 CuS 5 O 7 (LTC) is an oxysulfde semiconductor material. Diffuse reflectance spectroscopy reveals that the light absorption edge of LTC is 650 nm, which is equal to a band gap energy of 1.9 eV. 6 Our group has reported that LTC exhibited photocatalytic activity for both water reduction and oxidation under visible light irradiation in the presence of sacrifcial reagents, 6,7 which implied that the conduction band of LTC was more negative than the hydrogen evolution potential, E(H + /H 2 ), and that the valence band was more positive than the oxygen evolution potential, E(O 2 /H 2 O). Besides, the abundances of the elements Ti, Cu, and La in the earth's crustal rocks are 6320, 68, and 35 ppm, respectively, which are much higher than those of the elements widely used in semiconducting materials for photovoltaic and PEC applications, such as Ga (19 ppm) and In (0.24 ppm). 10 Thus, LTC is a prospective candidate for solardriven PEC water splitting. Our recent study revealed that the photocathodic current of LTC photoelectrodes fabricated by the particle transfer (PT) method was improved by a factor of eight by doping 1 mol% Sc into the LTC powder precursor. 8 In the present work, we studied the PEC properties of LTC photoelectrodes prepared by the PT method using different metal back contacts in the water splitting reaction. By exploiting the PT method, a monolayer of LTC particles was embedded into Au or Ti metal layers. 3,11 It was found that LTC photoelectrodes with Au and Ti as back contact layers showed photocathodic and photoanodic current under light irradiation attributable to PEC hydrogen and oxygen evolution, respectively. Furthermore, visible light irradiation of LTC/Ti and LTC/Au connected with each other drove PEC overall water splitting at a moderate Faradaic efficiency ($80%) without the aid of an external voltage or photovoltaic cells. The unique behaviour of LTC photoelectrodes can be explained by the relative barrier heights at the LTC/electrolyte and LTC/metal junctions. ## Experimental LTC powder was prepared by a solid state reaction as described in the ESI. † 5,6 X-ray diffraction patterns (Fig. S1 in ESI †) and scanning electron micrographs (Fig. S2 in ESI †) revealed that the produced LTC was a well-crystallized rod-shaped material. The diameter and length of LTC rods ranged from 0.7 to 1.4 mm and 2 to 6 mm, respectively. To fabricate photoelectrodes by the PT method, the LTC powder was densely dispersed on glass plates (Scheme S1 in ESI †). A thin ($2.5 mm) metallic layer of Au or Ti was deposited on top of stacked LTC particles by evaporation (for Au) or radio-frequency (RF) magnetron sputtering (for Ti). In this process, the top layer of LTC particles was embedded into the Au or Ti flm, which ensured an intimate electrical and physical contact between the semiconductor particles and the metal layer. The metal layer was bonded to a second glass plate by double-face tape and then peeled off of the primary glass plate. Thus, photoelectrodes of LTC/metal (Au or Ti)/tape/glass plate were fabricated. The LTC/Au electrode was loaded with Pt (1 nm) by sputtering and the LTC/Ti electrode was loaded with cobalt phosphate (CoPi) by electrodeposition 12 (see ESI †) to enhance PEC H 2 or O 2 evolution on the surface. Current-potential curves for the PEC water splitting reaction were acquired in an aqueous solution containing 0.1 M Na 2 SO 4 with a pH value of 12 adjusted by NaOH. A 300 W xenon (Xe) lamp equipped with a mirror module was used to irradiate visible light (420 nm < l < 800 nm). The power spectra of the Xe lamp and the standard AM1.5G irradiation are shown in Fig. S1. † The photon flux of the Xe lamp was 2.1 10 18 photon s 1 cm 2 in the wavelength region shorter than 650 nm, the absorption edge wavelength of LTC, and was roughly 19 times stronger than that of sunlight. ## Results Fig. 1 shows the current-potential curves of LTC/Au and LTC/Ti photoelectrodes during PEC water splitting under chopped light irradiation in an aqueous solution at pH 12. The photoresponses of the two electrodes were different from each other: the LTC/Au photoelectrode showed a photocathodic current typical of p-type semiconductor photoelectrodes (Fig. 1a), while the LTC/Ti photoelectrode exhibited a photoanodic current characteristic of n-type semiconductor photoelectrodes (Fig. 1c). The onset potentials of the photocathodic current on bare LTC/Au and the photoanodic current on bare LTC/Ti were 0.6 and 0.1 V vs. RHE, respectively. In addition, photoanodic and photocathodic currents were observed on LTC/Au and LTC/Ti when the electrode potentials were more positive than 0.6 V vs. RHE and more negative than 0.1 V vs. RHE, respectively. The photocathodic and photoanodic currents observed on the LTC/Au and the LTC/Ti photoelectrodes, respectively, suggest the occurrence of PEC H 2 and O 2 evolution on the surface via eqn (1) and ( 2): (1) Loading Pt as a H 2 evolution catalyst and CoPi as an O 2 evolution catalyst on the surface of LTC/Au and LTC/Ti increased the photocathodic and photoanodic currents, respectively, as shown in Fig. 1b and d. As a result, the difference between the photocathodic and photoanodic currents was increased. To confrm PEC gas production on the respective photoelectrodes, a sealed two-electrode system was constructed using Pt/LTC/Au or CoPi/LTC/Ti as the working electrode and Pt wire as the counter electrode. A half and a quarter of the total charge passing as photocathodic and photoanodic current correspond to the amounts of H 2 and O 2 generated during PEC water splitting, respectively, given that the Faradaic efficiency is unity. A comparison of the measured amount of gas generated on a Pt/LTC/Au photocathode and the amount calculated from the total charge is shown in Fig. 2a, along with the current-time profle in the inset. A photocathodic current was generated continuously on the Pt/LTC/Au photoelectrode under light irradiation at 0.9 V vs. counter electrode and stopped immediately when the illumination was turned off. The Faradaic efficiency of the Pt/LTC/Au photocathode was 97% for PEC H 2 production over 20 h. Fig. 2b shows the results of gas detection on a CoPi/LTC/Ti photoanode in PEC water oxidation at 0.8 V vs. counter electrode. After a 23 h reaction, the Faradaic efficiency of the CoPi/LTC/Ti photoanode for O 2 production was estimated to be 79%. The comparatively lower efficiency of the CoPi/LTC/Ti photoanode was most likely due to the partial photooxidation of sulphide ions in LTC. As shown in Fig. 1b and d, the onset potentials of the photocathodic and photoanodic currents were 0.8 and 0.1 V vs. RHE for Pt/LTC/Au and CoPi/LTC/Ti photoelectrodes, respectively. Therefore, unassisted PEC water splitting was feasible using these two photoelectrodes connected in series. Fig. 2c shows the time course of H 2 and O 2 evolution and the current-time curve (in the inset) for Pt/LTC/Au and CoPi/LTC/Ti photoelectrodes connected in series. During the frst 5 h, a photocurrent was generated without any additional external voltage under visible light irradiation. Subsequently, an external voltage of 0.5 V was applied to increase the photocurrent and to observe gas production more easily. As shown in Fig. 2c, the evolution of H 2 and O 2 occurred concurrently with the observation of photocurrent. After 20 h of illumination, the Faradaic efficiencies for PEC H 2 and O 2 evolution on Pt/LTC/Au and CoPi/LTC/Ti photoelectrodes were calculated to be 95% and 78%, respectively. These values were close to the Faradaic efficiencies observed when either of the photoelectrodes and a Pt counter electrode were used. ## Discussion It was found that the PEC behaviour of LTC photoelectrodes varied depending on the type of back metal conductor, even though the LTC powder used was the same. Specifcally, the use of Au and Ti as the back metal contacts resulted in the generation of photocathodic and photoanodic currents at relatively positive (0.8 V vs. RHE) and negative (0.1 V vs. RHE) potentials, respectively. Moreover, because the potential regions where the photocathodic and photoanodic LTC electrodes could generate photocurrents overlapped, Pt/LTC/Au and CoPi/LTC/Ti photoelectrodes connected in series were capable of splitting water into H 2 and O 2 at a moderate Faradaic efficiency ($80%) for oxysulfde photoelectrodes. The importance of the back metal conductor in the PEC properties such as the photocurrent and the onset potential of photoanodic current has been discussed in terms of the formation of Ohmic, Schottky, and Bardeen-type junctions in earlier studies on titanium oxide photoanodes. 13,14 However, the switching of the photoanodic response to photocathodic response was not observed although metal conductors with various work functions were used. In nanoparticulate systems such as PbS and copper zinc indium sulphide (ZCIS) quantum dots, in which the development of a depletion layer is not generally considered, both photoanodic and photocathodic are sometimes observed depending on the electrode potential. 15,16 A photocathodic response was observed for PbS photoelectrodes when the electrode potential was once held at a positive potential, presumably because the surface trap states were unflled with electrons and vice versa. 15 Electrochemical stripping of Cu + ions occurred in CZIS quantum dots when the electrode potential was held at a positive potential. Because of the formation of Cu vacancies, the electrochemically-treated CZIS photoelectrode showed enhanced photocathodic response. 16 However, the situation is likely different for photoelectrodes consisting of micrometre-sized LTC particles in which the band bending at the interface generally rectifes migration of carriers. In the following section, we will discuss the characteristic PEC behaviour and the onset potential on the basis of band diagrams, taking account of the properties of the back metal contact. For a qualitative understanding of the effect of metal substrates on the PEC properties of LTC, the band structures of electrolyte/LTC/Au and electrolyte/LTC/Ti were analysed using a semiconductor device simulator (AFORS-HET 17 ). The material parameters for LTC are shown in Fig. 3a: band gap E g ¼ 1.9 eV 6 and the top of the valence band E v ¼ 5.7 eV (see Fig. S3 †). The work functions of Au and Ti were reported to be 5.1 and 4.33 eV, respectively. 18 The electrolyte can be treated as a hypothetical metal with a work function between 4.4 eV (the hydrogen evolution potential) and 5.67 eV (the oxygen evolution potential). 19 In the following discussion, only the case of an electrolyte potential of 4.4 eV is described for simplicity because, qualitatively, the same conclusions are reached regardless of the potential difference between 4.4 and 5.67 eV. The other parameters used in the calculations are listed in Table S1. † Fig. 3b and c show the equilibrium band diagrams of LTC in contact with the electrolyte on one side and Au or Ti on the other. Although LTC was previously studied as a photocathode, 8,9 the LTC photoelectrodes function as both photocathode and photoanode depending on the kinds of back contact metals. Therefore, three types of conductivity should be considered for the LTC semiconductor, viz., n-type, intrinsictype, and p-type, with, respectively, the following Fermi levels (E F ): 4.0, 4.75, and 5.4 eV (relative to the vacuum level). The black, red, and blue curves represent the conduction band, valence band, and the Fermi level; the solid, dotted, and dashed curves correspond to the cases in which LTC is assumed to be a p-, intrinsic-, and n-type semiconductor, respectively. Note that the barrier height, f b , at interfaces was assumed to be f b ¼ f c, where f is the work function of the contact material and c is the electron affinity of the semiconductor. 20 Depending on the energy difference between the Fermi level and the work function of the contact materials, the band structure of LTC is expected to rearrange as shown in Fig. 3b and c. Taking the case of the p-type (n-type) LTC as an example, a two-sided downward (upward) bend is formed because the majority carriers in LTC diffuse into both the electrolyte and the metal layer. However, because of the different work functions of the Au and Ti substrates, the difference in band-bending height between the LTC/electrolyte and LTC/Au junctions is larger than the difference between the LTC/electrolyte and LTC/Ti junctions. Fig. 3d and e plot the band diagrams of LTC/Au and LTC/Ti electrodes under AM1.5G illumination at varying applied biases with the conditions that the photons having energies larger than the band gap of LTC are all absorbed and LTC is a p-type semiconductor. The solid and dotted curves in Fig. 3d and e depict the band structures at the potential with no bias (E ¼ 0) and at the potential where no photocurrent is generated, which corresponds to the onset potential (E ¼ E onset ). The pale black and red lines indicate, respectively, the conduction and valence bands for E ¼ E onset AE 0.5 V. For the LTC/Au electrode (see Fig. 3d), the cathodic current would be expected to occur at E ¼ 0 because the electron (hole) energy level at the electrolyte side is more positive than that at the metal substrate side. As the electrode potential shifts positively, the electron energy level on the metal substrate side also shifts positively. The cathodic current vanishes at E ¼ E onset , and the photoanodic current eventually starts to flow with a further anodic shift in the electrode potential. On the other hand, for the LTC/Ti electrode shown in Fig. 3e, E onset should be very close to zero because the electron (hole) energy level at the metal substrate side is hardly more negative than that of the electrolyte. The plots in Fig. 3d and e match well with the PEC behaviour shown in Fig. 1. In summary, the relative barrier heights at the LTC/electrolyte and LTC/metal junctions dominate the directions of the photogenerated charge migration, which eventually leads to the photocathodic and photoanodic response on LTC/Au and LTC/ Ti, respectively. Note that, although the present study is based on the p-type LTC, the dependence of E onset on the metal substrate type can be explained in a similar manner for intrinsic-type and n-type LTC. It is generally not believed that photoexcited electrons and holes generated in the same semiconductor are both used for PEC reactions, because the semiconductor/liquid junction rectifes the direction of charge migration. The unique properties of LTC are possibly due to the special band and crystal structures that allow charge migration over a long range. The density functional theory (DFT) calculations in our previous study suggested that the valence band and the conduction band of LTC were spatially localized around one-dimensional chains of CuS 4 tetrahedra and Ti(O,S) 6 octahedra, respectively. 6 Such a structure was considered to be favourable for charge separation and transport. In fact, it was recently suggested that the charge migration distance in LTC was on the order of microns in the case of LTC/Au photocathodes, because PEC deposition of noble metal nanoparticles occurred only onto the tip of rod-like LTC particles a few micrometres in length. 9 In order to further investigate the nature of photogenerated charge carriers in LTC powder, we employed transient absorption spectroscopy (TAS), which has been successfully used in monitoring carrier dynamics in many semiconductor photocatalysts 21,22 (see ESI † for experimental details). Since the powder was opaque, we collected and analyzed the diffuse reflected light from the sample, instead of transmitted light, to obtain information on the kinetics of transient species; femtosecond time resolved diffuse reflectance (fs-TRDR) spectroscopy was employed. The transient time profle of the LTC powder probed at 900 nm (l exc ¼ 535 nm) shown in Fig. 4 suggests the presence of longlived charge carriers. The experiment was performed under weak excitation conditions (pump power z 0.1 mJ) to rule out the second-order electron-hole recombination processes. As is evident from Fig. 4, even at a reasonably long delay time of 2000 ps, ca. 78% of photogenerated carriers survived. Thus, recombination was much slower in LTC than in other visiblelight-driven photocatalysts such as GaN:ZnO (66% at 500 ps), TaON (28% at 500 ps), Ta 3 N 5 (12% at 100 ps), and LaTiO 2 N (50% at 270 ps). 21,22 Even though at this preliminary stage of investigation we cannot conclusively assign the observed transient absorption to electrons or holes, such a long photoexcited carrier lifetime hints toward the validity of the long charge migration distance estimated in our previous study 9 and possibly also explains the sensitivity of the PEC properties of LTC electrodes to the backside solid/solid junction. Despite the long lifetime of the photoexcited state and the crystal and the band structure being favourable for charge separation and transport, the photocurrent observed in this study was not high or stable. Therefore, signifcant advancement in the PEC properties of LTC photoelectrodes is needed to make the best of the unique properties of the material. It was found that the photocathodic current could be increased almost by an order of magnitude by p-type doping of LTC. 8 Signifcant enhancement in the photocathodic and the photoanodic responses can be thus expected by appropriate compositional modifcations. On the other hand, the decrease of the photocurrent due to photocorrosion was more critical for the CoPi/ LTC/Ti photoanode. Therefore, upgrading the stability of LTC photoanodes is demanded. Recently, surface modifcations by amorphous titanium oxide layers were reported to make some non-oxide photoanodes durable during the PEC water oxidation reaction. 23 Advancement in such surface modifcation techniques for LTC photoanodes could offer a solution to the stability issue. ## Conclusion The PEC properties of LTC/Au and LTC/Ti photoelectrodes fabricated by the PT method were discussed in connection with water splitting. Under light irradiation, LTC/Au acted as a photocathode for H 2 evolution, while LTC/Ti acted as a photoanode for O 2 evolution. By combining the LTC/Au photocathode and LTC/Ti photoanode, zero-bias overall water splitting was achieved under visible light irradiation. Using a semiconductor device simulator (AFORS-HET), the p-type and n-type characteristics observed for LTC/Au and LTC/Ti photoelectrodes, respectively, were ascribed to the difference in barrier height between the LTC/electrolyte and the LTC/metal under light irradiation. As already mentioned above, LTC exhibited a carrier lifetime much longer than some of the other commonly investigated visible-light-driven photocatalysts. This may account for the uniqueness of this material. Our work on LTC opens up a new possibility, namely, that the conductivity type of a semiconductor photoelectrode can be controlled by engineering the barrier heights not only at the semiconductor/liquid junction, but also at the semiconductor/back metal junction.

chemsum

{"title": "Photoanodic and photocathodic behaviour of La₅Ti₂CuS₅O₇ electrodes in the water splitting reaction", "journal": "Royal Society of Chemistry (RSC)"}

separated-pair_approximation_and_separated-pair_pair-density_functional_theory

## Abstract: Multi-configuration pair-density functional theory (MC-PDFT) has proved to be a powerful way to combine the capabilities of multi-configuration self-consistent-field theory to represent the an electronic wave function with a highly efficient way to include dynamic correlation energy by density functional theory.All applications reported previously involved complete active space self-consistent-field (CASSCF) theory for the reference wave function. For treating large systems efficiently, it is necessary to ask whether good accuracy is retained when using less complete configuration interaction spaces. To answer this question, we present here calculations employing MC-PDFT with the separated pair (SP) approximation, which is a special case (defined in this article) of generalized active space self-consistent-field (GASSCF) theory in which no more than two orbitals are included in any GAS subspace and in which intersubspace excitations are excluded. This special case of MC-PDFT will be called SP-PDFT. In SP-PDFT, the electronic kinetic energy and the classical Coulomb energy, the electronic density and its gradient, and the on-top pair density and its gradient are obtained from an SP approximation wave function; the electronic energy is then calculated from the first two of these quantities and an on-top density functional of the last four. The accuracy of the SP-PDFT method for predicting the structural properties and bond dissociation energies of twelve diatomic molecules and two triatomic molecules is compared to the SP approximation itself and to CASSCF, MC-PDFT based on CASSCF, CASSCF followed by second order perturbation theory (CASPT2), and Kohn-Sham density functional theory with the PBE exchangecorrelation potential. We show that SP-PDFT reproduces the accuracy of MC-PDFT based on the corresponding CASSCF wave function for predicting C-H bond dissociation energies, the reaction barriers of pericyclic reactions and the properties of open-shell singlet systems, all at only a small fraction of the computational cost. ## Introduction There is strong interest in the development of quantum chemical methods for accurately treating large systems with inherently multiconfgurational electronic structures at affordable computational cost. 1 Such systems are also called multireference systems or strongly correlated systems, and they are usually treated, at least as a frst-order approximation, by mul-ticonfgurational self-consistent feld (MCSCF) methods. 2 This approach includes static electron correlation that would be neglected if a single electronic confguration were employed. In MCSCF methods, one simultaneously variationally optimizes all the orbitals and the coefficients of the various confgurations in a confguration interaction (CI) expansion of the electronic wave function. There are several possible ways to select the confgurations that are included. In the complete active space selfconsistent feld (CASSCF) method, a full confguration interaction (FCI) expansion of the wave function is constructed over an active space of n electrons in N orbitals, with other orbitals double occupied (inactive) or vacant. 3 The size of the FCI expansion grows exponentially as the active space is enlarged, such that an active space with n ¼ 18 and N ¼ 18, labeled as CAS (18,18), is already at the limit of what is affordable. For medium-to-large systems, the active space limit, CAS (18,18), is typically not large enough to describe bond-breaking, electronic excitations, and other chemical properties in a balanced fashion. Thus well-balanced CASSCF calculations are in practice limited to the study of small-to-medium systems. Generally, most of the confgurations in the FCI expansion of the active space in CASSCF computations make only small contributions to the total wave function. As a result, Ruedenberg and coworkers suggested that these confguration state functions (CSFs) are "deadwood" that can be excluded without signifcantly affecting the accuracy of the results. 4 The generalized active space (GAS), 5,6 restricted active space (RAS), 7,8 occupation restricted multiple active spaces (ORMAS) 9 and Split-GAS 10,11 approaches are some of the frameworks that attempt to remove deadwood CSFs by partitioning the active space into subspaces. We have previously shown that active spaces larger than the CAS (18,18) limit can be attained with the generalized active space self-consistent-feld theory, GASSCF. 6,10,11 These MCSCF-type wave functions (CASSCF, GASSCF, etc.) can recover static correlation effects well, but are impractically slowly convergent (with respect to active space size) for the dynamic correlation energy, which is necessary for chemically accurate energetic calculations. For higher accuracy they can be used as zeroth-order reference functions in post-SCF perturbative, multireference coupled-cluster (CC), or multireference confguration interaction (CI) calculations to obtain a good approximation to the dynamic correlation energy. CASPT2 is a popular example that applies second-order perturbation theory to a CASSCF zero-order wave function. 12,13 Such approaches, while capable of high accuracy, 8,14 are however not suited for studying large systems because their computational costs rise rapidly with system size. We have recently proposed an approach for treating strongly correlated systems at much lower computational costs than CASPT2, by combining CASSCF with density functional theory (DFT). This approach is called multiconfguration pair-density functional theory (MC-PDFT). 15,16 It may be considered to be a multiconfgurational analog of Kohn-Sham 16 density functional theory 16,17 (KS-DFT). In KS-DFT, the energy is computed as the kinetic energy and classical Coulomb energy of a Slater determinant (which is a single-confguration reference wave function) and a one-electron integral over an exchange-correlation functional of the one-electron density of the Slater determinant. The classical Coulomb energy includes the nuclear attraction of the electrons, the classical interelectronic repulsion of the electronic charge density, and the nuclear repulsion. The exchange-correlation density functional includes electron exchange, electron correlation, and the difference between the exact kinetic energy and that computed from the Slater determinant. The exact exchange-correlation density functional is unknown, so one uses approximations. In MC-PDFT, the energy is computed as the kinetic energy and classical Coulomb energy of an MCSCF reference wave function and a one-electron integral over an on-top density functional of the one-electron density and the on-top pair density of the reference wave function. The on-top density functional includes electron exchange, electron correlation, and the difference between the exact kinetic energy and that computed from the reference wave function. The MC-PDFT energy may be written as where orbital indices refer to the spatial molecular orbitals, i and j are the doubly occupied inactive orbitals, v, w, x, and y are the active orbitals, h vw and g pqrs respectively one-electron and two-electron integrals, D pq is the one-electron reduced density matrix, V N is the nuclear repulsion, and E ot [r,P] is an on-top density functional of the total density, r, and the on-top pair density, P. Functional expressions for E ot [r,P] when using r and P obtained from an MCSCF solution have been provided in ref. 15 and 18 KS-DFT is usually applied full self-consistently; that is, the exchange-correlation functional is included during the SCF step. MC-PDFT can also in principle be applied fully selfconsistently, but in all work reported so far and in the present article, we carry out the MCSCF calculation by CASSCF without the on-top density functional, and then calculate the fnal energy post-SCF from eqn (1). In this post-SCF mode, MC-PDFT is like the perturbation theory, multireference CC, and multireference CI wave function methods in that it attempts to use an MCSCF method to obtain a balanced reference wave function in an SCF step and to calculate an accurate energy in a post-SCF step. However, in the case of MC-PDFT, the cost of the post-SCF density functional step is negligible (if coded efficiently) compared to the cost of the SCF step, whereas in the wave function methods like CASPT2, the post-SCF step is more expensive than the SCF step. The cost of the SCF step though is still prohibitive for large systems if one uses CASSCF as the MCSCF method. In the present article we test whether MC-PDFT can yield accurate results when based on a GASSCF wave function. In particular, we present a systematic way to choose the active space in GASSCF theory. This new way of choosing the active space is called the separated-pair (SP) approximation. The method is intermediate between generalized valence bond (GVB) theory and complete active space self-consistent-feld (CASSCF) theory. We then use SP and CASSCF as reference wave functions for MC-PDFT. The MC-PDFT method based on a CASSCF and a SP reference wave function will be labeled as CAS-PDFT and SP-PDFT, respectively, when it is desired to distinguish the kind of MCSCF wave function being used as the reference. The next section presents the relevant theory and defnes the separated pair (SP) approximation. We then provide computational details, test sets, results, and discussion. ## On-top density functionals We have previously presented a prescription for translating existing exchange-correlation functionals of KS-DFT to on-top functionals. 15 As an example, tPBE is an on-top pair density functional developed by translating the PBE functional; 15 tPBE is a function of the electron density, its gradient, and the on-top pair density. We have also described a "fully" translated functional called ftPBE that also depends on the gradient of the ontop pair density. 18,19 ## Separated pair approximation The frst step in building a GASSCF wave function is to choose the number m of GAS subspaces and the number and type of orbitals in each GAS subspace. Note that not only in eqn (1) but also in the whole rest of the article, when we talk about orbitals, we are referring to spatial orbitals, not spin-orbitals. We use the notation GAS-m(n,N) for n electrons in N orbitals divided into m subspaces. But this is not a complete specifcation; in addition, for each irreducible representation, one specifes the accumulated minimum and maximum electron occupations after each GAS subspace is added. For a GAS-m(n,N) calculation, the number of electrons in each space, the number and nature of orbitals in each space, and the number of inter-subspace excitations can signifcantly affect the number of CSFs in the CI expansion, andby extensionthe quality of the results obtained. A GAS wave function includes all confgurations that can be defned within the restrictions imposed by the accumulated minimum and maximum electron occupations and by the restriction, if any, on inter-subspace excitations. The effects of these specifcations on the computed properties of various molecules have been previously reported. 5,6,11,20 In the present work, we only use GAS subspaces in which each subspace contains at most two orbitals, and interspace excitations are not allowed. A GASSCF calculation with these restrictions will be called the separated pair (SP) approximation, and when the number of subspaces is m, it will be abbreviated SP-m. If each subspace contains two electrons in two orbitals, this would be specifed in the language of GASSCF as GASm(2m,2m) with the additional specifcation that no intersubspace excitations are allowed. For singlet systems with an even number of electrons, we typically do have two electrons in two orbitals in each subspace, and the two orbitals in a given subspace are usually a bonding orbital and the corresponding antibonding orbital. This is reminiscent of the generalized valence bond perfect pairing (GVB-PP) algorithm, 21 but it is more general. The GVB-PP approximation has subspaces of two electrons in two orbitals coupled to a singlet; this involves two or three confgurations, depending on symmetry. In the SP approximation, when there are two electrons in two orbitals, they may be coupled into either a singlet or a triplet, and the various triplet pairs may be coupled in all possible ways to obtain CSFs with the desired overall spin symmetry of the system (for example, if the overall wave function is a singlet, one may have CSFs where four of the pairs are triplets and all the others are singlets, and the four triplet pairs may be coupled to each other in a variety of ways to obtain an overall singlet); thus the SP approximation involves more possible confgurations than does the GVB-PP approximation. Nevertheless, the SP approximation greatly reduces the number of CSFs in the CI expansion as compared to CASSCF. It is also important to note that we carry out a FCI expansion in each GAS subspace. This is because we allow both singles and double excitations in each subspace containing just two orbitals. The SP approximation is more similar to the generalized valence bond restricted pairing (GVB-RP) approximation 22 than to GVB-PP. A key advantage of SP and GVB-RP is that, unlike GVB-PP, they allow dissociation to high-spin fragments. 21,22 In the SP approximation, every GAS subspace contains one electron in one or two orbitals or two or three electrons in two orbitals, depending on the system. Intersubspace excitations are always excluded. For closed-shell systems (and for open-shell singlets that can be made by breaking a bond in a closedshell system) an SP-m approximation always corresponds to GAS-m(2m,2m). But the value of m depends on which pairs are included in the active space and which are treated as inactive (doubly occupied in all CSFs), and that is an individual choice. For example, we treat the molecular orbitals with parentage in the 2s atomic orbitals as active for C 2 but inactive for N 2 , O 2 , and F 2 . Moss and coworkers, 23 in their GVB-CI calculations on O 2 , also removed the fully occupied 1s g , 1s u , 2s g and 2s u molecular orbitals from the active space. For F 2 , we also treat the molecular orbitals with parentage in the 2p x and 2p y orbitals as inactive. The SP approximation we used for the carbon dimer, C 2 , is shown in Fig. 1. This molecule has a closed-shell singlet ground state, and the orbitals shown in Fig. 1 correspond formally (at equilibrium) to a double bond and a ground state confguration of 2s g 2 , 2s u 2 , 1p ux 2 , 1p uy 2 . This corresponds to a double bond as the occupied 2s u orbital is actually of antibonding character. Within C 1 symmetry, there are 150 CSFs in this reference for the closed-shell singlet, as compared to 1764 CSFs in the analogous CAS (8,8) reference (the analogous GVB-PP wave function would have only 16 CSFs). We note that the SP-4 reference correctly dissociates to two high-spin ( 3 P) carbon atoms, just like the CAS(8,8) reference. The SP scheme for open-shell systems depends on the type of open-shell character. The SP-3 approximation that we used for O 2 is shown in Fig. 2. O 2 differs from C 2 in that the s bonding Fig. 1 The four GAS subspaces used in the SP-4 approximation for the carbon dimer, C 2 . In this scheme, the 2s, 2p z , 2p x , and 2p y atomic orbitals form s g , s u , p(p x ), and p(p y ) (which are bonding or in the case of 2s u , GAS2, antibonding) orbitals respectively as well as their antibonding (or in the case of 2s u , GAS2, bonding) counterparts. These pairs are shown from left to right. The orbitals with an occupation close to two are placed at the top, while those that are nearly empty are placed at the bottom. Two electrons are placed in each GAS subspace. Intra-space excitations (up to double excitations) between a bonding orbital and its antibonding pair are allowed. Inter-subspace excitations between GAS subspaces are not allowed. Fig. 2 The three GAS subspaces used in SP calculations on triplet dioxygen, O 2 . In this scheme, the 2p z atomic orbitals form 3s and 3s* orbitals, and the 2p x and 2p y atomic orbitals form bonding p(p x ), and p(p y ) orbitals and correlating antibonding p*(p x ), and p*(p y ) orbitals. These are shown from left to right. GAS 1 contains two electrons while GAS 2 and GAS 3 each contain 3 electrons. Inter-subspace excitations between GAS spaces are not allowed. This journal is © The Royal Society of Chemistry 2016 Chem. Sci., 2016, 7, 2399-2413 | 2401 combination of 2p orbitals lies higher in energy than the p bonding combination for C 2 but lower for O 2 . In O 2 , as already mentioned, the 2s and 2s* molecular orbitals (which are predominantly formed from the 2s atomic orbitals) are kept inactive. Therefore the SP-3 approximation that we used for O 2 has GAS1 containing two electrons in the 3s g and 3s u orbitals (which are predominantly formed from the 2p z atomic orbitals), GAS2 containing three electrons in the 1p(p x ) and 1p*(p x ) orbitals, and GAS3 containing three electrons in the 1p(p y ) and 1p*(p y ) orbitals. This is a GAS-3 (8,6) reference. It contains 20 CSFs in comparison to 378 CSFs for the full valence CAS (12,8) and 105 CSFs for CAS (8,6). This GAS-3 (8,6) reference also correctly dissociates into two 3 P oxygen atoms. The SP approximations we used for SO and S 2 are isoelectronic to that for O 2 . For the Cr dimer, Cr 2 , we calculated the potential energy curve with an SP-6 approximation, equivalent to GAS-6 (12,12), with the twelve valence orbitals coupled in six GAS subspaces and two active electrons in each GAS subspace. Within D 2h symmetry, there are 1516 CSFs in the SP-6 CI expansion, as compared to 28 784 CSFs in the analogous CAS (12,12) CI expansion. The SP-6 approximation is sufficiently complete that the dimer correctly dissociates to two high-spin ( 7 S Cr) atoms. For methylene triplet or methylene open-shell singlet, a full valence CAS is (6,6). We can think of CH 2 as derived from methane by dissociating two C-H bonds, and the antibonding orbitals associated with those bonds have left with the hydrogens. Thus these systems each have two singly occupied orbitals, which are taken as their own GAS subspaces with one electron in one orbital in each. In addition, they have two GAS spaces that each have two electrons in two orbitals. Thus the separated pair approximation we use is SP-4, which is shorthand in this case for GAS-4 (6,6). There are two important points to note. First, the SP approximation allows one to design GAS subspaces that contain only the bonding and antibonding orbitals necessary to describe a particular process. For example to compute the C-H dissociation energies of acetylene, ethylene and ethane, we included only orbitals relevant to C-H bonding in the SP active space. This formally leads to a SP-3 active space for both acetylene and ethynyl, an SP-4 active space for both ethylene and vinyl, and an SP-6 active space for both ethane and ethyl. We illustrate this feature with several examples. For the ethyl radical, a full valence CAS space would be (13,13) with seven bonds. We think of this as derived from ethane by removing a hydrogen atom, and the antibonding orbital accompanies it. Constructing GAS subspaces with the same logic as explained above for methylene then yields an SP-7 approximation that is equivalent to GAS-7 (13,13). However, when we study C-H bond dissociation in this paper, we treat the C-C bonding orbital as inactive and use an SP-6 approximation corresponding to GAS-6 (11,11). Ethynyl has a full-valence CAS of size (9,9). Since we are interested in C-H bond dissociation, we made the four electrons in p and p* orbitals inactive, which yields an SP-3 approximation equivalent to a GAS-3 (5,5) reference. Vinyl has a full valence CAS of size (11,11). Since we are interested in C-H bond dissociation in, and because we are interested in seeing the effect of aggressively reducing the size of the active space, we removed the four electrons in the C]C bond and the associated s, s*, p, and p* orbitals from the active space, which yields an SP-4 approximation, equivalent to GAS-4 (7,7). Second, the SP-1 approximation is equivalent to CASSCF(2,2), a case which applies to lithium hydride (LiH), as an example. In addition, as we are performing a full CI for each subspace, SP and SP-PDFT are size consistent in so far as the active space is chosen correctly. For all other molecules, the SP approximation used here involves an active pair for all or some of the bonds, as specifed in each case. Nonbonding valence orbitals and core orbitals are always doubly occupied. ## Basis sets The aug-cc-pVTZ basis set is used to describe all the H, Li, B, C, N, O, F and S atoms in the molecules studied in this work. 24 For the Cr dimer we used the ANO-RCC basis set 25 containing [21s15p10d6f4g2h] primitive functions contracted to (10s10p8d6f4g2h). ## Symmetry For the Cr dimer, a D 2h point group was adopted. All other calculations in this work were carried out without symmetry. This is because the method is designed for large molecules that usually have no symmetry so we want to test it in that context. ## CASSCF calculations We used full valence active spaces in CASSCF calculations on all the molecules studied in this work. The exceptions are ozone, for which we used CAS (12,9), a-3-didehydrotoluene and 1,4didehydrobenzene, for which we used CAS (8,8), and the compounds involved in pericyclic reactions for which we included only the subset of p, p*, s, and s* orbitals of the carbon ring systems that are transformed during the reaction. The full details of the active space used for each compound are given in the ESI. † ## CASPT2 and CAS-PDFT To include dynamic correlation, the CASSCF solutions are used as references in MC-PDFT and CASPT2 calculations. For MC-PDFT, we used the CAS-tPBE and CAS-ftPBE functionals. 15 These are our translated and fully translated functionals that use CASSCF solutions as references. For CASPT2, an empirical ionization-potentialelectron-affinity (IPEA) shift of 0.25 atomic units (6.80 eV) is added to improve agreement with experiment. 26 To illustrate the dependence of CASPT2 on this empirical parameter and to allow for a more standard comparison with MC-PDFT, we performed analogous calculations without the IPEA shift. These calculations are labeled as CASPT2-0. For the Cr dimer we also employed an IPEA value of 0.45 atomic units, as suggested for this specifc system in previous work. 8 All CASPT2 and CASPT2-0 computations used a standard imaginary shift of 0.2 atomic units (5.44 eV) to prevent intruder states. 27 ## SP calculations As in CASSCF and GASSCF in general, the CI coefficients are optimized via a Direct-CI procedure 28 while the orbital parameters are optimized through the Super-CI approach. 29 Intraspace rotations (inactive-inactive, virtual-virtual, gas1-gas1, gas2-gas2, .) are redundant and are not included in the optimization step; only inter-subspace rotations are included in the orbital optimization procedure. ## SP-PDFT calculations SP-PDFT calculations are just like CAS-PDFT calculations, except that the reference wave function is a separated pairs approximation. ## KS-DFT calculations The results of calculations with CASPT2, CASPT2-0, CAS-tPBE, CAS-ftPBE, SP-tPBE and SP-ftPBE are compared with those obtained from KS-DFT calculations with the PBE 30 exchangecorrelation functional. ## Geometries We used the experimental geometries of acetylene and ethylene as well as those of the ethynyl and vinyl radicals. 31 We optimized the structures of ethane and the ethyl radical by M06-2X 32 /6-31G(d). For the pericyclic reactions, the geometries and zero point energies of the reactants and transition states were obtained at the B3LYP/6-31G(d) level by Houk and coworkers. 33 The geometries of methylene and ozone were optimized by scanning the bond lengths at various bond angles. The geometries of planar and twisted ethylene were obtained with the MR-CISD/SA-3-RDP/aug-cc-pVTZ method by Lischka and coworkers. 34 For a-3-didehydrotoluene and 1,4-didehydrobenzene, we use geometries optimized at the M06-2X 32 /6-31G(d) level while using unrestricted Kohn-Sham DFT (abbreviated as UDFT). ## Bond energies and atomization energies All bond energies and atomization energies in this paper are potential energy differences excluding vibrational energies. Usually these are obtained from the literature, but for CH 2 the thermal correction to the enthalpy at 298 K obtained by KS-PBE/ aug-cc-pVTZ is added to the empirical DH 298 of CH 2 . The frequency component of this correction was scaled using the scaling factor obtained from ref. 35 for this model chemistry. ## Soware All the CASSCF, CASPT2, CASPT2-0, SP, and SP-PDFT calculations in this work were carried out with a locally modifed version of the Molcas 7.9 program suite. 36 All KS-DFT calculations were carried out with the Gaussian 09 program. 37 ## Systems studied In order to provide a broad test of the performance of SP-PDFT, we have computed the structural properties and bond energies of twelve diatomic molecules (LiH, HF, B 2 , C 2 , CO, S 2 , SO, NH, N 2 , O 2 , F 2 , and Cr 2 ) and two triatomic molecules (CH 2 and O 3 ). We also studied C-H bond dissociation in three prototypical organic compounds (acetylene, ethylene and ethane) and the barrier heights of fve pericyclic reactions. The pericyclic reactions are the electrocyclic ring opening of cyclobutene, the ring closing of cis-1,3,5-hexatriene and ortho-xylylene, and the sigmatropic shift reactions of 1,3-pentadiene and 1,3-cyclopentadiene. Finally, we examined the performance of SP-PDFT for describing the properties of open-shell singlet (OSS) systems, specifcally the relative energies of planar and twisted ethylene, and the singlet-triplet separations in a-3-didehydrotoluene and 1,4-didehydrobenzene. ## Diatomic molecules The ability of an electronic structure method to provide potential energy surfaces or potential energy curves that accurately describe the formation and cleavage of chemical bonds is a very important test of its capabilities. This task is challenging for methods based on a single-confguration reference state; for example, coupled cluster theory with full inclusion of single, double, and triple excitations (CCSDT) fails to properly describe the dissociation of N 2 into two N atoms. 38 The spectroscopic constants (the equilibrium distances, R e , and dissociation energies, D e ) of diatomic molecules have been computed with many theoretical methods (see ref. 39 for examples), and they are good test cases to compare the results obtained from SP-PDFT to those obtained with CAS-PDFT and CASPT2 as well as to accurate experimental data. ## Equilibrium bond distances of diatomic molecules In Fig. 3, we show the performance of SP-PDFT and other methods for predicting the equilibrium bond distances of eleven diverse main-group diatomic molecules, LiH, HF, B 2 , C 2 , CO, S 2 , SO, NH, N 2 , O 2 , and F 2 , and one transition-metal diatomic molecule, Cr 2 . The dominant confgurations in the CASSCF wave functions when using full-valence complete active spaces have percentage weights of 98.0, 99.9, 78.5, 70.9, 94.3, 94.8, 95.0, 98.3, 92.8, 94.0, 93.2, and 44.3, respectively. Since molecules in which the dominant confguration has a weight of less than or equal to 95% are usually considered to be multireference, we see that nine of the twelve molecules are multireference ones, the Cr 2 case being the one least dominated by a single confguration, followed by B 2 and C 2 . It has previously been recognized that CASSCF solutions generally lead to equilibrium bond lengths that are too long, 13 and our results are consistent with this. CASSCF has a mean absolute error (MAE) of 0.146 when compared to experimental data. This statistic is however dominated by the result obtained for Cr 2 , for which CASSCF overestimates the equilibrium bond length by 1.52 . Without the results obtained for Cr 2 , the MAE of CASSCF (labeled as MAE-11) is 0.021 . This is similar to previous results. 13 The MAE-11 of KS-PBE (0.022 ) is similar to that of CASSCF. However, we fnd that the CAS-tPBE and CAS- For the highly multireference systems (B 2 , C 2 , and Cr 2 ), CAS-tPBE and CAS-ftPBE perform better than CASPT2 and CASPT2-0 for B 2 and C 2 , while CAS-ftPBE gives a similar error as CASPT2 for Cr 2 . When comparing SP and CASSCF, we see that restricting the active space with the SP approximation only marginally alters the MAE and MAE-11 of the calculated bond distances of these diatomic molecules. The largest difference between the results obtained with CASSCF and SP was found for B 2 and Cr 2 . In all other cases, the difference between these methods is in the range 0.002-0.007 , as shown in Table S1. † More importantly, there is no noticeable difference in the MAE obtained for SP-PDFT (SP-tPBE and SP-ftPBE) and MC-PDFT (CAS-tPBE and CAS-ftPBE), as shown in Fig. 3. Indeed, SP-PDFT performs equally as well as CAS-PDFT in all the cases that were tested; details are in Table S1. † ## Dissociation energies of diatomic molecules The calculated bond dissociation energies of these twelve diatomic molecules are presented in Table 1. The dissociation energies are calculated as the difference between the potential energy of the molecule at 12 and the energy of the molecule at equilibrium. With CASSCF, the dissociation energies of these diatomic molecules are generally underestimated, and the MAE with respect to experimental values is 19.3 kcal mol 1 . Without the results obtained for Cr 2 , for which it underestimates the experimental dissociation energy by 30.8 kcal mol 1 , CASSCF has an MAE (labeled as MAE-11) of 18.2 kcal mol 1 . This underestimation of the dissociation energy is associated with an underestimation of the force constant and is related to the excessive antibonding character of CASSCF solutions. 13 Imposition of restrictions on the CI expansion by enforcing the SP approximation raises the MAE to 21.7 kcal mol 1 and the MAE to 20.8 kcal mol 1 , corresponding to differences of 3.4 and 2.6 kcal mol 1 , respectively, or about 11-13%. Of the methods that were tested, SP has the largest error. This is not surprising since the SP calculations use smaller active spaces than the CASSCF calculations. As discussed above, we use full valence active spaces in the CASSCF calculations, whereas the SP calculations contain only selected pairs of orbitals with no interpair excitations. KS-PBE calculations perform much better than either CASSCF or SP in nearly all cases, as expected since neither CASSCF nor SP include dynamic correlation. However KS-PBE also has a rather large MAE (11.8 kcal mol 1 ) as well as a large MAE-11 (12.0 kcal mol 1 ). The only system for which KS-PBE approaches chemical accuracy is hydrogen fluoride (HF), a system in which the dominant confguration has a weight of 99.9%. This is by all measures a single-reference system. The importance of including dynamic correlation for correctly computing the dissociation energies of these diatomic molecules is seen by comparing CASSCF with CAS-tPBE, CAS-ftPBE, CASPT2, and CASPT2-0 as well as by comparing SP with SP-tPBE and SP-ftPBE. The CAS-PDFT and SP-PDFT methods both perform very well for B 2 and C 2 , which are two systems with strong multireference character (the dominant confguration has a weight of less than 80%). 41 They reduce the MAEs and MAE-11s of CASSCF and SP by factors of about 4 and 6 respectively. For the systems presented in Table 1, the MAEs and MAE- 11s of CAS-PDFT and SP-PDFT are close to those of CASPT2. When comparing CAS-PDFT and SP-PDFT to CASPT2, one has to bear in mind that the latter incorporates an empirical IPEA shift, specifcally designed to improve agreement with experimental results; 2.2 (2.3) kcal mol 1 separates the MAE (MAE-11) of CASPT2 and CASPT2-0, indicating the importance of the empirical IPEA shift. 26 The CAS-PDFT and SP-PDFT results are almost as good as CASPT2 and CASPT2-0. Examination of Table 1 shows that the worst results for CAS-PDFT and SP-PDFT are obtained for F 2 (and Cr 2 in the case of CAS-tPBE and SP-tPBE). It is particularly encouraging that SP-tPBE and SP-ftPBE essentially match CAS-tPBE and CAS-ftPBE, which are based on full-valence CASSCF solutions; this is one of the key fndings of this paper, and it is important because SP-PDFT can treat much larger systems that CAS-PDFT. In principle, as CASSCF is affordable to upwards of 35 million CSFs, it should be possible to create SP solutions that approach that limit as well. As such one can envisage using SP and SP-PDFT for systems that are for CASSCF and CAS-PDFT. As examples, SP and SP-PDFT can be used to describe the full p/p* manifold of chrysene (C 18 H 12 ) as well as the full valence space of benzene-tetracyanoethylene complexes. Two other interesting points are (1) that the results are stable as far as replacing tPBE by ftPBE or vice versa and (2) that ftPBE results in signifcant improvements in the results obtained for Cr 2 (both for bond distances, Table S1 † and for bond dissociation energies, Table 1), suggesting that it might be particularly well suited for transition metal systems. ## Potential energy curves of diatomic molecules The ground-state potential energy curves of these twelve diatomic molecules were also scanned from near equilibrium to dissociation. Static correlation effects are generally more dominant at dissociation, and it is therefore important to test the ability of SP-PDFT to predict potential curves all the way out to this limit. The calculated potential energy curves as functions of the bond distances are presented for N 2 and O 2 in ESI (Fig. SI1 †). CASSCF, SP, CASPT2, CAS-PDFT, and SP-PDFT all give smooth curves. The potential curves obtained with SP are similar to those obtained with CASSCF, the energies obtained with SP-tPBE are similar to those obtained with CAS-tPBE, and those obtained with SP-ftPBE are similar to those obtained with CAS-ftPBE. Thus we fnd that the restrictions in going from CAS to SP do not degrade the potential energy curves. For O 2 and N 2 in the bonding regions ($0.9-1.2 ), the total electronic energy obtained with SP deviates from the CASSCF energy by about 3-10 kcal mol 1 as shown in Table 2. This is because some of the CSFs deleted in going from the complete active space to the separated-pair active space contribute nonnegligible amounts of dynamic correlation in these cases. At greater internuclear separations (2.5 and 5.0 ), the differences between the total energies obtained with CASSCF and SP become small, as also shown in Table 2. In contrast, the total energies obtained with SP-tPBE and SP-ftPBE are much closer to those of CAS-tPBE and CAS-ftPBE respectively. In Table 2, we see that the largest difference between the total energies obtained with the SP-PDFT and CAS-PDFT approaches are about 1.7 kcal mol 1 , which shows that the PDFT approach recovers the static and dynamic correlation energy that were neglected by using the approximate SP approximation in a variational wave function calculation. This is extremely encouraging. The ability of PDFT to recover these electron correlation effects is the reason why the potential energy curves obtained with SP-PDFT are closer to those obtained with CAS-PDFT in Fig. SI1 † than SP is to CASSCF. We emphasize that the SP-PDFT and CAS-PDFT agree well both in the bonding regions of N 2 and O 2 , where there are dominant confgurations with greater than 90% weight, and in the limit of dissociation, where there are many confgurations Table 1 The experimental dissociation energies (kcal mol 1 ) of eleven main-group diatomic molecules and the chromium dimer, Cr 2 , are compared with the calculated results obtained with several levels of theory. The MAE obtained without the results for Cr 2 is labeled as MAE-11. For each theoretical method, the deviation of the calculated results from experimental values is given. A negative sign denotes underestimation of the bond energy, while a positive sign indicates overestimation. All experimental data are taken from ref. 40 ## Triatomic molecules In this section, the calculated bond lengths, bond angles, and atomization energies of the two lowest energy states of methylene (CH 2 ) are presented, along with the calculated adiabatic 3 B 1 -1 A 1 gaps. The calculated geometry and atomization energy of ground-electronic-state ozone (O 3 ) are also presented. For CH 2 , a full valence CASSCF(6,6) wave function is used for subsequent CASPT2, CASPT2-0, and CAS-PDFT calculations. In C 1 symmetry, this active space choice results in 189 and 175 CSFs for the 3 B 1 and 1 A 1 states, respectively. For SP and SP-PDFT calculations, we used the SP-4 active space, as described above. This leads to a total of 25 and 17 CSFs for the 3 B 1 and 1 A 1 states, respectively. For ozone, we used a CAS(12,9) reference for CASSCF and an SP-3 reference for the SP approximation, with the latter resulting in the reduction of the number of CSFs from 2520 to 37. In essence 98.5% of the CSFs in the CASSCF (12,9) solution are completely neglected in the SP-3 approximate wave function. These active space schemes are illustrated in Fig. 4. Methylene. The structural parameters that we obtained with CASSCF are in good agreement with the CASSCF results of Apeloig et al. 43 The calculated C-H bond lengths and bond angles of the 3 B 1 and 1 A 1 states of CH 2 are compared with experimental data in Table 3. The experimental values of the C-H bond length and bond angle of the 3 B 1 state of CH 2 are 1.085 and 135.5 respectively. For the 1 A 1 state, the C-H bonds are longer (1.107 ) and the bond angle is signifcantly smaller (102.4 ). 44 Similarly to what was seen for the diatomic molecules, CASSCF and SP overestimate the C-H bond lengths in the 3 B 1 and 1 A 1 states of CH 2 . The calculated bond angles are also too large, as seen in Table 3. CAS-PDFT reduces the errors in the calculated structural properties of CH 2 to within the margins provided by CASPT2 Table 2 Effect of imposing restrictions on the active space with the GAS scheme on the total electronic energies of N 2 , O 2 and Cr 2 as functions of inter-nuclear distance R (). The differences in the total electronic energies obtained with CASSCF and SP, CAS-tPBE and SP-tPBE as well as CAS-ftPBE and SP-ftPBE are reported in kcal mol 1 , where A : B denotes the absolute value of the energy difference between A and B R ( ) and CASPT2-0. In general, the C-H bond lengths obtained with CAS-PDFT are within 0.004-0.007 of the values obtained with CASPT2 and CASPT2-0. This is the case for the 3 B 1 and 1 A 1 states. For the 3 B 1 state, CAS-PDFT overestimates bond angles by about 0.7-1.5 while CASPT2 underestimates them by about 1.9 . Compared with CAS-PDFT, SP-PDFT gives almost the same C-H bond lengths, and the bond angles are about 2 larger. We note that Jensen and Bunker obtained a bond angle of 133.9 for the 3 B 1 state. 45 This is 1.6 below the experimental value shown in Table 3, and indicates that the results obtained with CASPT2, CASPT2-0, CAS-PDFT and SP-PDFT are within the range of available experimental data. For the 1 A 1 state, CAS-PDFT overestimates the bond angle by up to 1.0 , while CASPT2 and CASPT2-0 underestimate by 0.5 and 0.7 , respectively. Similar to the situation for the 3 B 1 state, SP-PDFT results in slightly larger bond angles. An earlier approach for combining MCSCF-type methods with DFT has been described by Cremer and coworkers. 46 This method, which they call CAS-DFT, does not perform as well as CAS-PDFT and SP-PDFT for predicting the structural properties of CH 2 . 47 It overestimates the C-H bond length of the 3 B 1 state by 0.017 . For the 1 A 1 state of CH 2 , it overestimates the C-H bond length by 0.031 . To calculate the atomization energy of CH 2 and O 3 , the C-H and O-O bond lengths are stretched to 12 , while keeping the equilibrium bond angle fxed at the value obtained with each method. (Our general conclusions remain unchanged if we use the experimental value of the bond angle.) CASSCF and SP underestimate the atomization energy of the 3 B 1 state of CH 2 by about 20 kcal mol 1 while PBE overestimates by about 8.5 kcal mol 1 , as shown in Table 3. Calculations with CAS-tPBE and SP-tPBE bring the error down to below 3.0 kcal mol 1 , which is similar to CASPT2, which overestimates the bond energy by 1.2 kcal mol 1 . Inclusion of the gradient of the on-top pair density, results in errors of 8.4 and 7.4 kcal mol 1 for CAS-ftPBE and SP-ftPBE, respectively, still better than KS-PBE but much worse than tPBE. Table 3 shows that the CAS-tPBE and SP-tPBE calculations underestimate the adiabatic singlet-triplet gap by about 1.0 kcal mol 1 while CAS-ftPBE and SP-ftPBE calculations underestimate the separation by 2.3 and 2.7 kcal mol 1 , respectively. The earlier CAS-DFT approach also underestimates the gap (by 1.7 kcal mol 1 ). 47 These results are quite encouraging when compared to CASPT2 and CASPT2-0, which overestimate the separation by 2.7 and 4.8 kcal mol 1 , respectively. Ozone. The ozone molecule has been studied with a large number of quantum-mechanical methods. 11, We highlight the work of Vogiatzis and coworkers in which they showed that a GAS-2(12,9)-1e active space provides the same MAE as CASSCF (12,9) for the computed vertical excitation energies, ionization potential, and electron affinity of O 3 . 11 The GAS-2(12,9)-1e notation corresponds to two subspaces containing 12 electrons in 9 orbitals with one excitation allowed between the subspaces. In the present work, we have used an even more restrictive framework, namely SP-3, which becomes GASSCF-3(6,6)-0e in the general notation. The frst two subspaces each contain a coupled pair of s and s* orbitals while the third space contains a coupled pair of p and p* orbitals. In contrast, we placed 12 electrons in 9 orbitals for the CASSCF calculations. The nine orbitals are those formed by combination of the 2p x , 2p y , and 2p z orbitals of the three oxygen atoms, as shown in Fig. 5. The dominant confguration in the CASSCF (12,9) wave function has a weight of only 84%, showing that this system has signifcant multi-reference character. Table 3 shows that the optimized geometry of O 3 obtained with CASSCF is in good agreement with the results of Tsuneda et al., 51 who used a similar active space with the cc-pVTZ basis sets augmented with s, p, and d diffuse functions. Also, the structural parameters obtained with CASPT2 are in agreement with the reports of Ljubic and Sabljic, who used the same active space. 50 The bond lengths and bond angle obtained with SP-PDFT are similar to those obtained with CAS-PDFT, despite the fact that the underlying SP wave function contains only about 1.4% of the number of CSFs in the CASSCF solution. CAS-PDFT slightly underestimates the O-O distances and slightly overestimates the bond angle, while CASPT2 and CASPT2-0 have opposite behaviors. Krishna and Jordan have previously reported that CASSCF underestimates the atomization energy of O 3 by about 57.7 kcal mol 1 . 52 Table 3 shows that CASSCF and the SP approximation are both poor for calculating the atomization energy of O 3 . These are the two methods that do not attempt to include most of the dynamic electron correlation. On the other hand, PBE overestimates the atomization energy by about 42 kcal mol 1 but CAS-tPBE and CAS-ftPBE reduce the error of PBE by 14 and 20 kcal mol 1 , respectively. SP-tPBE and SP-ftPBE behave similarly to CAS-tPBE and CAS-ftPBE respectively. However, CASPT2 and CASPT2-0 perform best for predicting the atomization energy of O 3 . ## C-H bond dissociation energies in organic compounds In this section we study C-H bond dissociation in acetylene, ethylene, and ethane, that is: The calculated energies for these reactions are compiled in Table 4, where they are compared to experimental values estimated by adding the thermal correction to the enthalpy at 298 K obtained by KS-PBE/aug-cc-pVTZ to the empirical DH 298 reported by Blanskby and Ellison. 53 The frequency component of this correction was scaled using scaling factors obtained from ref. 35 for this model chemistry. We used full valence active spaces for the CASSCF calculations in this table: CAS(10,10), CAS(9,9), CAS (12,12), CAS (11,11), CAS (14,14) and CAS (13,13) for acetylene, ethynyl, ethylene, vinyl, ethane and ethyl respectively. With C 1 symmetry, these active spaces result in 19 404, 8820, 226 512, 104 544, 2 760 615 and 1 288 287 CSFs respectively. In Table 4 we report the number of CSFs only for the parent compound and not for the dissociation radical species. The active spaces used in SP and SP-PDFT calculations are presented in the computational details section and are illustrated for ethane and the ethyl radical in Fig. 5. Only orbitals with signifcant C-H character are included in the SP active spaces. As previously noted, we used SP-3, SP-3, SP-4, SP-4, SP-6 and SP-6 active spaces for the acetylene, ethynyl, ethylene, vinyl, ethane and ethyl, respectively. These result in 37, 17, 150, 76, 3012 and 1704 CSFs respectively, a signifcant reduction compared with the full CASSCF calculation. In all the three cases presented in Table 4, CAS-PDFT performs much better than CASPT2, CASPT2-0, or PBE, and SP-tPBE and SP-ftPBE do even better with MAEs of only 1.6 and 1.7 kcal mol 1 , respectively. This is another demonstration that PDFT effectively recovers correlation that is left behind by the active space restrictions of the SP approximation, even though we were very aggressive in including only a small number of pairs. Fig. S2 † shows that the SP-PDFT potential energy curves for C-H cleavage in acetylene, ethylene, and ethane match those obtained with CASPT2 and CAS-PDFT very well, so the active space restrictions do not distort the shape of the potential energy curves. In general we are presenting the SP-PDFT results for just one small active space, as previously indicated. We did do some testing to see the effect of using different choices of which pair of orbitals to include, and we found that the effect of adding or removing spectator pairs was small. For example we used SP-5 for ethylene and vinyl and found a difference in the calculated C-H dissociation energy of only 0.2 kcal mol 1 as compared to the SP-4 results presented in the table. ## Barrier heights for pericyclic reactions We have previously shown that CAS-tPBE reduces the average error of PBE by a factor of 2.7 for predicting the forward and reverse barrier heights for chemical reactions involving small molecules. 19 CASPT2 however was found to have a lower MAE than CAS-tPBE. Houk and coworkers have collected datasets of the barrier heights of pericyclic reactions. 33,54 These datasets can be used to benchmark computational approaches. In Table 5, we compare the calculated barriers for fve pericyclic reactions with experimental data taken from the dataset of Houk and coworkers. 33,54 These reactions are shown in Fig. 6. SP and CASSCF overestimate the reaction barriers by 15 kcal mol 1 or more, in nearly all cases. Table 5 shows that the only exception is for reaction 1, for which they overestimate it by only 2.4 and 3.0 kcal mol 1 respectively. SP-PDFT and CAS-PDFT greatly improve on CASSCF and SP, reducing the MAEs by factors of 3-6. This is similar to the situation found for reactions involving small molecules. 19 The MAEs of SP-PDFT and CAS-PDFT are still somewhat large, at least when compared to CASPT2. However, regarding the central topic of this manuscript, we fnd that the calculated reaction barriers are stable to the use of a restricted SP wave function; that is, SP-PDFT and CAS-PDFT yield similar results for each reaction and overall have similar MAEs. ## Open-shell singlet systems Cremer and coworkers have previously used UDFT, brokensymmetry UDFT, and CAS-DFT 46 to study open-shell singlet diradicals. 47 Specifcally, they studied the energy of the 1 B 1 state of twisted ethylene relative to planar ethylene. In addition, they also studied the singlet-triplet gaps of 1,4didehydrobenzene and a,3-didehydrotoluene, which are shown in Fig. 7. They found that CAS-DFT predicts the 1 A 00 state of a,3-didehydrotoluene to be lower in energy than the 3 A 00 state. Also the 1 A g state of 4-didehydrobenzene is predicted to be lower in energy than the 3 B 1u state. These state orderings are in agreement with experimental data. 55,56 In Table 6, we compare the results obtained when PBE, CASSCF, SP, SP-PDFT, CAS-PDFT and CASPT2 are used to carry out similar computations to those performed by Cremer and coworkers. 47 We used CAS (8,8) and SP-4 active spaces in the calculations on the singlet and triplet states of 1,4- didehydrobenzene and a-3-dide-hydrotoluene. For the KS-DFT computations, the singlet states were treated as unrestricted broken-symmetry solutions. For planar and twisted ethylene, we used CAS (12,12) and SP-1 active spaces, respectively. For the gap between twisted and planar ethylene, SP-tPBE and SP-ftPBE provide similar results that are 2-3 kcal mol 1 above those obtained with CAS-tPBE and CAS-ftPBE respectively, as shown in Table 6. The results obtained with CAS-ftPBE and CASPT2 are the same. The results obtain for the fully translated functionals improve upon CAS-tPBE and SP-tPBE. Lischka and coworkers obtained a value of 69.2 kcal mol 1 at the MR-CISD+Q/SA-3-RDP level while using similar basis sets. 34 This is close to the values obtained by CAS-ftPBE and CASPT2. For 1,4-didehydrobenzene, the results obtained with CAS-tPBE, CAS-ftPBE, SP-tPBE and SP-ftPBE fall within the error bar of the experiment, which is 3.5 AE 0.5 kcal mol 1 . In contrast, CASPT2 and CASPT2-0 fall outside this range; they overestimate even as compared to the high end of the experimental results 55 by 0.7 and 0.5 kcal mol 1 , respectively. CAS-DFT predicted an energy separation of 2.6 kcal mol 1 between the 1 A g and 3 B 1u states of 1,4-didehydrobenzene. 47 For a,3-didehydrotoluene, CAS-tPBE, CAS-ftPBE, SP-tPBE and SP-ftPBE correctly predict that the 1 A 00 state is more stable than the 3 A 00 state. The separations provided by all these methods are similar and fall within the range of available experimental data. They are however about 1.5 kcal mol 1 smaller than the separations predicted by CASPT2 and CASPT2-0. Unfortunately, available experimental reports only indicate that the splitting should be lesser than 5.0 kcal mol 1 . ## Concluding remarks The present paper contains new methods for both wave function theory and density functional theory. Starting with wave function theory, we have presented a systematic way to choose the active space in generalized-active-space selfconsistent-feld (GASSCF) theory. This new way of choosing the active space is called the separated-pair (SP) approximation. The method is intermediate between generalized valence bond (GVB) theory and complete active space self-consistent-feld (CASSCF) theory. The SP wave function is a truncation of CASSCF, obtained by partitioning the CAS active space into an arbitrary number of generalized-active-space (GAS) subspaces that each contain at most two orbitals, and inter-subspace excitations are excluded. In the examples, only pairs required to describe a particular bond-breaking process are included in the GAS subspaces; all other orbitals are treated as doubly occupied in all confgurations. With such a choice, the SP methods can be used for large systems for which conventional CASSCF calculations are unaffordable. Just as for GVB and CASSCF, the precise choice of active space in the SP approximation is not completely unambiguous because in all three methods one must decide which orbitals to correlate. The orbitals to be correlated are an individual choice, although we expect that the most useful choice will usually be a bonding orbital and the corresponding antibonding orbital. In the present paper we include the orbitals involved in bond breaking and in some cases also additional orbitals closely close coupled to the bond breaking. A general objective might be to include the orbitals responsible for nondynamic correlation, which is also called static correlation, strong correlation, and near-degeneracy correlation. Although dynamic correlation tends to be very similar across systems, nondynamic correlation is usually system-specifc. Therefore, a practical multi-confguration approach may well have to be applied in a case-by-case manner, sometimes requiring chemical insight. But the examples presented here show that simple considerations lead to reasonably accurate results for a set of diverse cases and signifcantly reduce the computational cost of specifc problems. The defnition and exploration of SP may be useful for all methods that need to start from a strongly correlated reference wave function. We subsequently considered whether the SP approximation is useful for multi-confguration pair-density functional theory (MC-PDFT), and to put this in context we frst contrast MC-PDFT to Kohn-Sham density functional theory (KS-DFT). In KS-DFT, ones represents the density by a Slater determinant, and one writes the total energy as the sum of the kinetic energy computed from Slater determinant by standard wave function methods, the Coulomb energy computed classically from the density, and a remainder. The remainder is a functional of the density and is called the exchangecorrelation energy, and it includes not just the deviation of the true potential energy from the Coulomb energy computed classically from the density, but also the deviation of the true kinetic energy from the Slater-determinant kinetic energy. In MC-PDFT, we represent the density and the on-top pair density by a multi-confgurational wave function, and we write the total energy as the sum of the kinetic energy computed from the multi-confgurational wave function by standard wave function methods, the Coulomb energy computed classically from the density, and a remainder. The remainder is a functional of the density and the on-top pair density and is called the on-top energy, and it includes both the deviation of the true potential energy from the Coulomb energy computed classically from the density and the deviation of the true kinetic energy from the multi-confgurational-wave-function kinetic energy. In most previous attempts to combine multi-confgurational wave functions with density functional theory, one writes the total energy as the sum of the energy computed by wave function theory from the multi-confgurational wave function plus a remainder. Because the energy computed by wave function theory from the multi-confgurational wave function includes some of the effect of electron correlation on the true potential energy, one must be careful not to include this portion of the correlation energy in the remainder; this can be called the double counting problem. Because we use only the kinetic energy of the multi-confgurational wave function, we avoid this double counting problem. Note though that we do not know an exact on-top functional, just as an exact exchangecorrelation functional is not known in KS-DFT, so our treatment is not exact. A major goal of both KS-DFT and MC-PDFT is to fnd a better approximation to the corresponding exact functional. One motivation for MC-PDFT is that it might be "easier" to fnd a good on-top functional than to fnd a good exchange-correlation functional for two reasons: (1) our kinetic energy is based on a representation that better conforms to the true wave function when (as is often the case) the system is inherently multi-confgurational (as, for example, for describing the breaking of a covalent bond), and (2) our functional is allowed to depend on the on-top pair density, which brings in extra information. Many years of development have gone into modern exchange-correlation functionals, 57 whereas for on-top functionals we are still using frst-generation approximations. The present paper, however, is not about improving the functional but rather about testing how many and what kind of confgurations need be present in the multi-confgurational wave function in order obtain reasonable results with simple density functionals. We found that the new SP approximation, discussed in the previous paragraph for wave function theory, provides an economical multi-confgurational wave function that yields good accuracy with MC-PDFT. Thus we have presented a version of MC-PDFT called separated-pair pair-density functional theory, abbreviated SP-PDFT. The SP-PDFT method uses a separated-pair (SP) wave function to generate the kinetic and classical Coulomb contributions to the total electronic energy, and the remainder of the total electronic energy is computed from a functional of the total density and the on-top pair density taken from the SP wave function. The SP-PDFT methods can therefore be used for large systems for which conventional CASSCF calculations, CASPT2, and CAS-PDFT are unaffordable. Sometimes the SP approximation wave function calculations agree well with the CASSCF ones; in other cases they are less accurate, as would be expected. But even in cases where the energetic results obtained by wave function theory from the SP approximation are less accurate than those obtained by CASSCF, we show that the SP approximation is able to produce an accurate enough kinetic energy and on-top pair density that the SP-PDFT results are in generally good agreement with the CAS-PDFT results. The tests included in this article include structural properties and bond dissociation energies of eleven diatomic and two triatomic molecules, the C-H dissociation energies of prototypical organic systems, the reaction barriers of pericyclic reactions, and the description of open-shell singlet species. In all the cases that were tested, SP-PDFT provides approximately the same accuracy as CAS-PDFT. In most cases, both SP-PDFT and CAS-PDFT provide similar levels of accuracy as the much more expensive CASPT2 approach; the only exception to this is for the reaction barriers of pericyclic reactions. The key result for wave function theory is that the SP approximation often agrees quite well with CASSCF, at greatly reduced cost, and this extends the usefulness of the method to bigger systems. The key result for density functional theory is that the quality of results obtained from MC-PDFT calculations remains largely unchanged even with drastic reductions in the number of included CSFs, as we have in made in the SP-PDFT variant of the method. In addition the SP-PDFT approach, just as is the more general MC-PDFT framework, is free of doublecounting of electron correlation energies. This double-counting problem plagues nearly all other hybrid approaches for combining CASSCF and KS-DFT. Future work that one can anticipate includes testing the performance of the SP and SP-PDFT methods for transition metal complexes, developing better on-top functionals of the MCSCF density and on-top pair density, and developing an orbital optimization algorithm that includes the on-top functional in the self-consistent-feld step.

chemsum

{"title": "Separated-pair approximation and separated-pair pair-density functional theory", "journal": "Royal Society of Chemistry (RSC)"}

concurrent_suppression_of_aβ_aggregation_and_nlrp3_inflammasome_activation_for_treating_alzheimer's_

## Abstract: Alzheimer's disease (AD) is a neurodegenerative illness accompanied by severe memory loss, cognitive disorders and impaired behavioral ability. Amyloid b-peptide (Ab) aggregation and nucleotide-binding oligomerization domain (NOD)-like receptor protein 3 (NLRP3) inflammasome play crucial roles in the pathogenesis of AD. Ab plaques not only induce oxidative stress and impair neurons, but also activate the NLRP3 inflammasome, which releases inflammatory cytokine IL-1b to trigger neuroinflammation. A bifunctional molecule, 2-[2-(benzo[d]thiazol-2-yl)phenylamino]benzoic acid (BPBA), with both Abtargeting and inflammasome-inhibiting capabilities was designed and synthesized. BPBA inhibited selfand Cu 2+ -or Zn 2+ -induced Ab aggregation, disaggregated the already formed Ab aggregates, and reduced the neurotoxicity of Ab aggregates; it also inhibited the activation of the NLRP3 inflammasome and reduced the release of IL-1b in vitro and vivo. Moreover, BPBA decreased the production of reactive oxygen species (ROS) and alleviated Ab-induced paralysis in transgenic C. elegans with the human Ab 42 gene. BPBA exerts an anti-AD effect mainly through dissolving Ab aggregates and inhibiting NLRP3 inflammasome activation synergistically. ## Introduction Alzheimer's disease (AD) is a common form of dementia characterized by the accumulation of extracellular amyloid bpeptide (Ab) plaques, neuroinflammation and neuronal cell death in the brain. 1 About 50 million people are living with AD globally in 2019, which put an enormous economic and mental burden on the society and families. 2 Although great effort has been made, the pathogenesis and pathogenic factors of AD are not yet fully elucidated. 3 Existing anti-AD drugs merely delay the symptoms to some extent but cannot stop the progression of the disease and have various side effects. 4 The Ab cascade hypothesis is the most prevalent supposition about the pathogenesis of AD, which suggests that Ab deposits play a vital role in initiating the disease. 5,6 In the past few decades numerous studies focused on cellular Ab deposits as a pathological hallmark and target of therapeutic drugs. 7 The newly FDA approved aducanumab is the frst anti-AD drug based on the Ab cascade hypothesis, though its efficacy is inconclusive. 8 Recently, increasing evidence supported that innate immunity-mediated neuroinflammation plays a crucial role in the pathogenesis and progression of AD. 9 Inflammasome plays an important role in neuroinflammation and neurodegenerative diseases. 10,11 Particularly, the microglia-specifc nucleotidebinding oligomerization domain (NOD)-like receptor protein 3 (NLRP3) inflammasome mediates the pathogenesis of AD. 12,13 The NLRP3 inflammasome is an intracellular multimeric protein complex composed of the receptor protein NLRP3, the effector protein cysteine protease-1 (caspase-1), and the adaptor protein called apoptosis-associated speck-like protein containing a caspase activation and recruitment domain (ASC). 14 It affects a variety of physiological functions, including the innate immune process and caspase-1-dependent response. 15,16 Ab plaques activate the NLRP3 inflammasome, which releases inflammatory factors such as cytokines interleukin-1b (IL-1b), interleukin-18 (IL-18), and ASC. 17 The released inflammatory cytokines and ASC trigger chronic neuroinflammation and lead to cognitive impairment. 12,18 In reverse, neuroinflammation aggravates the formation of Ab fbers and plaques, and worse, boosts tau phosphorylation, thus leading to their aggregation to promote the pathogenesis of AD. 16 The activation of the NLRP3 inflammasome/caspase-1 axis contributes much to AD in vivo, 10 while the defciency of NLRP3 or caspase-1 markedly reduces the Ab burden and cognitive impairment in amyloid precursor protein/presenilin-1 (APP/PS1) mice. 19 The elevation of IL-1b in the brain has been associated with the progression and onset of AD. 20,21 The inhibition of IL-1b could signifcantly diminish brain nerve inflammation, alleviate cognitive impairment, and partially reduce Ab deposition in 3xTg-AD mice. 22 Various inhibitors of the inflammasome have been reported, 23 such as OLT1177, 19,24 CY-09, 25 tranilast, 26 oridonin, 27 benzenesulfonamide analogues, 28 sulphonamides (CRIDI, MCC950), and non-steroidal anti-inflammatory drugs (NSAIDs). 12,32 More inhibitors are targeted to Ab aggregation; however, inhibitors that emphasize both Ab aggregation and inflammasome are rare. Since the synergism between Ab oligomers or plaques and pro-inflammatory factors could increase the neural damage to the brain, 17 a combination therapy involving the inhibition of Ab aggregation and NLRP3 inflammasome activation may enhance the therapeutic effect on AD. Herein, we integrate benzothiazole, an Ab-targeting group, with o-aminobenzoic acid, an analogue of the NLRP3 inflammasome inhibitor mefenamic acid, 32 into a single molecule BPBA (Fig. 1), which may lead a dual inhibition of NLRP3 inflammasome activation and Ab aggregation simultaneously. A series of experiments demonstrate that BPBA remarkably inhibits the self-and metalinduced Ab aggregation, reduces the level of inflammatory cytokine IL-1b, restrains the activation of caspase-1 in vitro, and alleviates the formation of Ab oligomers and plaques as well as the Ab-associated toxicity in vivo. ## Design, synthesis and physicochemical properties The design of BPBA is based on the structures of benzothiazole and o-aminobenzoic acid; the former is a potential Ab-targeting group that has a specifc affinity for Ab aggregates rich in bsheet structures, and the latter is an analogue of mefenamic acid, which is a known inhibitor of the NLRP3 inflammasome. 32 In addition, the O, N, and S atoms in BPBA could chelate metal ions, which may inhibit the metal-induced Ab aggregation. We suppose that BPBA has the capability to prevent the formation of Ab plaques and restrain the activation of the NLRP3 inflammasome in the brain of AD sufferers. The synthesis and characterization of BPBA are shown in Scheme 1 and Fig. S1. † The synthetic intermediate BP was synthesized by reacting 2-aminobenzoic acid with 2-aminothiophenol in polyphosphoric acid (PPA) as reported in the literature. 36 BPBA was prepared by a modifed literature method, 37 which produced a yellow solid with a yield of 40%. BPBA is soluble in acetonitrile, methanol, and dimethyl sulfoxide (DMSO), but is insoluble in water. In the acetonitrile solution, two absorption peaks around 285 and 380 nm were detected by UV-vis spectroscopy, representing the existence of the benzene ring and the whole BPBA, respectively (Fig. S2A †). The maximum emission peak of BPBA is at 505 nm (l ex ¼ 380 nm) in the emission spectrum (Fig. S2B †). The blood-brain barrier (BBB) is the main obstruction for developing anti-AD agents. 38 The BBB-penetrating ability of BPBA was frstly evaluated on the basis of Lipinski's "rule of fve", which predicts that a compound would cross the BBB when logBB is larger than 0.3. 39 The calculated logBB of BPBA is 0.18 (Table S1 †), and all other indices meet Lipinski's rule except Clog P, implying that it may potentially cross the BBB. The log P octanol/water partition coefficient was further determined by the shaking flask method and UV spectroscopy. The lipophilicity (log P o/w ) was calculated to be 1.24 AE 0.08 (Table S2 †), suggesting that BPBA could penetrate the BBB. ## Inhibition of Ab aggregation Excessive Zn 2+ is associated with the generation of Ab aggregates, and Cu 2+ may lead to the aggregation of Ab and production of reactive oxygen species (ROS), 40,41 which would result in damage to neurons and synapses in AD patients. 42,43 Since the O, N, and S atoms in BPBA could coordinate to metal ions, the chelating ability of BPBA to Zn 2+ or Cu 2+ was investigated by fluorescence titration in Tris-HCl buffer. As shown in Fig. 2, with the addition of Zn 2+ or Cu 2+ , the maximum fluorescence intensity of BPBA decreased gradually and approached equilibrium when the ratio of [Zn 2+ ]/[BPBA] reached 2.0, whereas the fluorescence intensity kept decreasing even when the ratio of [Cu 2+ ]/[BPBA] reached 3.2. The results indicate that BPBA can bind to Zn 2+ or Cu 2+ . However, since BPBA contains at least 4 coordination atoms and could form different chelates with Zn 2+ or Cu 2+ , that is, the product is not unique, it is hard to identify these species in a complex mixture. Thus we only tentatively obtained an apparent binding constant of BPBA to Zn 2+ at 1 : 1 stoichiometry, which was calculated to be 0.254 mM 1 according to the reported method. 44,45 The chelation of BPBA to Zn 2+ or Cu 2+ was further confrmed by the HR-MS spectra in the Tris-HCl buffer, where the chelate cations of Zn 2+ or Cu 2+ with BPBA were observed (Fig. S3 †). The neurotoxicity of Ab aggregates largely originates from the b-sheet conformers. 46 The inhibition effect of BPBA on the bsheet formation of Ab was studied by the ThT assay, which is widely used to detect the content of b-sheets in Ab aggregates. ## Morphological alteration of Ab The morphology of Ab ## Inuence on the hydrophobicity of Ab Ab fbrils are highly hydrophobic, which can be detected by using the fluorescent probe 8-aniline-1-naphthalenesulfonic acid (ANS). 50 The fluorescence of ANS is enhanced signifcantly after binding to the hydrophobic structure on the surface of proteins. 51 Therefore, the influence of BPBA on the hydrophobicity of Ab can be reflected by the fluorescence changes of ANS. As shown in Fig. 5, the fluorescence intensity of ANS decreased and the maximum emission wavelength red-shifted once Cu 2+ was added to the solution of Ab 42 , suggesting that Cu 2+ reduced the number of surface hydrophobic structures or the hydrophobicity of Ab 42 fbrils. In contrast, the fluorescence intensity of ANS increased after Zn 2+ was incubated with Ab 42 , indicating that Zn 2+ increased the hydrophobicity of Ab 42 fbrils, and its impact on the hydrophobicity is different from that of Cu 2+ . Once BPBA was added into the above solutions, the fluorescence intensity of ANS decreased signifcantly, suggesting that BPBA has a remarkable inhibitory effect on the surface hydrophobic structures of self-and metal-induced Ab 42 aggregates. These results verify that BPBA can reduce the hydrophobicity or increase the hydrophilicity of Ab 42 aggregates. The fluorescence interference of BPBA with ANS is negligible on this occasion (Fig. S6 †). ## Effect of BPBA on nerve cells and Ab toxicity Before assessing the effect of BPBA on the Ab toxicity to neuron cells, we frst examined the possible neurotoxicity of BPBA. Mouse neuroblastoma N2a (or Neuro-2a) cells are often used to study the pathological mechanism of AD. 52 Therefore, the survival of N2a cells after incubation with BPBA for 24 h was tested by the MTT assay. As shown in Fig. 6A, the survival rate of N2a cells is around 100% even when the concentration of BPBA reached 140 mM, indicating that it is almost non-toxic to the neuron cells. The extremely low neurotoxicity suggests that BPBA per se is safe for neuron cells in the following experiments. The neurotoxicity of Ab 42 , Cu 2+and Zn 2+ -Ab 42 aggregates in the absence and presence of BPBA toward neuronal model cell PC12 from rat pheochromocytoma was tested using the MTT assay. As shown in Fig. 6B, Ab 42 , Cu 2+and Zn 2+ -Ab 42 aggregates were markedly toxic to the PC12 cells, with the cell viability declining by more than 30%. The toxicity of Ab 42 , Cu 2+and Zn 2+ -Ab 42 aggregates was signifcantly attenuated in the presence of BPBA, with the cell viability rising to more than 80%. The results show that BPBA can suppress the neurotoxicity of Ab 42 , Cu 2+and Zn 2+ -Ab 42 aggregates and enhance the viability of neuron cells. ## Inhibition on inammasome The NLRP3 inflammasome is an intracellular complex that activates caspase-1, which processes the IL-1b precursors into active molecules, and the mature IL-1b is released to the extracellular fluid. 13 Human acute monocytic leukemia THP-1 cells are similar to the phenotype and functional characteristics of human primary monocytes/macrophages, therefore are commonly used as their model cells. 53 THP-1 cells become mononuclear macrophages under the differentiation induced by phorbol ester (PMA). When the THP-1 macrophages were costimulated with lipopolysaccharide (LPS) and ATP, the NLRP3 inflammasome/caspase-1 was activated and a variety of cytokines like IL-1b were synthesized and released. 54 Therefore, the effect of BPBA on the NLRP3 inflammasome was evaluated using the THP-1 macrophages. The NLRP3 inflammasome requires the adapter protein ASC to activate caspase-1. After inflammasome activation, ASC assembles into a large protein complex called "speck", which can be detected by immunocytochemistry as the size reaches around 1 mm. Therefore, the formation of ASC specks is regarded as a simple upstream indicator of inflammasome activation. 25,55 As shown in Fig. 7, red ASC specks were formed in the cytoplasm of THP-1 macrophages after stimulation with LPS and ATP; when the stimulated cells were treated with BPBA subsequently, the formation of ASC specks was dramatically inhibited, suggesting that BPBA can inhibit the activation of the NLRP3 inflammasome. Theoretically, the inhibition of BPBA on ASC specks should depend on its concentration; however, a quantitative relationship is unavailable because the exact position and amount of ASC specks are difficult to determine. According to the inhibition effect, 20 mM BPBA seems to be an appropriate concentration for the inhibition. As a result of ASC decline, the activation of caspase-1 was inhibited accordingly. As shown in Fig. 8, caspase-1 mainly exists as inactive pro-caspase-1 with or without stimulation; the activated caspase-1 p12 and caspase-1 p10 only increased after THP-1 macrophages were co-stimulated with LPS and ATP. However, in the presence of BPBA, the expression of activated caspase-1 decreased to the unstimulated level. IL-1b is a main inducer of inflammation and one of the main mediators of innate immune response, which is produced as an inactive precursor (pro-IL-1b) that requires cleavage by caspase-1 for activation and secretion. 56 Its level is elevated in the brain of AD patients and is associated with the progression and onset of AD. 20 As shown in Fig. 9, a large amount of IL-1b was secreted when THP-1 macrophages were stimulated with LPS and ATP, whereas the secretion of IL-1b was dramatically suppressed by BPBA, hence manifesting that BPBA can inhibit the release of IL-1b. All these results suggest that BPBA could inhibit the activation of the NLRP3 inflammasome and suppress the maturation and release of IL-1bthe fnal product of the inflammasome, thus predicting that BPBA could reduce the inflammatory response in AD patients. ## Production of ROS and alleviation of paralysis in C. elegans The transgenic C. elegans expressing Ab gene in muscles or neurons is widely used as an in vivo model of AD to test the effect of compounds on Ab aggregation and toxicity. 57 Ab plaques can induce ROS accumulation and oxidative stress, thus aggravating the pathological progress AD ulteriorly. 58 The changes of ROS in C. elegans CL4176 strains with or without BPBA were detected with a 2 0 ,7 0 -dichlorodihydrofluorescein diacetate (DCFH-DA) probe. The ROS level was reflected by the DCF fluorescence that is positively dependent on the oxidization of DCFH by intracellular ROS. 59 As shown in Fig. 10A, the green fluorescence intensity of BPBA-treated worms is signifcantly weakened as compared with that of the control, suggesting that BPBA can inhibit the production of ROS in the worms. The quantitative data of the fluorescence intensity are shown in Fig. 10B. The relative fluorescence intensity of the BPBA-treated group markedly decreased as compared to that of the control, indicating that BPBA can inhibit the level of ROS in C. elegans. CL4176 specifcally expresses the Ab 42 gene in the SMG-1 mRNA temperature induction system, which results in the time-dependent aggregation of Ab and paralytic phenotypes. 60 The activated SMG-1 system recognizes the Ab gene and degrades it at 15 C, causing the worms to produce a low level of Ab and move as rollers; when the temperature is raised to 25 C, the SMG-1 system is inactivated and Ab aggregates are formed in the muscle cells, which paralyze the nematode gradually. The effects of BPBA on the Ab-induced paralysis in CL4176 are shown in Fig. 10C. The BPBA-treated CL4176 nematodes were not paralyzed until 42 h after the temperature upshift; however, the paralysis rate of the untreated nematodes began to rise quickly thereafter, reaching more than 50% at 44 h. The paralysis rate of the worms treated with BPBA (68 mM) is only 7% at 48 h, while that of the control group is 73%. Evidently, BPBA can inhibit the Ab-induced paralysis in transgenic CL4176 strains; in other words, it can downregulate the expression of the Ab 42 gene in CL4176 or alleviate the Ab toxicity to neurons of C. elegans. ## Reduction of Ab and IL-1b in AD mice The potential of BPBA to eliminate the pathogenic factors of AD was further verifed in mice. APP/PS1 double-transgenic mice that produce elevated levels of Ab by expressing human APP and PS1 mutants from 6 months of age were used as AD models. 61 To select an appropriate dose for the assay, the acute toxicity of BPBA to wild type (WT) C57BL/6J mice was frst evaluated. No obvious toxicity was observed at a dose of 10 mg kg 1 BPBA based on the mortality in 2 weeks (Fig. S7 †). The APP/PS1 mice Fig. 7 Immunofluorescence images of THP-1 macrophages before and after stimulation with LPS (1 mg mL 1 ) plus ATP (5 mM) and subsequent treatment with BPBA. Red, ASC specks; blue, DAPI-stained nuclei. Fig. 8 Expressions of pro-caspase-1 and caspase-1 in the LPS-ATPactivated THP-1 macrophages before and after the treatment with BPBA determined by western blotting. were treated with BPBA at a dose of 5 mg kg 1 every 3 days for 3 months. The brain tissues of the BPBA-treated and untreated AD mice were taken out and the expression of Ab was analyzed by western blotting. As shown in Fig. 11A and S8, † the Ab species with a molecular weight #55 kDa (oligomeric species) decreased signifcantly, thus confrming that BPBA can inhibit the formation of Ab oligomers and plaques in the brain of AD mice. The results also imply that BPBA could pass through the BBB of the mice. The effect of BPBA on pro-inflammatory IL-1b in the brain of AD mice was also investigated. As shown in Fig. 11B, the content of IL-1b in the brain of saline-treated wide-type (WT) and AD mice is 125 and 250 pg mL 1 , respectively; however, that in the brain of BPBA-treated mice is 100 pg mL 1 , which is even lower than that in the WT mice. The results demonstrate that BPBA can reduce the level of IL-1b in vivo and thereby douse the inflammatory responses. Since IL-1b is an immunomodulatory cytokine, the decrease of IL-1b production may further affect the innate immune cells in the brain. ## Efficacy on APP/PS1 mice Finally, the in vivo therapeutic effect of BPBA on the learning and cognitive abilities of APP/PS1 mice was assessed by the Morris water maze (MWM) test according to reported procedures. 61 As shown in Fig. 12A, the time required for the BPBAtreated mice to fnd the survival platform is similar to that for the saline-treated WT mice, but is shorter than that for the AD control mice. On the 6th day, the survival platform was removed. The frequency of presence in the survival platform area (quadrant II) for the BPBA-treated mice increased within 60 s as compared with that for the saline-treated AD mice (Fig. 12B). Preliminary behavioral experiments show that BPBA can slightly alleviate the memory impairment of AD mice. The benefcial effect of BPBA on the memory of AD mice seems not so effective as expected, because the recovery of memory involves ceasing of neuron damage and regeneration of damaged neurons. 62 BPBA can eliminate the pathogenic factors of neuron damage but cannot restore the damaged neurons. It is very possible that when the treatment began, the neuron damage had occurred, so the improvement of memory is not evident. This may be the reason why so many AD drug candidates failed in the clinic. Fortunately, the AD symptom did not get worse under the treatment of BPBA; in other words, BPBA can stop the progression of the disease, which is superior to most AD drugs. ## Conclusion Ab aggregation is believed to be a key factor in the pathogenesis of AD; likewise, chronic neuroinflammation mediated by the activation of the NLRP3 inflammasome also plays a crucial role in the AD pathogenesis. 10,63 Ab deposits activate the NLRP3 inflammasome, leading to an overproduction of IL-1b and neuroinflammation. Therefore, dissociating Ab aggregates not only directly reduces the Ab-induced neurotoxicity to neurons, but also inhibits the activation of the NLRP3 inflammasome, curbs inflammatory responses, and decreases the release of neuro-destructive inflammatory cytokines. 17,31 BPBA possesses the basic structural characters of both ThT and NSAID mefenamic acid, thus showing Ab-targeting and anti-inflammatory abilities. Its function includes inhibiting Ab aggregation, reducing ROS production, alleviating Ab toxicity, deactivating the NLRP3 inflammasome, and restraining IL-1b release. In vivo studies on transgenic C. elegans show that BPBA can allay the Ab-associated paralysis or Ab toxicity to the neural system of C. elegans. Although BPBA cannot effectively recover the learning and cognitive abilities of AD mice, it can terminate the progression or deterioration of AD. The synergistic impact of BPBA on Ab aggregation and neuroinflammation may bring about a new inspiration for the design of anti-AD drugs. Nevertheless, the exact mechanism of regulating the NLRP3 inflammasome as well as the interactions between NLRP3 inflammasome activation and other signaling pathways in AD remain to be clarifed.

chemsum

{"title": "Concurrent suppression of A\u03b2 aggregation and NLRP3 inflammasome activation for treating Alzheimer's disease", "journal": "Royal Society of Chemistry (RSC)"}

theoretical_study_of_a_derivative_of_chlorophosphine_with_aliphatic_and_aromatic_grignard_reagents:_

## Abstract: The proposed SN2 reactions of a hindered organophosphorus reactant with aliphatic and aromatic nucleophiles [Ye et al., Org. Lett. 19, 5384-5387 (2017)] were studied theoretically in order to explain the observed stereochemistry of the products. Our computations indicate that the reaction with the aliphatic nucleophile occurs through a backside SN2@P pathway while the reaction with the aromatic nucleophile proceeds through a novel SN2@Cl mechanism, followed by a frontside SN2@C mechanism. To the best of our knowledge, this is the first time that a SN2@Cl mechanism is reported. We also found that on reducing the bulkiness of substituents around the phosphorus atom, the backside SN2@P mechanism is preferred. The conclusions made from investigating the steric effect should help experimentalists to decide for the organophosphorus reactant to achieve the products of desired stereochemistry. ## Introduction Organophosphorus derivatives possess significant value in disciplines such as asymmetric metal catalysis, 1 pest control, 2 medicine 3 and bioorganic chemistry. 4 An insight into their synthetic routes is thus essential. 1, One of the reactions often encountered in the synthesis of organophosphorus derivatives is the bimolecular nucleophilic substitution (SN2) reaction at the trivalent tricoordinate phosphorus atom. 5, The SN2 reaction at a trivalent phosphorus atom (SN2@P) refers to the replacement of a substituent (leaving group, L) around the phosphorus atom by an electron-rich molecule (nucleophile, Nu − ). An SN2@P reaction occurs via the inversion and retention pathways in which molecular configuration around the phosphorus atom is inverted (P-inverted) and retained (Pretained), respectively [Figure 1(a)]. 10 The inversion pathway takes place by the attack of the Nu − opposite to L which is synonymous to a backside SN2 pathway (SN2@P-b). Nucleophilic attack can also occur opposite to the lone pair of electrons on the P atom and on the same side as L in the retention pathway; this is analogous to a frontside SN2 attack (SN2@P-f). In the inversion and retention pathways of an archetypal SN2@P reaction, the initial interaction of the organophosphorus and the Nu − reactants (R) generates a reactant complex (RC) which forms a pentacoordinate transition state (TS). The TS dissociates to form the product complex (PC) and finally the separate products (P). The R, RC, TS, PC and P make up the typical double-well potential energy surface (PES) [Figure 1(b)]. 11 An example is the retention pathway of the PH2Cl + Cl − reaction which occurs through a double-well PES with a Berry pseudorotation about the phosphorus atom. 12 Studies also reported atypical PESs for SN2@P reactions when the bulkiness of the substituents around the phosphorus atom is changed. 11, An example is the nucleophilic attack at unhindered trivalent phosphorus atom centres which may result in a single-well PES [Figure 1(c)] with a relatively stable pentacoordinate intermediate, which is known as the transition complex (TC). The TC is in accordance with experimental reports. Such a reaction is spontaneous; the reactants form a stable TC followed by the products without the formation of the RC or PC. An increase in the bulkiness of substituents around the trivalent phosphorus atom separates the TC from the R and P by the pre-TS and the post-TS, respectively. This results in a triple well PES [Figure 1(d)] which constitutes a stepwise Walden inversion with favourable energy barriers in an addition-elimination fashion. 6,10,11,14,20,21 The bulkiness of substituents around the phosphorus atom is one of several factors governing the stereochemical consequences of SN2@P synthesis of organophosphorus derivatives. Other factors include the effect of neighbouring groups, 6,8 solvent 6,22 and choice of nucleophile. 9, An experimental study performed by Ye et al. highlights the choice of nucleophile as a factor influencing the stereoselectivity of SN2@P reactions. 25 The study deals with the synthesis of tertiary phosphines from diastereomeric mixtures of secondary phosphine oxides in tetrahydrofuran (THF) in the presence of boron trihydride (BH3) [Figure 2(a)]. 25 One of the steps in the synthesis [highlighted in blue in Figure 2(a)] involves the replacement of a Cl atom bonded to the trivalent tertiary phosphorus atom by an R group from a Grignard reagent. The authors reported diastereomeric ratios (dr) of 98:2 and 1:99 for the P-inverted and P-retained organophosphorus products in the reactions involving an aliphatic and aromatic Grignard reagent, respectively. 25 Based on these observations, Ye et al. proposed that the aliphatic Grignard reagent attacks opposite to the leaving group Cl (SN2@P-b) and the aromatic Grignard reagent attacks opposite to the lone pair of electrons on the phosphorus atom (SN2@P-f). In continuation to our research programme to understand substitution reactions and our reports on the influence of ion-pair nucleophiles on SN2 reactions, we attempted a computational verification of the reaction mechanisms which were proposed by Ye and co-workers. 25 We investigated the steric effect on SN2 reactions with Grignard reagents (a and b) as ion-pair nucleophiles in the gas phase and in THF [Figures 2(b) and 2(c)]. The presence of BH3 was considered when investigating the reactions in THF so as to mimic the solvation conditions in the experiment, 25 as it is reported that solvent systems constitute a crucial factor in influencing the stereoselectivity of reactions. 29 The occurrence of activation barriers was explained using the distortion/interaction-activation strain model (D/I-ASM) of chemical reactivity. 30,31 The mechanisms that we determined do not only explain the experimental outcome 25 but also constitute a novel synthetic route, via SN2@Cl mechanism, for organophosphorus derivatives which experimentalists may explore in cases where a control over stereoselectivity is required. In addition, to the best of our knowledge, SN2@Cl is being proposed for the first time. ## Computational Details Full optimisation with no symmetry constraints was performed using the B3LYP hybrid functional and the 6-31++G(d,p) basis set in the gas phase. We selected this method based on previous reports involving SN2@P. 20, The stationary points were characterised by frequency computations to verify that the TSs have only one imaginary frequency 39 and these were checked by intrinsic reaction coordinate (IRC) computations. The same computational method was employed based on the polarisable continuum model (PCM) 40 to perform bulk solvation with THF by incorporating BH3 in the reaction system to mimic the experiment conditions. 25 This procedure is reported in the literature. 6 All the computations were carried out at 298.15 K and 1 atm. The reactions are reported in terms of the relative Gibbs free energies (ΔG) with respect to the separate reactants, unless otherwise specified. We performed a non-covalent interaction (NCI) analysis of selected TSs using the Multiwfn program. 41 The NCI plots generated allow for a visualisation of non-covalent interactions between molecular fragments as real-space surfaces. Within the NCI framework, the sign of λ2 enables identification of the interaction type. Attractive interactions are negative, van der Waals interactions occur close to zero and steric repulsions are positive. The Visual Molecular Dynamics program was used to visualise the molecules in 3D. 42 The Gnuplot 4.2 program 43 and Ghostscript interpreter 44 were employed to generate the 2D plots. Moreover, the D/I-ASM of chemical reactivity was employed to explain the trends in reactivity along the reaction. 30,31,45,46 In the D/I-ASM, the energy along the reaction coordinate is decomposed into its strain and interaction contributions [Equation S1; supplementary information (SI)] which originate from the rigidity and orbital overlap of deformed reactants, respectively. The D/I-ASM was also applied in THF to investigate the effect of solvation (ΔEsolvation) (Equations S2 and S3). The reaction coordinate was considered as the IRC projected onto P‧‧‧Cl bond stretch, which constitutes the activation strain diagram (ASD). In another attempt to explain the outcome of the experiment computationally under the experiment conditions employed by Ye et al., 25 the effect of a temperature of 193.15 K was considered on the ΔG; the results are reported in the SI. Relative electronic energies (ΔE) at both at 298.15 K and 193.15 K are also reported in the SI. The ExcelAutomat tool facilitated file manipulation and file extraction. GaussView 6, 50 Chemcraft 51 and CylView 52 were employed to visualise structures. All computations were carried out in Gaussian 09 53 and Gaussian 16 54 suites using the SEAGrid computing facilities. Explicit details about the computational methods are found in the SI. ## Mechanistic explanation of diastereoselectivity In this section, we report on the mechanism by which the R1I + a and R1I + b reactions occur in order to explain their diastereoselectivity. 25 Depending on the orientation of the nucleophiles, we located several SN2@P-b and SN2@P-f pathways in the gas phase and in THF (See SI). Only those with the lowest Gibbs free activation barriers are reported in the manuscript (Figure 3). Both the gas-phase SN2@P-b and SN2@P-f pathways occur through a TS in which the Mg 2+ cation interacts with both the Cl and C atoms and this causes the Cl-P-C bond angle to bend (Figure 3). The bent Cl-P-C bond angle of the SN2@P-b pathways is contrary to the typical near-linear TSs which are reported for anionic SN2-b reactions. 14,37 This reflects the effect of the presence of the countercation Mg 2+ . The small relative Gibbs free energy barriers (ΔΔG ≠ ) between the TSs of SN2@P-b and SN2@Pf pathways do not explain the drs (98:2 and 1:99 for the P-inverted and P-retained organophosphorus products, respectively) obtained in the experiment. This mismatch between experiment and computations prompted us to consider the solvent effect. We studied the solvent effect by incorporating a BH3 molecule in the R1I reactant and performing bulk solvation with THF as the solvent. Under such conditions, the SN2@P-b pathways feature TSs with near-linear Cl-P-C bond angles. The PESs adopt a unimodal or a hill-and-well shape, which differs from anionic reactions. 14,15,29 The ΔΔG ≠ of 2.6 kcal mol −1 of the R1I + a reaction (Figure 3) indicates that the SN2@P-b pathway is preferred. Indeed, the P-inverted organophosphorus stereoisomer was obtained in a higher dr (98:2). 25 Hence, the aliphatic nucleophile a prefers to react via the inversion pathway, as proposed by Ye et al. 25 The ΔΔG ≠ (5.2 kcal mol −1 ) of the R1I + b reaction in THF shows that, similar to the R1I + a reaction, the P-inverted stereoisomer should be obtained. However, the dr, higher for the P-retained stereoisomer (1:99), shows that the SN2@P-b mechanism is not at play. Hence, to explain the observations of Ye et al. 25 , we went back to the basics of charge distribution in a molecule. We therefore employed ESP maps to determine the possible manners of attack, or approach, of the nucleophile b. Owing to the Coulombic nature of molecular interactions, ESP maps serve as good indicators of molecular regions which favour particular interactions; for instance, Bhasi et al. expanded the use of ESP maps to the successful location of TSs for atmospheric reactions which were previously thought to be barrierless. 60 The ESP map of the organophosphorus reactant R1I [Figure 4(a)] shows a σ-hole at the extremity of and a belt of negative potential on the Cl atom. Previous reports related the σ-hole to the anisotropic charge distribution of the Cl atom in the molecule and to the occurrence of halogen-bonded complexes. In this study, the σ-hole gives rise to a halogen-bonded RC [RC in Figure 4(c)] with an electronic interaction energy of −1.9 kcal mol −1 ; this qualifies as a halogen bonding interaction (that is, interaction between a halogen and a base) based on the study by Metrangolo et al. 64 Hence, we investigated the possibility of a reaction mechanism which is initiated by a halogen-bonded RC. The ΔG of TS1 [+43.9 kcal mol −1 , Figure 4(b)] is lower than the Gibbs free activation energy of the inversion pathway of the SN2@P reaction (+53.9 kcal mol −1 ) by 10 kcal mol −1 . Hence, this explains the 1:99 dr obtained by Ye et al. for the P-retained product. 25 We thus confirmed that the reaction with nucleophile b occurs via an SN2@Cl mechanism followed by an SN2@C-f mechanism. In order to investigate the solvation, strain and interaction factors behind the trend in reactivity of the R1I + a and R1I + b reactions, we employed the D/I-ASM. An analysis of the ASDs in Figure 5 shows that the SN2@P-b pathways are more stabilised by bulk solvation by THF than the SN2@P-f pathways. This happens because of the change in geometries along the PESs; the reactants along the SN2@P-b pathways deform to generate TSs with near-linear Cl-P-C bond angles (169.4° in TSIa-i and 175.3° in TSIb-i), whereas along the SN2@P-f pathway, the reactants deform and lead to a TS with bent Cl-P-C bond angles (61.7° in TSIa-r and 72.0° in TSIb-r). There is more charge separation along the pathways forming the TSs with linear Cl-P-C bond angles than along pathways leading to cyclic geometries. The ΔEsolvation curve for the SN2@Cl pathway indicates that the PES is most stabilised by THF at the start of the reaction (C-Cl-P bond angle is 174.7°). As the INT1 is formed, the drop in charge separation causes less stabilisation by THF. The R1I + a reaction experiences more strain in the SN2@P-b pathway which we ascribe to the umbrella flip about the phosphorus atom and to more geometrical deformation to reach the linear TSIa-i in THF. However, the SN2@P-b pathway has a lower Gibbs free energy barrier than the SN2@P-f pathway due to more favourable interaction between the deformed reactants and due to more stabilisation by THF. The initial stages of both the SN2@P pathways of the R1I + b reaction have similar strain curves while the SN2@Cl pathway experiences higher strain. However, as TS1 forms, the ΔEstrain lowers slightly [shown in magnified section in Figure 5(b)] and the ΔEinteraction between the deformed reactants becomes comparable to the SN2@P pathways. As the reaction progresses, the ΔEstrain stays lower than the SN2@Pf pathway and comparable to the SN2@P-b pathway and eventually becomes highest due to the extent of deformation of the reactant fragments involved in reaching the INT1I. The interplay between the stabilisation by THF and reduction in strain while the TS forms that causes the SN2@Cl mechanism to be the most preferred. Hence, we confirmed that (i) the R1I + a reaction yields the P-inverted product in higher dr through an SN2@P-b pathway and (ii) the R1I + b reaction has a preference for the halogen bond-assisted SN2@Cl pathway over the SN2@P-b pathway by a factor of 10 kcal mol −1 . The next question that we raised was whether the use of less bulky substituents can tune the reaction towards a product of specific stereochemistry. ## Steric effect The effect of steric congestion around the P atom in an organophosphorus reactant is a well-investigated aspect of the SN2@P reaction: 11,14,15 An increase in the bulkiness of the substituents of the organophosphorus reactant changes the PES from a single-well to a triple-well and back to the archetypal double-well PES (Figure 1). However, this consensus was reached upon investigation of anionic SN2@P reactions. In this section, we built on this consensus and showed that TCs, and hence single-and triple-well PESs, are not located in ion-pair SN2@P reactions of trivalent organophosphorus reactants. Figure 6 shows the TSs of the SN2@P-b and SN2@P-f reactions in the gas phase and in THF. The ΔG of the RCs, TSs, PCs and Ps are listed in Table 1. Table 1: ΔG, in kcal mol −1 , of the RCs, TSs, PCs and Ps along the SN2@P-b and SN2@P-f pathways of the reactions of R1n with nucleophiles a and b in the gas phase and in THF. The values corresponding to the bulk-solvated phase are in brackets. An increase in the substituent size in the R1n organophosphorus reactants causes the Cl-P-C bond angle of the TSs of the SN2@P-b pathways to bend (Figure 6), allowing for more interaction between the Mg 2+ cation and the Cl leaving group. In the SN2@P-f pathways, the Mg 2+ cation interacts with both the C atom of the nucleophile and the Cl atom of R1n on the same side of the P atom. An analysis of Table 1 shows that gas-phase ion-pair SN2 reactions occur mostly through unimodal PESs or through hill-andwell PESs with the PC as the minimum point, contrary to anionic SN2 reactions. 11,14,15 The preference of the pathway depends on the steric hindrance around the P atom of the organophosphorus reactants. ## Inversion pathway On increasing the bulkiness of the substituents of R1n, we observed that (i) the PESs of the SN2@P-b pathways (R1I + a and R1II + b) shift back to the typical double-well shape; (ii) there is a general decrease in exergonicity in the reactions and; (iii) there is a general decrease in the ΔΔG ≠ of the pathways. In line with the third observation, R1II has a preference for the SN2@P-f with both nucleophiles a and b. The trend in the PESs is explained through the D/I-ASM in the SI. In the THF-solvated SN2@P-b pathways, the Mg 2+ cation is attracted to the high electron density of the B atom of the BH3 moiety. The TSs of the SN2@P-b pathways have near-linear Cl-P-C bond angles. The combined effect of bulk solvation by THF and the ion-pair nucleophile renders the PESs of the SN2@P-b and SN2@P-f pathways (mostly) unimodal with only the TS as the stationary point. The reactions also become more exergonic than in the gas phase. As the bulkiness of the substituents increases, the Gibbs free energy barrier that reactants must overcome to reach the TS increases and the general decrease in exergonicity persists (Table 1). The SN2@P-b pathways have a lower Gibbs free energy barrier than the SN2@P-f pathways with both nucleophiles a and b in THF. Next, we compared the SN2@P-b and SN2@Cl pathways of the R1n + b reactions in THF to investigate how the presence of less bulky substituents affects the preference of the pathway. Figure 7 illustrates the bulk-solvated TSs of the SN2@Cl and the consequent SN2@C-f pathways of the R1n + b reactions. Table 2 lists the ΔG of the RCs, TSs, INTs and PCs of the SN2@Cl and SN2@C-f pathways. A comparison of Tables 1 and 2 shows that, similar to the R1I + b reaction, the Gibbs free energy barrier of the SN2@P-b pathway of the R1II + b reaction is higher (+57.8 kcal mol −1 ) than that of the SN2@Cl pathway (+48.5 kcal mol −1 ). The reactions of R1III, R1IV and R1V investigated occur through the SN2@P-b pathway in THF. This shows that the nucleophile b may approach the reactive backside of the P atom better when the P atom is less crowded by the presence of small substituents. As the bulkiness of the substituent increases, competing mechanisms, such as the SN2@Cl mechanism, become more favourable. Between the C and Cl atoms and the Cl and P atoms, an attractive isosurface is observed enclosed by a repulsive wall at sign(λ2)ρ = 0.02 a.u. Finally, the non-bonded overlap between the C atoms of the phenyl ring causes a steric repulsion shown by the red region at sign(λ2)ρ = 0.04 a.u. Similar isosurfaces were observed in all the TS1n. In the TS1I and TS1II, areas of van der Waals interaction and steric repulsion were also observed between the B atom and surrounding H atoms of the substituents, and between the B atom and surrounding C atoms, respectively. Figure 9 shows the changes in strain, interaction and solvation which are responsible for the trends in the PESs of the R1n + a and R1n + b reactions. A comparison of ΔEsolvation along the SN2@P-b and SN2@P-f pathways in Figures 9 (a) shows that THF stabilises all points on the PES of the SN2@P-b pathways to a greater extent than those of the SN2@P-f pathways. This happens because of the charge separation along the PESs. The reactants along the inversion pathways deform to generate TSs with linear Cl-P-C bond angles, whereas along the retention pathways, the reactants deform and lead to TSs with a cyclic geometry. There is more charge separation, and hence more stabilisation by THF, along the pathways forming the TSs with linear Cl-P-C bond angles than along pathways featuring TSs with cyclic geometries. At the start of the reactions, in the presence of small substituents (R1III, R1IV and R1V), the deforming reactant fragments experience more ΔEstrain and less stabilising ΔEinteraction in the SN2@P-f pathway than in the SN2@P-b pathway. As the reactions progress, the ΔEstrain becomes comparable along both pathways and ΔEint becomes more stabilising along the SN2@P-f pathway. The SN2@P-b pathways are preferred to the SN2@P-f pathways due to more stabilisation by THF and due to lower deformational strain in the initial stages of the reaction. As the substituents of the organophosphorus reactant increase in size in R1II [Figure 9 (a)]and R1I [Figure 5 (a)], the ΔEstrain is higher throughout the inversion pathway. However, the lower Gibbs free energy barrier (Table 1) of the SN2@P-b pathway is due to more stabilising ΔEinteraction between deforming reactants and more stabilising ΔEsolvation along the pathway. Hence, the R1n + a reactions proceed through the SN2@P-b pathway. When comparing the ASDs for the R1n + b reactions [Figure 9 (b)], the same trends in strain, interaction and solvation as in the R1n + a reactions were observed for the SN2@P pathways. The SN2@Cl pathways experience the most stabilisation by solvent before the TS is formed, at the TS and even after the formation of the TS. This is because of higher charge separation caused by the linearity of the C‧‧‧Cl‧‧‧P backbone. The ΔEsolvation becomes less stabilising than the SN2@P pathways as the geometry becomes increasingly non-linear while the PESs progress towards the INT1n. The deformed reactants along the SN2@P-f pathway of the R1II + b reaction experience the least strain throughout the PES. However, the ΔEinteraction and ΔEsolvation have the least stabilising effect on the PES of the SN2@P-f pathway. Similar to the R1I + b reaction, the SN2@Cl pathway is favoured due to lower strain and better stabilisation by THF. As the size of the substituents decreases, the ΔEinteraction of the SN2@P-f pathways of the R1n + b reactions becomes the most stabilising. However, the SN2@P-f pathways have the highest energy barrier due to the highest ΔEstrain and the least stabilisation by THF. The ΔEstrain along the SN2@Cl pathways is lower to or comparable to the SN2@P-b pathway in the initial stages. As the reactions progress, the strain component becomes higher than that of the SN2@P-b pathway. The ΔEinteraction remains less stabilising than that of the SN2@P-b pathway. Although THF stabilises the SN2@Cl pathways the most, the interplay between the ΔEstrain and ΔEinteraction makes the SN2@P-b pathways preferred as the size of the substituents decreases. ## Conclusions Ye et al. substituted the chlorine atom in a derivative of chlorophosphine and obtained the P-inverted and P-retained organophosphorus products on using aliphatic and aromatic Grignard reagents, respectively. 25 In this manuscript, we determined the reaction mechanisms of these two reactions (R1I + a and R1I + b) using the B3LYP/6-31++G(d,p) method. We also highlighted the use of a model of the THF-solvated reaction system with a dative bond between the P and B atoms to account for the presence of the BH3 molecule. The lower Gibbs free activation barrier of the SN2@P-b pathway of the R1I + a reaction explains the formation of the P-inverted product in THF. Our D/I-ASM analyses attribute the preference of the SN2@P-b pathway to more favourable interaction between the deformed reactants and more stabilisation by THF. In the case of the R1I + b reaction, the SN2@P mechanism fails to explain the formation of the P-retained product. In fact, our computations point to a halogen bond-assisted SN2@Cl mechanism. The SN2@Cl mechanism occurs with a lower free Gibbs activation barrier than the SN2@P pathways and is followed by an SN2@C-f mechanism. The preference of the SN2@Cl mechanism resulted from greater stabilisation by THF and reduction in strain while the TS1I forms. We also investigated the effect of decreasing steric crowding of the phosphorus atom of the R1n organophosphorus reactants. The combined effect of the ion-pair nucleophile and bulk solvation by THF renders the SN2@P PESs mostly unimodal and the reactions become more exergonic than in the gas phase. As the bulkiness of the substituents decreases, the R1n + a reactions occur through the SN2@P-b pathways, the decrease in bulkiness lowers the Gibbs free energy barriers of the SN2@P-b pathways on account of more stabilisation by THF and lower deformational strain in the initial stages of the reactions. As the bulkiness of the substituents decreases in the R1n + b reactions, the relative Gibbs free energy barrier of the SN2@P-b pathways becomes lower than that of the SN2@Cl pathway due to the interplay between the strain and interaction components. Hence, a consideration of competing interactions, such as halogen bonding, becomes significant when determining the course of an SN2 reaction, especially in the presence of bulky substituents which hinder the reactive centre. This study should help experimentalists decide which organophosphorus reactant to use so as to obtain a product of required stereochemistry.

chemsum

{"title": "Theoretical study of a derivative of chlorophosphine with aliphatic and aromatic Grignard reagents: SN2@P or the novel SN2@Cl followed by SN2@C?", "journal": "ChemRxiv"}

replacement_of_the_l-iduronic_acid_unit_of_the_anticoagulant_pentasaccharide_idraparinux_by_a_6-deox

## Abstract: One critical part of the synthesis of heparinoid anticoagulants is the creation of the L-iduronic acid building block featured with unique conformational plasticity which is crucial for the anticoagulant activity. Herein, we studied whether a much more easily synthesizable sugar, the 6-deoxy-L-talose, built in a heparinoid oligosaccharide, could show a similar conformational plasticity, thereby can be a potential substituent of the L-idose. Three pentasaccharides related to the synthetic anticoagulant pentasaccharide idraparinux were prepared, in which the L-iduronate was replaced by a 6-deoxy-L-talopyranoside unit. The talo-configured building block was formed by C4 epimerisation of the commercially available L-rhamnose with high efficacy at both the monosaccharide and the disaccharide level. The detailed conformational analysis of these new derivatives, differing only in their methylation pattern, was performed and the conformationally relevant NMR parameters, such as proton-proton coupling constants and interproton distances were compared to the corresponding ones measured in idraparinux. The lack of anticoagulant activity of these novel heparin analogues could be explained by the biologically not favorable 1 C 4 chair conformation of their 6-deoxy-L-talopyranoside residues. Venous thromboembolism is a major cause of mortality and morbidity in the western countries, and epidemiological studies indicate that the aging of the population will increase the incidence of this illness worldwide 1,2 . Medical therapy for venous thromboembolism has been limited to the use of the thrombin inhibitor heparin and the vitamin K antagonists 4-hydroxycoumarins (e.g. warfarin) over 70 years since the 1930's 3 . Then, the 21st century opened a new era for the anticoagulant treatment. The first breakthrough was the approval of fondaparinux, the synthetic analogue of the antithrombin-binding pentasaccharide domain of heparin as a new antithrombotic drug. Fondaparinux is an indirect, selective factor Xa inhibitor possessing a higher safety profile and a longer elimination half-life compared to the animal-originated heparin products 4 . Just a few years later new oral anticoagulant drugs, the direct thrombin inhibitor dabigatran etexilate and the direct factor Xa inhibitors rivaroxaban, apixaban and edoxaban have been approved for clinical use, revolutionising the anticoagulant therapy 5 . Although these new oral anticoagulants have major pharmacologic advantages over vitamin K antagonists, they also have drawbacks and are not approved for some clinical situations . These data predict that the classic anticoagulant drugs, particularly the heparin derivatives will continue to be important medicines for antithrombotic therapy 9 . In the field of heparinoid anticoagulants, current research focuses on unmet issues such as low oral bioavailability of heparin 10 , lack of specific antidote for low molecular weight heparins 11 , and chemoenzymatic production of heparin oligosaccharides . Furthermore, many research efforts have been devoted to the development of highly efficient synthetic routes to the outstanding anticoagulant pentasaccharides fondaparinux and idraparinux 19,20 as well as the preparation of various analogues of the antithrombin-binding pentasaccharide unit of heparin as novel anticoagulant candidates . For a successful synthesis of heparin oligosaccharides a number of factors must be considered such as access to L-idose or L-iduronic acid (IdoA) unit, the choice of uronic acid or the corresponding non-oxidised precursor as building blocks, stereochemical control in glycosylation, suitable protecting-group strategy and efficient assembly of the backbone sequence 26,27 . A range of synthetic approaches have been described to generate heparinoids including solid-supported synthesis 28 , modular approach 29,30 and nonglycosylating strategy 19 . To avoid the inherent low reactivity and base-sensitivity of uronic acid donors, typically the corresponding glycopyranosides are used as glycosyl donors and the formation of the uronic acid can be performed at the disaccharide level 31 or by TEMPO-mediated late-stage oxidation at the higher oligosaccharide level 20,32 . However, each strategy toward chemical synthesis of heparin oligosaccharides faces the same difficulty: the lengthy, laborious and low-yielding synthesis of the L-iduronic acid building block, which is a critically important structural component for the anticoagulant activity. Despite recent progress , the short and efficient synthesis of an orthogonally protected L-idose or iduronic acid glycosyl donor useful for heparinoid synthesis remained unmet. We envisioned that replacing the IdoA with a more easily available sugar unit would solve the problem and L-talopyranuronic acid could be a good candidate as a potential structural substituent. It is known, that a unique conformational plasticity of the L-iduronic acid, shift the 1 C 4 -2 S O equilibrium to the bioactive 2 S O skew-boat conformation, is required for the antithrombotic activity 36 and it was also shown, that its conformation is regulated by the sulphation pattern of nearby saccharides 14,37 . We assume that L-talose, which only differs from L-idose in the C3 configuration, can also adopt the required skew-boat conformation. Moreover, an attractive, short synthesis of L-talopyranosyl thioglycoside, ready for glycosylation has been developed recently 38 . This method, based on iridium-catalyzed CH-activation of the corresponding 6-deoxy derivative 39 can potentially utilize in the synthesis of the L-talopyranuronic acid-containing heparinoid oligosaccharides. As a first step towards this goal we decided to prepare idraparinux-analogue pentasaccharides in which the iduronic acid unit is substituted by a 6-deoxy-L-talopyranoside, a very easily accessible 6-deoxy-L-hexose (Fig. 1). Idraparinux ( 1) is a synthetic anticoagulant pentasaccharide based on the heparin binding domain possessing a higher anti-Xa activity and a longer half-life than the synthetic anticoagulant drug fondaparinux 4 . Being a fully O-sulfated, O-methylated non-glycosaminoglycan analogue, its synthesis is easier than that of fondaparinux, which makes it an ideal model compound. Although the 6-deoxy-L-talopyranoside lacks the biologically important carboxylate moiety, it is suitable for studying the conformational behavior of a talopyranose built in the highly sulphated pentasaccharide. Herein, we report the synthesis and NMR-based conformational analysis of three idraparinux analogue pentasaccharides (2-4) in which the iduronic acid unit (unit G) is substituted by a 6-deoxy-L-talopyranoside moiety (Fig. 1). ## Results Over the last years, we have developed several new methods for the synthesis of idraparinux 20,40,41 . The most efficient strategy involves a 3 + 2 coupling of a FGH trisaccharide acceptor and a DE disaccharide donor, and application of acetyl groups to mask the hydroxyls to be methylated and benzyl ethers to protect the hydroxyls to be sulfated in the final product 20 . We applied the same strategy for the synthesis of compounds 2-4. The preparation of the orthogonally protected 6-deoxy-L-talopyranosyl glycosyl donor 9 was accomplished from the commercially available and cheap 6-deoxy L-hexose, L-rhamnose (Fig. 2). Peracetylation and thioglycosylation of the starting L-rhamnose gave 5 42 , which was converted to the L-talo-configured 6 39 by the well-established C4 epimerization method involving oxidation followed by stereoselective reduction of the corresponding 2,3-O-acetalated rhamnose derivative 43 . Protection of the 4-OH by (2-naphthyl)methylation gave 7 in 93% yield. Finally, the isopropylidene protecting group was replaced by ester groups by deacetalation followed by acetylation to result in the 6-deoxy-L-talopyranoside donor 9 having a C2 participating group capable of ensuring the desired 1,2-trans-selectivity upon glycosylation. Glycosylation of acceptor 10 44 with donor 9 in the presence of N-iodosuccinimide (NIS) and trifluoromethanesulfonic acid (TfOH) led to the formation of the expected 1,2-trans-α-linked disaccharide 11 in 96% yield (Fig. 3). Oxidative cleavage of the 4′-O-(2-naphthyl)methyl (NAP) group using 2,3-dichloro-5,6-dicyanobenzoquinone (DDQ) 45 gave the disaccharide acceptor 12. This disaccharide was isolated together with a small amount of the 2′,4′-di-O-acetyl byproduct due to the undesired acetyl-migration occurred during the column chromatographic purification. Condensation of acceptor 12 with donor 13 20 in the presence of NIS and trimethylsilyl trifluoromethanesulfonate (TMSOTf) resulted in an inseparable 1:1 mixture of the αand β-linked trisaccharides 14 in a moderate yield (Fig. 3). The complete lack of stereoselectivity of the glycosylation was surprising because analogous reactions of 13 with either an L-idose 20 or an L-iduronate 25 acceptor proceeded with exclusive α-selectivity. After an oxidative cleavage of the temporary NAP-protecting group of 14 by DDQ, the desired trisaccharide acceptor 15 was isolated successfully, albeit with low yield. In order to avoid side reactions and increase the yield, the assembly of the FGH trisaccharide was attempted by using the isopropylidenated 6-deoxy-L-talopyranoside derivative 7 as the donor in the first glycosylation step (Fig. 4). To our delight, the glycosylation with donor 7, equipped with a non-participating group at C2, proceeded with exclusive 1,2-trans-selectivity providing the desired α-linked disaccharide 17 as the only product. Although the yield of 17 was moderate upon NIS-TMSOTf promotion, it was significantly increased by changing the Lewis acid in the promoter system to silver triflate (AgOTf). The role of the hindered base sym-collidine in the coupling reactions was to protect the acid-labile isopropylidene group against the acidic conditions of glycosylation. The 4′-OH of 17 was freed by oxidative cleavage of the NAP group using DDQ and the obtained disaccharide acceptor 18 was glycosylated with donor 13 in the presence of a NIS-AgOTf promoter system. Pleasingly, the condensation reaction occurred with complete α-selectivity affording the desired FGH trisaccharide 19 in 60% yield. Finally, removal of the NAP-ether from the terminal glucose unit gave acceptor 20 in 63% yield. Although most difficulties of the first synthetic route to the acceptor FGH was overcame by changing the ester protected 6-deoxy-L-talopyranosyl building block 9 for the 2,3-acetal-protected donor 7 in the second route, we were not satisfied with the overall yield of this latter procedure. Therefore, we tested a third route for the preparation of trisaccharide FGH in which the phenylthio-α-L-rhamnopyranoside derivative 5 was used as the talose-precursor building block and the C4 epimerization of this unit was carried out at the disaccharide level (Fig. 5). Condensation of the L-rhamnose donor 5 and acceptor 10 led to the exclusive formation of the α-linked disaccharide 21 in 98% yield. After Zemplén deacetylation and a subsequent isopropylidenation, the two-step C4 epimerization, involving oxidation with pyridinium chlorochromate (PCC) followed by stereoselective reduction using NaBH 4 , proceeded with high efficacy affording the 6-deoxy-L-talose-containing disaccharide 18 in an excellent 71% overall yield from 21 via 22 and 23. Disaccharide 18 was glycosylated with 13, as described in the previous route. The 2′,3′-O-isopropylidene acetal moiety of 19 was changed to ester groups by acidic hydrolysis of the acetal group followed by acetylation of the obtained 24 to give 14 in 90% yield over the two steps. Oxidative removal of the 4″-O-NAP ether provided the desired FGH acceptor 15 in 77% yield. The assembly of the targeted pentasaccharides was carried out by coupling of the trisaccharide acceptors 15 and 20 with the non-glucuronide type DE disaccharide donor 25 46 (Fig. 6). According to our previously established strategy, we planned the oxidation of the glucose precursor E into D-glucuronic acid at the pentasaccharide level 20,24,41 . Condensation reaction of the isopropylidene-containing trisaccharide acceptor 20 and donor 25 upon NIS-TMSOTf activation provided the needed pentasaccharide 26 together with its diol derivative 27 formed by partial loss of the acetal group under the acidic conditions of the coupling. The products were unified by cleavage of the isopropylidene group of 26 to give diol 27 which was then acetylated to obtain the fully protected pentasaccharide 28 in 74% yield. The advantage of this protecting group pattern was that all hydroxyls to be methylated or freed in the final products were masked with the same acetate ester groups while the hydroxyls to be sulphated were protected in form of benzyl esters. Condensation of the 2′,3′-di-O-acetylated trisaccharide acceptor 15 and donor 25 provided a direct access to pentasaccharide 28. While the coupling showed only moderate efficacy upon NIS-TMSOTf promotion, changing the promoter system to NIS-TfOH compound 28 was formed in an excellent 90% isolated yield. Towards synthesis of the final products 3 and 4, the first transformation at the pentasaccharide level was the removal of the temporary (2-naphthyl)methyl group from unit E followed by (2,2,6,6-tetramethylpiperidin-1-yl) oxyl (TEMPO) and [bis(acetoxy)iodo]benzene (BAIB) mediated oxidation 47,48 of the freed primary hydroxyl group of 29 to produce the glucuronate-containing pentasaccharide 30 in form of a sodium salt (Fig. 7). Then, catalytic hydrogenolysis gave the debenzylated 31 in 94% yield, sulphation of which using excess SO 3 / Et 3 N complex in DMF afforded, after treatment with Dowex Na + ion-exchange resin, the partially acetylated idraparinux-analogue final product 3 as an octasodium salt in 77% yield. The cleavage of the acetyls with the use of 3 M aqueous NaOH solution resulted in 4, another idraparinux-analogue derivative containing free hydroxyl groups at units E and G. Transformation of compound 28 into the fully O-methylated and O-sulphated final product required a different pathway. It is known that uronic acid residues are prone to suffer β-elimination in the basic conditions of the etherification 46,49 . Therefore, prior to the oxidative formation of the glucuronide residue, compound 28 was deacetylated under Zemplén conditions and the obtained 32 was methylated in the presence of methyl iodide and sodium hydride. After the efficient methyl etherification, compound 33 was converted into the glucoronate derivative 35 in high yield by NAP-deprotection followed by TEMPO-BAIB oxidation of the freed hydroxyl at unit E. Finally, the hydroxyl groups to be sulphated were debenzylated by catalytic hydrogenolysis to give 36 in 86% yield. Subsequently, simultaneous O-sulphation of the seven freed hydroxyls was achieved with high efficacy using excess SO 3 /Et 3 N to give, after treatment with Dowex Na + ion exchange resin, the target compound 2 in 91% yield. The structure of all synthesized pentasaccharide derivatives was corroborated by 1 H and 13 C NMR spectroscopy. The unambiguous assignment of NMR resonances was achieved by combined use of 1D and 2D homo-and heteronuclear NMR spectroscopic methods, including COSY, TOCSY, ROESY, HSQC and HMBC experiments. In heparin-related oligosaccharides the L-iduronic acid and L-iduronic acid 2-sulphate residues exist in a dynamic equilibrium of the 1 C 4 , 2 S O and 4 C 1 conformers where the chair 1 C 4 and the skew-boat 2 S O are the predominant conformational forms (Fig. 8A) 14,15,36 . Each conformer has a unique set of three bond proton-proton coupling constants 3 J(H,H) which have recently been calculated by Liu and co-workers 14 using Amber 14 with GLYCAM06 parameter 50 (Table 1, calculated values). To investigate the conformational behavior of the functionally most critical unit G and to assess the relative population of the skew-boat ( 2 S O ) conformer, known to be essential for binding to antithrombin, the structurally relevant 3 J(H,H) couplings were measured for the novel talo-derivatives (2-4) and compared to the calculated values as well as to the corresponding data of idraparinux 1 40 , which was used as reference compound in the present study (Table 1). The distributions of conformers were also studied by 1 H-1 H NOE analysis. The NOE cross peak intensities are uniquely sensitive to detect the presence of the skew-boat conformer due to the significant difference in the atomic distance between H2 and H5 in the 2 S O conformational form (2.6 ), compared with the 1 C 4 (4.0 ) (Fig. 8A). The shorter H2-H5 interproton distance in the 2 S O conformer offers a stronger NOE intensity. Liu and co-workers demonstrated 14 that together with the analysis of 3 J(H,H) couplings, the ratio of H2-H5 and H4-H5 NOE intensities can be used to estimate the conformer population of the L-iduronic acid residue. As shown in Table 1, the 3 J(1,2) and 3 J(2,3) values for each talo-analogue are considerably smaller than the ones measured in idraparinux, thus suggesting that the relative population of the biologically favorable skew boat ( 2 S O ) conformer is reduced in the talo-derivatives and the conformational equilibrium is shifted towards the functionally less preferred 1 C 4 chair conformer (Fig. 8B). The analysis of H2-H5 and H4-H5 NOE intensity ratios (Table 1) also confirms this shift of the conformational equilibrium. The small values of NOE ratios, ranging between 0.1 and 0.2, suggest that a major 1 C 4 chair conformer with longer atomic distance between H2 and H5, correspondingly with weaker H2-H5 NOE crosspeak, indeed exists in D 2 O solution. (The interproton distances within unit G estimated from the volume integrals of ROESY cross peaks are summarized in Table S2). Finally, the inhibitory activities of pentasaccharides 2-4 towards blood-coagulation proteinase Xa were determined by using Berichrom heparin assay (Table S1). Unfortunately, almost complete loss of the factor Xa-inhibitory activity was observed. ## Discussion Heparin and heparinoid anticoagulants exert their anticoagulant activity by binding and activation of antithrombin (AT) which is an endogenous inhibitor of serine proteases in the coagulation cascade 51 . It has been known that plasticity of the L-iduronic acid unit of heparin, an easy shift from the equilibrium state of the preferred 1 C 4 and 2 S O conformations to the 2 S O skew-boat form, is crucial for the stabilization of the activated conformation of AT 36,52 . A detailed knowledge has been accumulated on how the sulphation pattern of neighbouring sugar residues and the sulphation of iduronic acid itself influence the conformational preference of this pyranosyl unit 14,15,37,53,54 . However, the precise structural requirements of the conformational plasticity is not known and it has not been studied whether other L-sugars, incorporated in a heparin structure, can adopt the bioactive 2 S O conformation. Considering the complicated synthesis of the iduronate building block, the possible substitution of this unit by a more easily available monosaccharide, possessing the required conformational plasticity, would be of great importance. In this work, we replaced the iduronate residue of the synthetic anticoagulant idraparinux by a 6-deoxy-L-talopyranose, which is the most easily available L-hexose epimer of L-idose, and studied the conformational behavior and biological activity of the obtained pentasaccharides. Beside the closest, fully O-sulphated and fully O-methylated analogue of idraparinux, a partially methylated and a methylated/acetylated derivative was also synthesized. The assembly of pentasaccharide skeleton was accomplished by coupling a 6-deoxy-L-talopyranoside-containing trisaccharide acceptor to a non-glucuronide disaccharide donor and the glucose precursor was oxidized to the required glucuronide at the pentasaccharide level. The key building block, the 6-deoxy-L-talose-containing trisaccharide FGH was prepared through three different reaction paths, the shortest and most efficient route was when a phenylthio-α-L-rhamnopyranoside was used as the precursor building block and its conversion to the talo-configured unit G was performed at the disaccharide level. While we successfully simplified and shortened the preparation of a heparinoid anticoagulant by replacing the synthetically most demanding unit, the biological activity was almost completely lost. The conformational analysis of pentasaccharides 2-4, based on 1 H NMR ROESY measurements, revealed that the critically important unit G predominantly populated the functionally less preferred 1 C 4 chair conformation. It was also shown that differences in the methylation pattern of the pentasaccharides had no effects on the conformer distribution of the talo-residue. The observed loss of the biological activities could be attributed to the lack of the essential carboxylic moiety of unit G as well as to the less abundance of the bioactive 2 S O conformer in the conformational equilibrium. We assume, that conversion of the 6-deoxy-L-talose residue to L-taluronic acid using established methods 38,47 and introduction of a 2-O-sulphate moiety, which is present in the natural AT-binding sequence, might push the conformational equilibrium of unit G toward the crucial 2 S O form. It is very important to note that although literature data show correlation between the affinity of heparin oligosaccharides and the relative population of the 2 S O form of iduronic acid in solution 55 , recent results of Liu, Guerrini and co-workers 14,15 revealed that predominant population of the 2 S O skew boat conformer of iduronic acid in free form is not a prerequisite for the activation of AT. They demonstrated that a synthetic heparin hexasaccharide, iduronate residue of which is displayed 73% of 1 C 4 conformer in solution, can efficiently activate antithrombin and the iduronic acid adopts a 2 S O conformation when bound to AT. These results indicate that although conformational analysis is very important, the biological test can not be avoided before making a final judgment on the anticoagulant activity of a compound. Further studies to find the aurea mediocritas, to simplify the structure and synthesis of heparin oligosaccharides to an extent that the compounds could keep the biological activity, are under way. ## Methods Optical rotations were measured at room temperature on a Perkin-Elmer 241 automatic polarimeter. TLC analysis was performed on Kieselgel 60 F 254 (Merck) silica-gel plates with visualization by immersing in a sulfuric-acid solution (5% in EtOH) followed by heating. Column chromatography was performed on silica gel 60 (Merck 0.063-0.200 mm) and Sephadex LH-20 (Sigma-Aldrich, bead size: 25-100 mm). Organic solutions were dried over MgSO 4 and concentrated under vacuum. One-(1D) and two-dimensional (2D) 1 H, 13 A solution of compound 31 (78 mg, 0.073 mmol) in dry DMF (4.0 mL) was treated with SO 3 /Et 3 N (461 mg, 2.545 mmol). After stirring for 48 h at 50 °C, the reaction mixture was neutralized with a saturated aqueous solution of NaHCO 3 (1.069 g, 12.72 mmol). The resulting mixture was concentrated. The crude product was treated with Dowex ion-exchange resin (Na + ) and purified by column chromatography on Sephadex G-25 (H 2 O) to give compound 3 (100 mg, 77%) as a white foam. ## -L-talopyranosyl)-(1→4)-(2,3,6-tri-O-sulfonato-α-D-glucopyranoside)] (4) To a solution of compound 3 (50 mg, 0.028 mmol) in MeOH (1.2 mL) a solution of NaOH (3 M, 600 μL) was added at 0 °C. When complete conversion of the starting material into a main spot had been observed by TLC analysis (24 h at room temperature), the mixture was neutralized with AcOH and all volatiles were evaporated. The crude product was purified by column chromatography on Sephadex G-25 (H 2 O) to give compound 4 (35 mg, 78%) as a white foam.

chemsum

{"title": "Replacement of the L-iduronic acid unit of the anticoagulant pentasaccharide idraparinux by a 6-deoxy-L-talopyranose \u2013 Synthesis and conformational analysis", "journal": "Scientific Reports - Nature"}

preventing_the_coffee-ring_effect_and_aggregate_sedimentation_by_<i>in_situ</i>_gelation_of_monodisp

## Abstract: Drop-casting and inkjet printing are virtually the most versatile and cost-effective methods for depositing active materials on surfaces. However, drawbacks associated with the coffee-ring effect, as well as uncontrolled aggregation of the coating materials, have impeded the use of these methods for applications requiring high control of film properties. We now report on a simple method based on covalent cross-linking of monodisperse materials that enables the formation of thin films with homogeneous thicknesses and macroscale cohesion. The coffee-ring effect is impeded by triggering gelation of the coating materials via a thioacetate-disulfide transition which counterbalances the capillary forces induced by evaporation. Aggregates are prevented by monodisperse building blocks that ensure that the resulting gel resists sedimentation until complete droplet drying. This combined strategy yields an unprecedented level of homogeneity in the resulting film thickness in the 100 nm to 10 mm range. Moreover, macroscale cohesion is preserved as evidenced by the long-range charge transfer within the matrix. We highlight the impact of this method with bioelectrocatalysts for H 2 and NADPH oxidation. Peak catalytic performances are reached at about 10-fold lower catalyst loading compared to conventional approaches owing to the high control on film cohesion and thickness homogeneity, thus setting new benchmarks in catalyst utilization. Preventing the coffee-ring effect and aggregate sedimentation by in situ gelation of monodisperse materials † Introduction Drop-casting methods and ink-jet printing are attractive for the assembly of functional flms owing to their scalability, their ease and speed of implementation, as well as their suitability for most types of surfaces and materials. However, the interplay of mass and heat transport processes taking place within the drying droplets leads to deposits which often considerably depart from the desired flm morphology 1 and thus restrain their applicability. In particular, the so-called "coffee-ring effect" (CRE) is omnipresent in drop-casting. 2 It is due to solvent evaporation during flm assembly that stimulates capillary flows within the drop that in turn displace particles to the three-phase contact line. As a result, particle accumulation at the dry flm boundaries is accelerated, and a coffee-ring is formed. 3 The width of the ring is often in the micrometer range, meaning that the major fraction of the area initially covered with the droplet remains mostly unmodifed. Common solutions to circumvent the CRE 4 rely on the engineering of the properties of the droplet, 5 of the surface, 6 of the solvent and cosolvent, 7,8 or of the particles to be deposited. In particular, aggregate sedimentation induced by physical interparticle agglomeration, addition of gelling agents, 12,13 chemical or photochemical cross-linking 7,14 can be used to promote vertical deposition while suppressing radial flow and thus a more uniform deposit is formed. However, the properties of drop-cast flms containing additives or made of aggregates often suffer from insufficient cohesion, homogeneity and flm thickness uniformity. These requirements are fundamental for numerous applications based on active flms such as organic semiconductors 15 for photovoltaics, 16 inorganic catalysts for water splitting, 17 or (bio-) molecular catalysts for sensing 18 and energy conversion. 19,20 The thickness of the flm is of particular importance since it defnes the factors governing the catalytic current or photocurrent generation. Therefore, if the thickness is heterogeneous because of the presence of aggregates, some parts of the flm will have sub-optimal dimensions which will be detrimental to the overall catalytic or photoactive performances. Moreover, breaks in matrix structure due to aggregate boundaries would interrupt key features such as long-range charge transfer pathways in electrocatalytic/photovoltaic flms, 25 energy transfer in light emitting systems, 26 plasmonic effects in macroscopic 3D superlattices 11 or network integrity in macroscale stimuli responsive materials. 27 For these reasons, methods for reproducible formation of flms with controlled and nearly homogeneous thicknesses as well as macroscale interconnection are highly desirable to enable their practical applications. Here, we demonstrate that drop-casting of monodisperse building blocks followed by in situ gelation prior to complete droplet drying circumvents both coffee-ring effects and aggregate formation. We take advantage of a mild crosslinking chemistry based on thioacetate terminated macromolecules 19,21,28 to induce gelation. The coating materials based either on polymeric or on dendrimeric scaffolds were functionalized with viologen moieties serving both as redox relays and fluorescence reporters (Fig. 1a). The key factor enabling the formation of uniform and highly cohesive flms is the ability to maintain gel homogeneity during droplet drying. While dropcasting and gelation of polydisperse polymeric materials efficiently suppresses the CRE, the sedimentation of polymer aggregates that are larger than the nominal thickness of the resulting flm leads to excessive thickness distributions (Fig. 1b). In contrast, the use of dendrimers as monodisperse building blocks in place of polydisperse macromolecules results in flms that have highly homogeneous thicknesses and that are free of breaks in their macroscale 3D matrix as evidenced by the electron accessibility throughout the complete volume of the flm (Fig. 1c). The applicability of the dendrimer hydrogel flms as the immobilization matrix is verifed for bioelectrocatalytic systems, namely hydrogenase and ferredoxin-NADP + -reductase (FNR), which both yielded high current densities and excellent stability under constant turnover for H 2 oxidation and NADPH oxidation, respectively. ## Structure of the viologen-modied macromolecules The viologen-modifed dendrimer (Scheme S1, ESI †) was synthesized starting from a polyamidoamine (PAMAM) core (3 rd generation, 32 head groups). The synthesis of the viologen-modifed polymer was based on a branched polyethylenimine (PEI) backbone according to a previously reported procedure. 19 In both cases, the same thioacetate terminated viologen groups were used for post-functionalization (Fig. 1a). The high density of primary amine groups on the PAMAM dendrimer and on the PEI backbone enables high loadings with viologen moieties upon covalent attachment via isothiocyanate chemistry (Scheme S2, ESI †). 1 H NMR characterization of the viologen-modifed dendrimer demonstrates almost quantitative functionalization (99% yield, Fig. S1, ESI †). A similarly high post-functionalization yield is achieved for the viologen-modifed polymer according to UV-Vis measurements (Fig. S2, ESI †). ## Spontaneous crosslinking and gelation The thioacetate termination on the viologen head groups serves as a crosslinking functionality for in situ gelation under mild conditions (Fig. 1). The thioacetate hydrolyzes in aqueous solutions, 29 and the resulting free thiol groups subsequently oxidize in the presence of O 2 to form disulfde bonds, leading to crosslinking between the macromolecules. Moderately basic buffers (pH 9) were used to accelerate both the thioacetate hydrolysis and the thiol oxidation reactions. Gelation of the viologen-modifed macromolecules was investigated in solution by dynamic light scattering (DLS), which is well suited for monitoring changes in macromolecule size and size distributions without disturbing the gelling system. 30 Under anaerobic and reducing conditions, the viologen-modifed polymer shows a high polydispersity with the main fraction of coil size in the 10-30 nm range, as well as presence of aggregates in the 100-500 nm (0.6%) and 3-4 mm (0.2%) range (Fig. 2a). In contrast, the dendrimers display an average diameter of 5.57 AE 0.04 nm (Fig. 2b) with very low distribution which is in agreement with the nanoscale dimension 31 and molecular nature of the viologen-modifed dendrimer. When the solutions are exposed to air, the average size for both the polymer and the dendrimer increases due to crosslinking via oxidation of the thiol groups by O 2 (Fig. 2c and d). After 6.5 h, the polymer displays a broadening of the main peak as well as an increase of the shoulder in the range from 30 nm to 50 nm (Fig. 2c). In the case of the dendrimer, only a very small fraction of particles (<0.2%) with a diameter of a few hundred nanometers appears after 6.5 h, and the main peak shows an increase in particle diameter from around 5 nm to 8 nm, while maintaining a very low polydispersity (Fig. 2d). This indicates that the monodisperse dendrimeric systems undergo crosslinking without signifcant aggregation in contrast to the gelation of the branched polymeric materials. The absence of large particles in the DLS of the crosslinked dendrimers implies the formation of a homogeneous gel. The relatively long time needed for gelation for both macromolecules under the DLS conditions is attributed to slow diffusion of O 2 into the large volume of the DLS cuvette (2 mL) and to the time needed for initial neutralization of the reducing agent (tris(2-carboxyethyl) phosphine, TCEP). ## Hydrogel lm formation and lm morphology Film formation was performed by drop-casting microliter sized droplets (1 to 4 mL) onto glassy carbon surfaces. Fluorescence microscopy was used to qualitatively study the effect of dropcasting conditions on flm morphology (Fig. 3). Fluorescence arises from the excitation of the viologen moieties. The measurements were performed in a confocal setup so that fluorescence intensities are related to the local concentrations and local thicknesses of the viologen modifed flms (see Methods section). In a frst set of experiments, the crosslinking of the macromolecules was hindered by using an excess concentration of TCEP as the reducing agent to prevent disulfde bond formation. The fluorescence microscopy images of the flms derived from the polymer and from the dendrimer both display ring-shaped structures with intensive fluorescence emission on the periphery of the flms (Fig. 3a and b). These are the typical coffee-ring features which result from the transport of the macromolecules to the three phase boundary of the droplet due to the capillary forces induced by solvent evaporation. 7 In a subsequent set of experiments, the concentration of the reducing agent was lowered so that crosslinking can take place before droplet drying. The low droplet volume combined with the large air-solution interfacial area allows for fast saturation of the solution with O 2 , ensuring a rapid establishment of oxidative conditions which induces the crosslinking via disul-fde bond formation. The fluorescence microscopy images reveal that the resulting flms now cover the complete drop-cast area and the coffee rings are fully suppressed (Fig. 3c and d). This demonstrates that the crosslinking via spontaneous disulfde bond formation results in a material that is unaffected by the capillary forces induced by droplet drying. Remarkably, the disulfde crosslinking eliminates the coffee-ring effect even in the case of the dendrimer molecules despite their spherical shape which is particularly prone to undergo coffee-ring deposition. 9 The second important observation is that the morphologies of the flms based on the monodispersed dendrimer are highly uniform in contrast to the polymer based flms which exhibit numerous and large aggregates. This is in agreement with the respective gelation behavior observed for the two materials in the DLS investigation. The aggregates of polymers precipitate and thus yield a rough flm while the dendrimer gel remains in solution and deposits evenly upon complete droplet-drying. Remarkably, the smoothness of the latter flm matches previously reported systems obtained by the addition of a gelling agent. 13 The ability to achieve flm homogeneity without additives is a valuable feature of the present approach since it ensures that the flm composition can be entirely defned by the intended application rather than the parameters required to obtain uniform deposits. The effect of aggregation on the morphology of the flm is further characterized by optical and atomic force microscopy (AFM). The optical microscopy images of the flms resulting from drop-casting of viologen-modifed polymers display numerous aggregates having diameters in the 20-50 mm range (Fig. 4a), whereas the images of the flms of viologen-modifed dendrimers are relatively smooth without any visible aggregates over an equally large sampling area (Fig. 4b). These observations corroborate the fluorescence microscopy investigations. AFM imaging of the polymer flms (Fig. 4c) coupled to phase analysis (Fig. 4e) reveals that the aggregates and the relatively smooth adjacent regions are of the same composition, implying full but uneven coverage of the surface by the polymer flm. These aggregates are in the micrometer magnitude according to the surface roughness measurements (Fig. 4g and i). In contrast, the AFM images of the dendrimer flm display a uniform, smooth surface over an area of 100 mm 100 mm (Fig. 4d). Phase analysis, surface roughness and 3D images also confrm the homogeneous coverage with the dendrimer flm over the complete sampled area (Fig. 4f, h and j). The mean thickness of the dendrimer flm obtained by AFM upon scratching of the surface was S4 (ESI †). For both the polymer and the dendrimer, the glassy carbon disks were modified with a surface coverage of 200 mg cm 2 and the gelation process was carried out with a total volume of 2.5 mL, including 0.5 mL Tris buffer solution (100 mM, pH 9.0) at RT in a water saturated atmosphere. This journal is © The Royal Society of Chemistry 2018 Chem. Sci., 2018, 9, 7596-7605 | 7599 found to be (3.5 AE 0.3) mm and was consistent for variable locations on the electrode surface (Fig. S3, ESI †). ## Electroactive lm thickness distribution The electroactive thickness distributions of the flms were subsequently investigated through cyclic voltammetry. 32 The main beneft of using an electroanalytical method in comparison to AFM is that it probes the portion of the flm that is accessible to electrons and therefore, the portion of the flm matrix that is break-free and able to contribute to electrochemical processes. Additionally, the whole surface of the electrode is probed instead of a selected area. In this electrochemical method, cyclic voltammograms (CVs) of the redox flm are measured at different experimental time scales defned by the scan rates (n) (Fig. S4A and B, ESI †), and the normalized peak currents (i p /n 1/2 ) are plotted versus n 1/2 . The experimental i p /n 1/2 vs. n 1/2 curve for the dendrimer flm shows only minor deviations from the theoretical current response for a perfectly smooth flm which indicates a highly uniform thickness distribution (Fig. S4D, ESI †). In comparison, the corresponding dimensionless peak current plots for the viologen-modifed polymer substantially deviate from the curve corresponding to a perfectly homogeneous flm, implying that the polymer flms are highly aggregated (Fig. S4C, ESI †). Probability distribution functions of the flm thickness for the polymer and the dendrimer flms (Fig. S4E and F, ESI †) were extracted by analysis of the deviations from the theoretical i p /n 1/2 vs. n 1/2 curves, and 3D representations of these flms were generated with arbitrary positions of the individual flm subsections (Fig. 4k and l). The relative standard deviations (s), expressed as a percentage of the mean, were determined for the dendrimer and polymer flms based on the underlying flm thickness distribution, and were found to be 10.7% and 169% respectively. The 3D representation of the polymer flm (Fig. 4k) illustrates the presence of the micrometer scaled aggregates, which are in good general agreement with the one observed in the optical and AFM images (Fig. 4a and g). In contrast, the 3D representation of the dendrimer flm shows a very smooth surface in which both "coffee rings" and aggregates are absent (Fig. 4l). The high degree of uniformity of the flm thickness distribution and the average thickness (2.6 mm) obtained from the electrochemical method are in quantitative agreement with the corresponding AFM investigations, which indicates that the entire redox-active flm is accessible to electrochemical processes, and thus, that the electron transfer pathway within the complete flm volume is free of blockages or breaks. ## Effect of gelation conditions on lm homogeneity The morphology of flms obtained from the precipitation of a colloidal solution within a drying droplet generally depends on the wetting properties of the surface 33 as well as on its tilt angle. 34 In particular, the coffee ring effects are exacerbated for pendant droplets and on hydrophobic surfaces. To test the general applicability of the in situ gelation method for uniform flm formation presented in this work, we compared the assembly of the dendrimer based hydrogel flms on gold and carbon surfaces. Optical microscopy (Fig. S5A and B, ESI †) and AFM images (Fig. S5C and D, ESI †) show flms of low roughness on both Au and C substrates. The flm thickness distributions extracted from electrochemical investigations (Fig. S5E-H, ESI †) are low on both materials (s Au ¼ 8.7% and s C ¼ 10.7%). This demonstrates that interactions between the materials for deposition and the surface of the substrate do not signifcantly impact the resulting flm thickness homogeneity. Additionally, flm preparations were peformed flm preparations on Au surfaces with different tilt angles (Fig. 5) to validate the mechanism for uniform flm formation. The flm resulting from a pendant droplet (180 ), as well as the flm from a droplet sticking to a vertically orientated surface (90 ), is highly homogeneous and comparable to the one obtained from a sessile droplet (0 ). The low thickness distributions obtained for these flms (s 0 ¼ 8.7%, s 90 ¼ 16.8% and s 180 ¼ 4.8%) regardless of the orientation and surface properties indicate that the flm formation is not a result of aggregation and sedimentation of the cross-linked macromolecules. This corroborates that the mechanism is based on in situ gelation of the dendrimers followed by even deposition of the resulting hydrogel onto the surface as it dries. ## Electrocatalytic applications In electrocatalytic applications, flm thickness homogeneity and electron transfer throughout the entire volume of the flm are important factors that defne the resulting catalytic performance. Film morphology from AFM images for tilt angles of (b) 0 , (e) 90 and (h) 180 . 3D representations derived from the electrochemical method for determination of the film thickness distribution for tilt angles of (c) 0 , (f) 90 and (i) 180 . The corresponding electrochemical data, secondary plots and probability distribution functions are given in Fig. S6-S8 (ESI †). The droplet volume for film formation was 2 mL for the tilt angles of 0 and 180 , and was 0.5 mL for the 90 tilt angle. The smaller droplet size in the latter case was chosen to prevent droplet deformation due to gravity. The applicability of the dendrimer based hydrogel was tested for NADPH oxidation catalyzed by the redox enzyme ferredoxin-NADP + -oxidoreductase (FNR) (Fig. 6a) as well as for H 2 oxidation catalyzed by the metallo-enzyme Desulfovibrio vulgaris MF (DvMF) NiFe-hydrogenase (Fig. 6b). In both cases, the viologen moities serve as electron relays between the electrode and the redox enzymes within the hydrogel flm. The bioelectrocatalytic flms were prepared by drop-casting a mixture of dendrimer and of the respective enzyme on Au electrodes. The viologen moieties confer the redox properties to the flm, which appears colorless in the oxidized state and dark blue in the reduced state (Fig. 6c). Long-range charge transfer takes place through electron hopping between viologens and is measured as an apparent electron diffusion coefficient (D e ). 25 The value of D e for the dendrimer flms is (1.15 AE 0.1) 10 8 cm 2 s 1 (see Fig. S9, ESI †), which is about twice as high as the D e value obtained for viologen-modifed polymer flms ((4.7 AE 1.7) 10 9 cm 2 s 1 ). 21 For FNR immobilized in dendrimer modifed flms, the catalytic current for NADPH oxidation reaches a maximum of 0.65 mA cm 2 (Fig. 6d). Similarly, the hydrogenase immobilized in the dendrimer flms delivers catalytic current density up to 0.4 mA cm 2 for H 2 oxidation (Fig. 6e). In contrast, the use of viologen-modifed polymer instead of dendrimer as the building block for the formation of flms containing FNR resulted in maximum catalytic currents for NADPH oxidation of only 0.1 AE 0.04 mA cm 2 (Fig. S11, ESI †). This 6-fold lower current is rationalized not only by the lower D e value in the polymer, but also by the aggregates present in the polymer flm, which results in breaks in the electron transfer pathway and electrochemical inaccessibility of the catalysts. Moreover, the flm thickness (l) defnes the factors limiting the catalytic current (i cat ). Using theoretical models which account for the reaction-diffusion processes at play for a given set of flm parameters, predictions of i cat vs. l can be constructed. For thin flms, catalyst loading limits the current, and therefore, i cat vs. l is linear with a slope defned by the catalytic properties (enzyme concentration C E and catalytic rate constants k A ) according to eqn (1). 22 For thicker flms, electron hopping within the flm typically becomes limiting and therefore a plateau in i cat is expected, reaching a value which depends on the electron transfer properties according to eqn (2). 22 where A, n, F and C A are the electrode surface area, the number of electrons transferred, the Faraday constant, and the concentration of the electron mediator (the viologen moiety) within the flm, respectively. Additional limiting cases related to mass transport or enzyme kinetics may also arise. 22 If the thickness is heterogeneous, some parts of the flm will be in one limiting regime, whereas other parts of the flm will be in another limiting regime which will result in i cat vs. l behaviors that deviate from the theoretical predictions. Hence the i cat vs. l plots can be used to qualitatively assess the heterogeneity of the fraction of the flm that is electrocatalytically active. Previous reports have shown that the use of polymers enables the formation of flms with redox enzymes which display behaviors following eqn (1) or (2) but typically require more complex procedures, such as layer-by-layer assembly. 22,35 In contrast, such control on bioelectrocatalytic response through flm thickness adjustment has never been reported through simple drop-casting of redox-active polymers. In the case of drop-casting of dendrimers on electrode surfaces, the average flm thicknesses can be adjusted with the amount of the dendrimer and enzyme used for flm formation (Fig. S12, ESI †). Remarkably, trends for i cat vs. l of the modifed electrodes are in agreement with the behavior expected from the theoretical reaction/diffusion models for electrocatalytic flms (eqn (1) and ( 2)). 22,23 For FNR immobilized in a dendrimer matrix, the catalytic current for NADPH oxidation linearly scales with the flm thickness (l) up to approximately 3.5 mm and subsequently transitions to a plateau (Fig. 6d). A similar behavior is observed for catalytic current for H 2 oxidation obtained from hydrogenase modifed electrodes (Fig. 6e). The clear experimental transition from a catalyst loading limitation in the thin flm regime (linear region) to a limitation by electron transfer as the flm thickness increases (plateau region) confrms that the use of viologen-modifed dendrimers enables the formation of bio-electrocatalytic flms with controlled and uniform thicknesses, and therefore allowing for the optimal utilization of the redox catalyst. The long term stability of dendrimer modifed flms for bioelectrochemical processes within viologen-modifed dendrimer/ enzyme electrodes was investigated under conditions of continuous turn-over. These conditions were set by applying a constant potential, which induced the oxidation of the substrates (NADPH or H 2 ) over long time periods ranging from hours to weeks (Fig. 6f and g). The catalytic current for the dendrimer-FNR electrodes remained completely stable within the frst 5 hours of continuous NADPH oxidation. The dendrimer-hydrogenase electrode was maintained under continuous turnover conditions for H 2 oxidation for 2 weeks, which resulted in only a 39% decrease in current. In addition, considering that the catalytic current is recorded under strongly oxidizing conditions by applying a potential of 0 V vs. Ag/AgCl at the electrode, which are deactivating for NiFehydrogenases, 36 the stability of the current in the dendrimer modifed flms exemplifes how the hydrogenase-dendrimer flms remain immune to this oxidative stress owing to the Nernst buffering properties of the viologen containing matrix. 19,28 The stability and bioelectrocatalytic performances obtained for the thin dendrimer flms (2 mm for the FNR-modifed electrode and 4.1 mm for the hydrogenase-modifed electrode) compete with the one previously reported for the viologen-modifed polymer flms with comparably much higher thicknesses (100-300 mm). 21 ## Conclusions The crosslinking and gelation of monodisperse materials open up the possibility to use drop-casting methods for the assembly of functional flms with a high degree of homogeneity and fne control of the flm thickness by circumventing aggregate sedimentation and coffee-ring effects. Dynamic light scattering, fluorescence, optical and AFM imaging, as well as electrochemical and electrocatalytic investigations, support a mechanism for flm formation involving in-droplet gelation followed by homogeneous deposition on the complete drop-cast area. Aggregate formation, which is intrinsic to classical methods based on precipitation processes used for coffee-ring suppression, is completely absent when using in situ gelation of dendrimeric materials as building blocks. Importantly, the cohesion of the flm, as demonstrated by the long-range charge transfer pathways accessing the full volume of the flm matrix, enables the optimal exploitation of the immobilized catalysts with performance in agreement with theoretical prediction for perfectly homogeneous flms. Although the present study was based on dendrimeric materials, the underlying flm formation mechanism allows extrapolation of the applicability of this drop-casting concept to any materials that enable homogeneous gel formation before droplet drying. Moreover, from a practical perspective, a remarkable feature of this drop-casting method is that the resulting flm thickness homogeneity is independent of the nature or tilt angle of the surface and the mild crosslinking conditions are compatible even for highly sensitive catalytic systems such as redox enzymes. As such, we anticipate that the concept will serve as a general basis for flm formation of various active materials for applications ranging from electrocatalysis to organic photovoltaics or organic electronics. To this end, the extension from drop-casting to inkjet printing 1 of monodispersed building blocks will be particularly valuable for small surface area coating or non-flat surfaces that are not accessible to classical deposition methods as well as for low-cost, industrial-scale printing for applications requiring a high degree of flm homogeneity. ## Materials and methods Unless stated otherwise, all reagents used in the experiments were purchased from Sigma-Aldrich. b-Nicotinamide adenine dinucleotide phosphate tetrasodium salt (NADPH) was purchased from GERBU Biotechnik GmbH. All the materials were directly used as received without further purifcation. ## Synthesis of PAMAM G3 dendrimer with thioacetate functionalized viologen terminal groups (Schemes S1 and S2, ESI †) PAMAM G3 dendrimer in methanol solution (290 mL, 7.2 mmol) was placed in an oven-dried Schlenk tube ftted with a stir bar, and the methanol was removed under vacuum. The resulting viscous oil was dissolved in dry DMSO (5 mL) and degassed with several cycles of vacuum and argon. After addition of isocyanate terminated viologen 19 (1.5 equivalents, 201 mg, 347.4 mmol), the solution turned deep blue. The mixture was then stirred at room temperature for three days. After the mixture was decanted into an aqueous solution of KNO 3 (0.1 M, 150 mL), the solution instantly turned yellow-green. The KNO 3 solution was placed into a fltration unit ftted with a 5 kDa size exclusion membrane, and was fltered overnight with distilled water. The water was then removed under reduced pressure resulting in a glassy, dark green solid (145 mg, 88%). ## Protein purication DvMF [NiFe] hydrogenase was purifed as described previously. 37 The petH gene coding for the FNR of Nostoc sp. PCC7120 was amplifed from genomic DNA excluding the 5 0 -end which covers 138aa of the cpcD like domain. The PCR product was cloned into the pASK-IBA7 expression vector (IBA GmbH) lacking the Strep tag II sequence. Wild type FNR proteins of Nostoc sp. PCC 7120 were overexpressed in 3 liter LB medium (Lennox) using E. coli strain BL21(DE3) as a host. Expression was induced upon attaining an OD 600 of 0.6 by the addition of 200 mg anhydrotetracycline per liter host culture. After further cultivation for 4-5 h the cells were harvested by centrifugation and resuspended in 100 mM Tris-HCl pH 8, 10% glycerol. Cell disruption was achieved by passing twice through a French pressure cell press (1000 psig). Cell debris was separated from soluble proteins by ultracentrifugation (45 000 rpm, 1 h, 4 C) and the supernatant subsequently cleaned from residual impurities by batch incubation treatment with 1 g DEAEcellulose (Whatman DE52) and two fltration steps (folded flter 2V Whatman, Filtropur S 0.2 mM). The clear fltrate was supplemented by ammonium sulfate ((NH 4 ) 2 SO 4 ) to 40% saturation and a frst fraction of contaminating soluble proteins separated from the target protein by centrifugation (10 000 rpm, 10 min, 4 C). Further (NH 4 ) 2 SO 4 was added to the supernatant up to 70% saturation. After centrifugation (see above) the yellowish precipitate containing the FNR was resuspended in buffer 2 (5 mL, 25 mM Tris-HCl, pH 7.5) and dialyzed overnight against the same buffer. During the following chromatography steps fractions containing the FNR were easily identifed by their deep yellow coloring. Anion exchange chromatography on a DEAE Sepharose (CL-6B, Sigma) column (1.5 20 cm) was performed using a 50 mM step gradient of NaCl (50-500 mM) in buffer 2, yielding the FNR within the 200 mM fraction. The eluate was supplemented with (NH 4 ) 2 SO 4 up to 30% saturation prior to hydrophobic interaction chromatography done on a phenyl sepharose column (Fast Flow, Sigma). The protein is recovered between 20 and 17% in a 3% ammonium sulfate step gradient of (32-2% saturation). Yellow fractions were pooled, dialyzed against buffer 2 and concentrated up to 1 mM FNR. Target protein purity was verifed by SDS PAGE and Coomassie staining. ## Fluorescence characterization Fluorescence measurements were performed on a Leica Microsystems TCS SP8 CARS laser scanning microscope in a fully confocal optical setup. An argon gas laser excited the samples at 488 nm through a HC PL Fluotar 10x/0.3 dry objective while the emission was detected in EPI direction with photomultipliers roughly between 500 and 650 nm. Each measurement consisted of 512 or 1024 pixels in both x and y directions and up to 38 stacks in z direction, with pixel sizes of down to 1 mm and stack heights of 4 mm. The extension of the measurements into the third dimension allows qualitative assessment of the thickness of the sample and its (sub-) surface structure (e.g. the coffee ring effect or aggregates). Up to 25 (5 5) measurements were combined to one mosaic to show the full extent of the samples. At the end all z-planes were projected (maximum intensity projection) onto a single plane to allow a two dimensional comparison of the mosaics. ## AFM characterization The AFM measurements were conducted in the AC mode by NanoWizard 3 (JPK) with cantilever of the type NSC15 (Mikro-Marsch). The details of flm assembly are provided in the related fgure legends. ## Electrochemical and size characterization All electrochemical measurements were performed at room temperature through a Gamry Potentiostat and an Autolab PGSTAT12 Bipotentiostat. A platinum wire and Ag/AgCl/3 M KCl were used as the counter electrode and reference electrode, respectively. The details of electrode modifcation are provided in the related fgure legends. DLS was applied to determine the size and size distribution via a Malvern Zetasizer Nano ZS.

chemsum

{"title": "Preventing the coffee-ring effect and aggregate sedimentation by <i>in situ</i> gelation of monodisperse materials", "journal": "Royal Society of Chemistry (RSC)"}

synthetic_control_and_empirical_prediction_of_redox_potentials_for_co₄o₄_cuban

## Abstract: The oxo-cobalt cubane unit [Co 4 O 4 ] is of interest as a homogeneous oxygen-evolution reaction (OER) catalyst, and as a functional mimic of heterogeneous cobalt oxide OER catalysts. The synthesis of several new cubanes allows evaluation of redox potentials for the [Co 4 O 4 ] cluster, which are highly sensitive to the ligand environment and span a remarkable range of 1.42 V. The [Co III 4 O 4 ] 4+ /[Co III 3 Co IV O 4 ] 5+ and [Co III 3 Co IV O 4 ] 5+ /[Co III 2 Co IV 2 O 4 ] 6+ redox potentials are reliably predicted by the pK a s of the ligands. Hydrogen bonding is also shown to significantly raise the redox potentials, by $500 mV. The potential-pK a correlation is used to evaluate the feasibility of various proposed OER catalytic intermediates, including high-valent Co-oxo species. The synthetic methods and structure-reactivity relationships developed by these studies should better guide the design of new cubane-based OER catalysts. ## Introduction Research on catalysts for the oxygen evolution reaction (OER), motivated by the goal of creating an artifcial photosynthesis system, has generated a number of hypotheses that challenge the traditional limits of transition-metal oxidation states and bonding. Thus, catalytic OER cycles often invoke unusually high and rare oxidation states for the metal centers, which are bound to reactive terminal oxo ligands. In particular, for intensely studied cobalt-based OER catalysts, a Co IV -oxo intermediate is commonly invoked. 3,6,9 However, the Co IV oxidation state is quite rare, and ligand feld considerations seem to suggest a high instability for a terminal oxo ligand bound to Co IV . 10 Recently, several experiments have confrmed the presence of Co IV species during water oxidation catalysis, but evidence for a discreet, terminal Co IV oxo species remains indirect. 3,9,11,12 Remarkably, many density functional theory (DFT) calculations of cobalt-catalyzed OER suggest an even more unconventional species, a Co V -oxo intermediate that is sometimes described as a Co IV -oxyl radical. 4,5 Experimental evidence for this type of intermediate is provided by kinetic analysis of OER mediated by a molecular Co 4 O 4 cubane, which is a rare example of a structural and functional model for the active site of OER catalysis. 6a For the oxo cubane Co 4 (m-O) 4 (OAc) 4 (py) 4 a singly oxidized state, formally Co III 3 Co IV , is frmly established by isolation and spectroscopy. 3,6a The doubly oxidized state, formally Co III 2 Co IV 2 , is observed at highly positive potentials, by cyclic voltammetry. 13 While the kinetic analysis points to involvement of the doubly oxidized state in the oxygen evolution mechanism, the nature of intermediates associated with this oxidation state remains largely unknown. For evaluation of possible Co III 2 Co IV 2 (or alternatively Co III 3 Co V ) intermediates, it would be quite useful to establish reliable methods for predicting the redox potentials associated with particular coordination environments. For example, the Co 4 O 4 cubane Co 4 (m-O) 4 (OAc) 4 (py) 4 has been shown to form the hydroxide complex [Co 4 O 4 (OAc) 3 (OH) 2 (py) 4 ] by displacement of acetate, and kinetic studies on OER indicate that this species is oxidized to high-valent species. 6 The resulting, transient intermediates have not been directly observed; therefore, alternative, indirect experimental methods for evaluation of candidate structures are useful. As shown here, a strategy for estimation of Co IV /Co V redox potentials for transitory intermediates is based on extrapolation of linear free-energy relationships (LFERs). This analysis requires a large and diverse set of related cubanes with various ligand sets and redox potentials, to provide a useful LFER from which redox properties can be confdently predicted. In this report, we demonstrate that the Co 4 O 4 core is readily and precisely manipulated to tune its chemical and electronic properties over an unprecedented range. While the parent cluster Co 4 O 4 (OAc) 4 (py) 4 has been well-studied, its controlled structural modifcation has never before been demonstrated; such clusters are generally prepared by "self-assembly" methods rather than by rational syntheses. 14 Synthetic methods were used to obtain electron-rich or electron poor cubanes, cubanes with mixed-carboxylate ligand sets, and cubanes possessing secondary-sphere hydrogen-bond donors. It is noteworthy that this synthetic methodology allows introduction of secondary-sphere hydrogen bonding into the cubane structure since the role of hydrogen bonding in electron transfer and water oxidation (especially in the OEC of photosystem II) has been an important topic for many years. 1b-h Analysis of a substantial library of oxo cubanes provides empirical linear correlations between ligand pK a values and redox potentials for singly and doubly oxidized species, over a range of 1.42 V. This analysis also quantifes the effect of hydrogen bonding on redox properties in this cubane system. These relationships offer a useful predictive tool for evaluating potential intermediates in water-splitting mechanisms. They should also provide important guidance in catalyst design studies for OER. ## Synthesis of new cobalt oxo cubanes The synthesis of Co 4 O 4 cubanes has previously been accomplished by the "self-assembly" route. 14b While this method is simple to execute, its harshly oxidizing conditions and less predictable nature prevents the targeted synthesis of many structurally diverse cubane clusters. Notably, the "self-assembly" route only leads to symmetrically ligated cubanes, and has only been demonstrated with ligands that are not very electron rich or electron poor. Given potentially important applications of the Co 4 O 4 cubane unit (vide supra), general synthetic routes that allow ready access to a diverse library of cubanes, including unsymmetrically substituted, electron poor and electron rich examples, are desired. The various cubane complexes to be discussed in this study possess a Co 4 O 4 cubane core with different ligand sets that are described in a concise manner by a convenient descriptor. Compounds of the general formula [Co III 4 O 4 X x L y ] (4x)+ are abbreviated [xX-yL] (4x)+ , where x and y are integers equal or greater than 0, and denote the stoichiometry of X (anionic ligand) and L (neutral ligand), respectively. If multiple types of X or L are present in the same complex, then the additional xX or yL is appended in the fashion Three general methods were used to synthesize the new cubane complexes, shown in Chart 1. The "self-assembly" route (method A, eqn (1)), patterned after the Das synthesis of Co 4 -O 4 (OAc) 4 (py) 4 from Co(NO 3 ) 2 (H 2 O) 6 , NaX, L, and H 2 O 2 , has been used to synthesize a small number of closely related cubanes such as Co 4 O 4 (OAc) 4 (p-cyanopyridine) 4 (4-OAc-4CNpy). 14b,20 However, attempts to extend this method to many of the cubane derivatives described below were unsuccessful. Thus, this does not appear to be a generally successful synthetic strategy for the synthesis of new cubane complexes. Exchange of neutral ligands (method B, eqn (2)) is useful in the synthesis of cubanes with L ¼ DMAP. The cubane Co 4 O 4 (OAc) 4 (DMAP) 4 (4-OAc-4DMAP) was produced in this way, in 67% recrystallized yield. Method B works well in other cases (vide infra), under conditions where pK a (L 0 H + ) > pK a (LH + ). Similarly, synthesis of the cubane with X 0 ¼ CF 3 COO was achieved by heating 4OAc-4py with a slight excess of CF 3 CO 2 H to give Co 4 O 4 (O 2 CCF 3 ) 4 (py) 4 (4TFA-4py) in 56% yield. This procedure (method C, eqn (3)) is driven by release of HX; thus, it works well when pK a (HX') < pK a (HX). Methods B and C provide predictable and divergent routes for cubane diversifcation (Chart 1), and in contrast to method A, give crude products that are mostly free of Co(II) impurities, as determined by thin-layer chromatography. Remarkably, substitution of ligands by methods B and C is highly stereoselective, always placing the X ligands around equatorial faces and the L ligands on the "top" and "bottom" faces, as evidenced by NMR spectroscopy and crystallography (Fig. 1). Signifcantly, method C allows for the synthesis of mixedcarboxylate cubanes. Mono-, di-, or tri-substituted carboxylate cubanes were produced using two RCO 2 H equivalents per cubane. This mixture of mono-, di-, tri-, and tetra-substituted cubanes is generally separable by column chromatography since the compounds are highly colored and exhibit signifcantly different R f values. By this method, Co 4 O 4 (OAc) 4n (CF 3 CO 2 ) n (py) 4 (nTFA,(4-n)OAc-4py) and Co 4 O 4 (OAc) 4n (p-nitrobenzoate) n (py) 4 (nNBA,(4-n)OAc-4py) complexes, where n ¼ 1, 2, and 3, have been synthesized and fully characterized. Interestingly, the formation of disubstituted cubanes is stereoselective for cis-substitutionthat is the two carboxylate (X 0 ) ligands are on adjacent faces of the cubane, as revealed by X-ray crystallography of 2TFA,2OAc-4py (Fig. 2). The 1 H NMR spectra reveal the presence of only one product, suggesting that the crystal structures represent the bulk material. Cubanes bearing electron-rich X ligands 2-pyridonate (pK a (O-H) ¼ 8.05) 15 and 7-azainodolate (pK a (N 1 -H) $ 15) 16 were synthesized in a manner similar to that of method C, but the products have a somewhat different composition (Scheme 1). Thus, heating 4OAc-4py with excess 2-hydroxypyridine in acetonitrile gave Co 4 O 4 (2pyridonate) 4 (2-hydroxypyridine) 3 (py) (4pyrO-3pyrOH,py) in 37% yield, after column chromatography. Elemental analysis and mass spectrometry are consistent with the stoichiometries determined by NMR spectroscopy. Similarly, heating 4OAc-4py with an excess of 7-azaindole in acetonitrile gave Co 4 O 4 (OAc) 2 (7-azaindolate) 2 (7azaindole) 4 (2OAc,2Az-4AzH) in 48% yield. Though the pK a values of 2-hydroxypyridine and 7-azaindole are higher than those of HOAc, the reaction is driven by use of excess ligand and the insolubilities of 4pyrO-3pyrOH,py and 2OAc,2Az-4AzH in acetonitrile. These reactions provide different types of cubane structures, in that protonated versions of the X ligands serve as the L ligands (2-hydroxypyridine and 7-azaindole possess pyridyl nitrogen atoms), and substantial substitution of the L-positions on the cubane is observed. Interestingly, the protonated, L-type ligands engage in hydrogen bonding with their O-H or N-H bonds interacting with the m 3 -oxo ions of the [Co 4 O 4 ] core. The 2-hydroxypyridine ligand can be substituted for other L ligands by method B to produce 4pyrO-4py and 4pyrO-4DMAP. The latter two compounds were characterized by 1 H NMR spectroscopy and single-crystal Xray diffraction (Fig. 3). 1 H NMR spectroscopy suggests that 4pyrO-3pyrOH,py coexists as C 1 -symmetric species in solution at room temperature, suggesting that hydrogen-bonds slow the rotation about the Co-N bonds. For 2OAc,2Az-4AzH, two C 2symmetric isomers co-crystallized, and differ by the relative orientation of the 7-azin ligands. The O(X)/O(X) distances of 2.44 in 4pyrO-3pyrOH,py, and the N(X)/O(X) distances of 2.69 in 2OAc,2Az-4AzH are consistent with strong to moderate hydrogen Scheme 1 Synthesis of cubane complexes containing intramolecular hydrogen-bonding. bonding. 17 Thus, these cubanes possess an additional secondary coordination sphere that affects the structural properties of the cubane core. Secondary coordination sphere effects are highly important in biology, and mimicry of this feature in inorganic complexes is a long-standing challenge that often involves special ligand design. 4). The reversibility of E 2 seems to depend on the combination of X and L, rather than the individual identities of X and L. In general, when there is a large asymmetry in the electron-donating properties of X and L (DpK a > 4), E 2 is irreversible. Four examples that illustrate this trend are 4OAc-4DMAP, 4pyrO-4py, 4pyrO-4DMAP, and 4OAc-4py. The frst two cubanes have two sets of ligands with DpK a > 4, and their cyclic voltammograms indicate an irreversible oxidation process at all measured scan rates (E 2 ). However, the two sets of ligands in 4pyrO-4DMAP and 4OAc-4py have a DpK a < 2, and reversible E 2 events. This observation may suggest that the E 2 redox couple for cubanes with a ligand DpK a > 4 may involve some degree of ligand non-innocence. For example, if there is a large difference between the donating properties of X and L, electron-hole character may concentrate on the more donating ligand, leading to oxidative degradation. Conversely, when the X and L ligands are both good donors, the electron-hole character is distributed more evenly across all eight ligands to give a more stable complex. The electronic structure of the species resulting from this second oxidation will be discussed in later sections. In general, and as expected, the observed redox couples reflect a dependence on the donor properties of the cubane ligand set, such that they shift anodically with electron-withdrawing ligands and cathodically with electron-donating ligands. To evaluate the influence of the L ligands on redox potentials, sixteen cubanes were analyzed holding X ¼ OAc constant while varying L (4OAc-4L, series 1), and conversely holding L ¼ py constant while varying X (4X-4py, series 2). The redox potentials for both series (1 and 2, respectively) correlate linearly with a convenient descriptor for the ligand donating ability, the average pK a (aqueous) 15,16,18,19 of HX and HL + per cobalt center (eqn (4); Fig. 5A). This average pK a value descriptor will herein be referred to as the effective basicity of the ligand set. Here, pK a is employed as a free-energy representation for the electron-donating ability of the ligand. A previous report used Hammett parameters to correlate ligand donor ability with Co 4 O 4 redox activity, but this analysis is limited to only substituted aryl-based ligands, whereas pK a is a much more widely useful parameter. 20 A plot of E 1 vs. effective basicities for the complexes in series 1 is linear with a slope of 120 AE 7 mV dec 1 , and series 2 also provides a linear relationship with a slope of 69 AE 9 mV dec 1 . Similarly, the slopes of lines derived from the E 2 values are very similar, at 128 AE 9 mV dec 1 (series 1) and 51 AE 9 mV dec 1 (series 2). The magnitudes of these slopes demonstrate that the redox couples for the cubanes are about twice as sensitive to the X vs. the L ligand. A possible cause for this effect will be discussed in the next section. Most surprising is that the redox potentials of unsymmetrical, mixed-carboxylate cubanes could be predicted simply by considering the summation of the pK a values. This fact would seem to suggest that the Co 4 O 4 cubane core levels out the electronic effects of coordination asymmetry through an electronically communicative mechanism. In this context, the four cobalt ions in the Co 4 O 4 unit essentially behave as a single "superion" entity. Both series of potentials ft to planar surfaces (eqn (5) and (6), Fig. 5B) with R 2 values of 96% and 98%, respectively. Effective basicity ¼ 1 4 To explore the generality of eqn ( 5) and (6) derived from compound series 1 and 2, the E 1 and E 2 values for cubanes based on additional combinations of X and L were measured and compared to the predicted values. Including these additional combinations of X and L, the E 1 redox potentials observed for tetracobalt oxo cubanes range from 0.30 to 1.12 V (a span of 1.42 V; Table 1). A smaller range of 830 mV was observed for E 2 , perhaps because some of the corresponding oxidation events exist outside the solvent's electrochemical window. Satisfyingly, the redox potentials of all the cubanes (without intramolecular hydrogen-bonds) are predicted by eqn (5) and ( 6) with maximum absolute errors of 0.099 and 0.050 V, respectively. In addition, reported redox potentials for cubanes of the type [Co 4 O 4 X 2 L 2 4 ] 2+ (L 2 ¼ 2,2 0 -bipyridine) follow the relationships described by eqn (5). 21 This observation indicates that the relative number of X to L ligands does not affect the validity of eqn (5), and therefore the coefficients must reflect the intrinsic donor properties of X-versus L-type ligands. In summary, it appears that eqn (5) and (6) hold for (1) a wide range of X and L, (2) symmetric and asymmetric ligand coordination environments around the cubane, and (3) X ligands other than carboxylates. ## Effects of protic solvent and hydrogen bonding The redox potentials reported above were all measured in polar aprotic solvents (MeCN, DCM, or DMF). When measured in water, the E 1 potentials experience an increase in the observed potential, by an average of +553 AE 26 mV (Table 2). This solvent dependence on redox potentials was previously noted for 4OAc-4py, but its origin has not been discussed. 6,22 Unfortunately, only a few complexes are water-soluble enough to be examined in this solvent, and E 2 potentials were not observed in water since these oxidation events occur beyond the electrochemical window for water. We propose that the observed increases in redox potentials in water originate from hydrogen-bonding interactions between the cubane m 3 -oxo ligands and water, which reduces the m 3 -oxo donation to cobalt. The relative electronic effects of the cubane ligands appear to persist in water, however, as indicated by a similar E 1 /pK a slope. The hydrogenbonding effect is clearly demonstrated, even in aprotic solvent, by comparing 4pyrO-3pyrOH,py with 4pyrO-4py. The pK a values of 2-hydroxypyridinium cation and pyridinium are essentially identical (5.23 and 5.25, respectively), and according to eqn (5), should give nearly identical E 1 values for the corresponding cubanes. However, the observed E 1 value for 4pyrO-3pyrOH,py is 550 mV higher than that of 4pyrO-4py. In the case of 2OAc,2Az-4AzH, the experimental E 1 value is 519 mV higher than predicted by eqn (5). The observed increases in redox potential with introduction of hydrogen bonding to the cubane (519 and 550 mV) are remarkably similar to that observed by changing the solvent to water (553 mV). ## DFT calculations of cubane electronic structures To quantify the ligands' influence on the redox potentials of cubanes, an energy decomposition analysis (EDA) was performed. The results reveal that the X ligands stabilize the . The frozen electron density term, (FRZ) which describes electrostatic and steric contributions to bonding, is at least an order of magnitude larger for the acetate ligand, and doubles in magnitude upon oxidation. The polarization and charge transfer terms are also generally larger for acetate than for pyridine; however, they are less than an order of magnitude different. The size of the FRZ term indicates that electrostatic forces dominate the stability of the oxidized cubane. This fnding seems intuitive as Co III and Co IV are both considered hard ions, and according to hard-soft acid-base theory should prefer ionic bonding interactions over dative interactions. To uncover trends in the bonding, EDA was also applied to 4TFA-4py and 4OAc-4CNpy. The energy difference in frozen components between [4X-4L] + and 4X-4L (FRZ ox FRZ red ¼ DFRZ) for the [Co 4 O 4 ]-X and [Co 4 O 4 ]-L bonds plot against pK a yields slopes of 0.8 kcal per (mol pK a ) and 0.5 kcal per (mol pK a ), respectively (Fig. S1 †). Note that the energy values are within the error of DFT calculations, but there is a qualitative correlation between theory and the experimental redox trends. The doubly oxidized state, [4OAc-4py] 2+ , was analyzed by DFT calculations. These calculations determined that the S ¼ 1 state is essentially degenerate with the S ¼ 0 state (within 3.3 kcal mol 1 ). Interestingly, the calculations suggest two valence-trapped Co IV centers, as opposed to a valence delocalized system. There have been conflicting reports of hole delocalization in cobalt cubane systems by DFT, and delocalization has been suggested as a product of a self-interaction error. 4,25 In support of a localized valence, UV-vis-NIR ## Cubane Effective basicity X 15,16,18,19 Effective basicity L 15,16,18,19 E 1 (V) Implications for cubane intermediates with hydroxide and oxide ligands Perhaps the most important utility of the energy relationships of eqn (5) and ( 6) is their ability to predict electronic properties for cubane-based intermediates that might be considered in mechanistic investigations. For example, these relationships allow thermodynamic evaluations of proposed hydroxyland oxo-intermediates in the recently proposed mechanism for cobalt cubane-mediated OER. 7)). 6 The bis(hydroxide) was characterized by NMR spectroscopy, and is proposed to be the resting state of the cubane during OER catalysis. However, as noted in this study, the formation of [3OAc,2OH-4py] is low yielding since the equilibrium strongly favors 4OAc-4py, which makes the isolation and direct electrochemical study of this species difficult. No prior knowledge of the relative redox potentials contributed to the formulation of this disproportionation step; its existence is strongly inferred from the second-order rate dependence on [4OAc-4py] + . While this mechanistic study revealed that the [Co 4 O 4 ] 6+ state was achieved and only terminal oxygen ligands were involved, the level of water/hydroxide/oxo substitution and protonation states throughout the OER cycle are not known. For example, does the active species contain two terminal oxo-ligands, or only one, or none? The redox steps involved in this mechanism are undoubtedly coupled with proton transfers, as implicated by the frst-order rate dependence on [OH ], but the ordering and extent of these processes are as yet unknown; in other words, kinetic data alone cannot defnitively distinguish between a mechanism involving only aquo-and hydroxo-ligated cubanes and one involving the more unusual oxo ligand. Thus, if the redox potentials for corresponding oxo-, hydroxo-, and aquo-substituted Co 4 O 4 clusters can be known (or well-estimated), it should be possible to rule out certain possible intermediates (those that cannot reach the [Co 4 O 4 ] 6+ state by disproportionation with [4OAc-4py] + ), and discover which candidates are most plausible. The LFER equations (eqn (5) and ( 6)) allow simple evaluations of these candidate intermediates. Using the conjugate acid pK a values of 15.7 for the OH ligand, and $36 for O 2 , 18,26 E 1 and E 2 redox potentials were calculated for the most likely catalytic intermediates (Scheme 2, Table 3). These E 1 values suggest that oxidation of the catalytic intermediates in Table 3 3. Table 3 Predicted redox potentials for hydroxo-and oxo-ligated cubanes 6 we favor the localized Co V formalism when a strong s and p donor such as O 2 is bound to a single Co ion, a viewpoint that is also supported by DFT. However, when multiple basic ligands (OH and O 2 ) are present, it is reasonable to expect the more delocalized states (e.g. Co IV 2 ) to signifcantly contribute. With this information, a revised OER catalytic cycle can be proposed (Scheme 3). This updated mechanism allows for two kinetically indistinguishable O-O coupling pathways, an acidbase mechanism or a radical-coupling pathway. Recently, Nocera and coworkers obtained direct evidence for the coupling of adjacent cobalt-oxygen species at pH 7 in the CoP i catalyst, and several DFT calculations have shown this pathway to be energetically reasonable. 4,27,28 Nonetheless, these predicted redox values show that cubanes with hydroxo-and oxo-ligands can reach the a formal Co IV 2 or Co V oxidation state by redox disproportionation reactions. ## Conclusions Cobalt oxo cubane complexes are readily manipulated by convenient syntheses that allow exchange of the nitrogen-donor and carboxylate ligands, the latter in a sequential, selective manner. This synthetic chemistry allows tuning of the cubane electronic properties over a wide range. Primary variables that influence the cubane's redox properties are the donor properties of the anionic and neutral ligands, and the incorporation of hydrogen-bond interactions. Two important principles that result from this work, that should guide OER catalyst design, are that (1) the [Co 4 O 4 ] cubane is predictably modifable over a remarkable range of electrochemical and chemical properties, and (2) the accessibility of a formal Co IV 2 /Co V -oxo intermediate by the cubane is energetically feasible given the appropriate ligand set. The strong influence of hydrogen bonding suggests that catalyst design must not only address the primary coordination sphere, but also second-sphere interactions between the catalytic core and the reaction medium. The fundamental reaction chemistry developed in this study provides new strategies for functionalization of the cubane with a vast array of chelating ligands and N-heterocyclic donors, and should allow incorporation of catalytic cubanes into more complex architectures involving biological molecules, surfaces, and 3dimensional networks. Importantly, the linear free energy relationships described above are useful in predicting the redox properties of newly designed cubane-based materials. ## Conflict of interest No competing fnancial interests have been declared.

chemsum

{"title": "Synthetic control and empirical prediction of redox potentials for Co₄O₄ cubanes over a 1.4 V range: implications for catalyst design and evaluation of high-valent intermediates in water oxidation", "journal": "Royal Society of Chemistry (RSC)"}

phytochemical_profiling_and_anti-fibrotic_activities_of_plumbago_indica_l._and_plumbago_auriculata_l

## Abstract: This study aimed at investigating the chemical composition and the hepatoprotective activities of Plumbago indica L. and P. auriculata Lam. LC-MS/MS analyses for the hydroalcoholic extracts of the aerial parts of the two Plumbago species allowed the tentative identification of thirty and twentyfive compounds from P. indica and P. auriculata, respectively. The biochemical and histopathological alterations associated with thioacetamide (TAA)-induced liver fibrosis in rats were evaluated in vivo where rats received the two extracts at three different dose levels (100, 200 and 400 mg/kg p.o, daily) for 15 consecutive days with induction of hepatotoxicity by TAA (200 mg/kg/day, i.p.) at 14th and 15th days. Results of the present study showed a significant restoration in liver function biomarkers viz. alanine transaminase (ALT), aspartate transaminase (AST), gamma glutamyl transferase and total bilirubin. The liver homogenates exhibited increased levels of antioxidant biomarkers: reduced glutathione (GSH) and catalase (CAT), accompanied with decline in malondialdehyde (MDA). Furthermore, treated groups exhibited a significant suppression in liver inflammatory cytokines: tumor necrosis factor-α (TNF-α) and interlukin-6 (IL-6), and fibrotic biomarker: alpha smooth muscle relaxant. Histopathological examination of the liver showed normality of hepatocytes. Noteworthy, P. indica extract showed better hepatoprotective activity than P. auriculata, particularly at 200 mg/kg. To sum up, all these results indicated the hepatoprotective properties of both extracts, as well as their antifibrotic effect was evidenced by reduction in hepatic collagen deposition. However, additional experiments are required to isolate their individual secondary metabolites, assess the toxicity of the extracts and explore the involved mechanism of action. The high incidence of liver toxicity has recently been linked to many factors, the most important of which are significant exposure to high drug doses, massive metabolic activity, and the presence of numerous enzymes believed to be accountable for the generation of reactive metabolites, most notably reactive oxygen species (ROS). Toxic compounds, among them ROS, have the ability of targeting macromolecular structures or specific molecules such as nuclear receptor family members or bile acid transporters, as well as intracellular lipids, nucleic acids, or proteins. These targeted molecules have become defective units that activate secondary pathways, resulting in programmed events such as apoptosis, necrosis, and autophagy, mitochondrial failure, immunological responses, and deposition of collagen which can ultimately lead to liver fibrosis 1 . In the development of chronic tissue injury, fibrogenesis, a dynamic process defined by the ongoing formation of fibrillar extracellular matrix (ECM), is accompanied with constant breakdown and remodeling. When OPEN 1 In vivo investigation. Acute toxicity study. Acute oral toxicity was conducted according to the OECD 423 guidelines 12 . Wistar albino rats (n = 5) were randomly selected and fasted for 4 h with free access to water. Hydroalcoholic extracts of both Plumbago species (suspended in 0.5% NaCMC) were daily administered orally at doses 250, 500 and 1000 mg/kg for each plant extract. The animals were monitored for behavioral alterations, toxic symptoms or mortality. Both plant extracts didn't show any toxic signs, behavior changes or mortality at the tested doses. Experimental animals and design. Fifty-four adult male albino rats weighing 200-220 g were utilized in this study. They were purchased from the animal house of Faculty of Science, Cairo University. The animals were housed in separated in steel mesh cages and kept under standard conditions (ventilation, temperature (25 ± 2 °C), humidity (60-70%) and light/dark condition (12/12 h)) and fed on a standard rat pellet diet and fresh, clean drinking water. The animals were acclimatized for a period of 15 days prior to the beginning of study. Rats were randomly allocated into nine groups (6 rats each). All groups excluding group I received TAA (El-Gomhouria Company for drug and chemicals, Egypt, 200 mg/kg; i.p.) on the 14th and 15th days of the experiment. Group I; rats served as control negative group received saline on the 14th and 15th days of the experiment and vehicle (0.5% NaCMC, Sigma-Aldrich (Merck Millipore, Darmstadt, Germany) daily throughout the experiment. ## Statistical analysis. The significant differences between the means of the groups (mean ± standard error of mean) were analyzed with one-way analysis of variance (ANOVA) followed by the Tukey's post hoc multiple comparisons test. Results were considered as statistically significant when p < 0.05. The statistical analysis and figures were generated using GraphPad Prism 8.4.3. ## Results Phytochemical analyses. Profiling of the phenolic metabolites in the hydroalcoholic extracts of P. indica and P. auriculata aerial parts were performed by LC-MS/MS. In total, thirty and twenty-five secondary metabolites were detected in P. indica and P. auriculata, respectively. Twenty-one compounds were common in both extracts. The compounds were identified and tentatively annotated based on their retention times, molecular weight, and fragmentation pattern as well as comparison with reported data found in literature and online data base (Mass Bank) (Table 1). ## Pharmacological analyses. Effect of pretreatment with hydroalcoholic extracts of the aerial parts of P. indica or P. auriculata on serum hepatic functions biomarkers in TAA-induced liver toxicity in rats. Injection of TAA significantly elevated serum ALT, AST, GGT, and bilirubin levels compared to normal control group values by 3.71, 3.57, 5.74, and 4.99 folds, respectively. Both plants-treated groups significantly reduced the elevated levels of serum liver biomarkers compared to TAA group. Pretreatment with P. indica showed remarkable amelioration of the elevation of ALT, AST, GGT, and bilirubin levels by about 26.69%, 27.61%, 33.24% and 31.19%, respectively at a dose of 100 mg/kg, by about 63.85%, 55.39%, 57.72% and 61.12%, respectively at 200 mg/kg, and by about 42.91%, 42.38%, 45.42% and 56.35%, respectively at 400 mg/kg compared to TAA group. Plumbago auriculata treated groups controlled the rising of ALT, AST, GGT, and bilirubin levels at a dose of 100 mg/kg by about 19.88%, 21.40%, 24.16% and 29.05%, respectively at a dose of 200 mg/kg by about 50.45%, 46.87%, 48.59% and 57.80%, respectively and at a dose of 400 mg/kg by about 37.00%, 35.43%, 40.55% and 39.74%, respectively when compared to TAA group. Treatments of rats with both plant extracts before TAA administration showed an improvement in liver function versus TAA group. These observed results were comparable to those of the reference drug, silymarin (Table 2). Effect of pretreatment with hydroalcoholic extracts of the aerial parts of P. indica or P. auriculata on the hepatic oxidative stress biomarkers in TAA-induced liver toxicity in rats. Induction of liver injury by TAA significantly decreased in antioxidant enzymes evidenced by GSH levels by about 58.17% and CAT levels by about 67.85%, as well as a significant elevation in lipid peroxidation that was measured by MDA values to 5.7 fold of a thiobarbituric acid reactive substance as compared to normal group. On the other hand, Administration of P. indica hydroalcoholic extract at different doses, 100, 200, and 400 mg/kg, significantly increased the lever contents of GSH levels about 56.25%, 125.11% and 75.92%, respectively and CAT levels by about 46.43%, 140.92% and 93.82%, , respectively as well as restrict the MDA elevation by about 70.0% (Fig. 1). Effects of pretreatment with hydroalcoholic extracts of the aerial parts of P. indica or P. auriculata on the hepatic inflammatory biomarkers in TAA-induced injury in rats. Initially, we virtually explored the interactions of the individual secondary metabolite from both extracts with the active site of both hepatic inflammatory biomark- www.nature.com/scientificreports/ ers (TNF-α and IL-6). The docked compounds furnished considerable binding energies and several hydrogen bonding and hydrophobic and ionic interactions, Table 3 and Fig. 2. These results were further confirmed in vivo where injection of TAA significantly elevated liver TNF-α and IL-6 by 1.33 and 3.01 fold, respectively compared to normal group values. Doses of the hydroalcoholic extract of P. indica aerial parts (100, 200 and 400 mg/kg) significantly decreased the raised TNF-α levels by about 8.43%, 19.04% and 11.93%, respectively, as well as IL-6 levels by about 24.44%, 60.54% and 37.40%, respectively compared to TAA control group. The rats pretreated with the hydroalcoholic extracts of P. auriculata aerial parts at doses of 100, 200 and 400 mg/kg showed remarkable reductions in the elevated TNF-α by about 5.92%, 15.57% and 9.12%, respectively and IL-6 by about 17.21%, 53.25% and 35.01%, respectively, compared to TAA control groups. Both extracts showed an obvious reduction in TNF-α and IL-6 levels. Moreover, pretreatment with silymarin showed a reduction by about 23.78% and 65.55% for TNF-α and IL-6 levels, respectively (Fig. 3). Effect of pretreatment with hydroalcoholic extracts of the aerial parts of P. indica or P. auriculata on hepatic α-SMA in TAA-induced liver injury in rats. Induction of liver dysfunction in rats using TAA significantly provoked the hepatic α-SMA by 6.04 fold as compared to the normal group (Fig. 4). Both plant extracts revealed a noticeable protection against the elevation of α-SMA by about 33.13%, 69.61% and 51.45% for 100, 200 and 400 mg/ kg of P. indica extract, respectively and 33.79%, 63.44%, and 44.95% for 100, 200 and 400 mg/kg of P. auriculata extract, respectively compared to TAA group. In spite of this observable protection accompanied with both plant extracts against fibrosis which is comparable by the protection showed by silymarin, there is still different in α-SMA values compared with control group indicating an antifibrotic effect. ## Histopathological examination of liver tissue. In regards to the histopathological examination, there was no histopathological alteration and the normal histological structure of the central vein and surrounding hepatocytes in the parenchyma were detected in normal control rats (Fig. 5A) while, examination of the livers of TAA rats revealed swelling besides vacuolar degeneration observed in hepatocytes (Fig. 5B) and associated with focal mononuclear leucocytes inflammatory cells aggregation in the hepatic parenchyma (Fig. 5C). The hepatic parenchyma showed also focal necrosis, as well as congestion in the portal and central veins (Fig. 5D-F). Collagen fibers deposition was noticed around central vein and portal tract besides pseudolobulation of hepatocytes with fibroblasts was also observed (Fig. 5G, H). Normal hepatic architecture was observed at silymarin treated group (Fig. 5I). Rats protected by P. indica (100 mg/kg) showed fibrosis in the portal area of liver and few inflammatory cells with many apoptotic cells (Fig. 5J) and parenchymal cells with marked apoptosis was also observed (Fig. 5K). Much better improvement was detected in P. indica group (200 mg/kg), as there was a normal hepatic architecture observed (Fig. 5L) while P. indica (400 mg/kg) treated group showed also normal parenchymatous architecture of the liver but there is a little cellular infiltration, inflammatory cell and vascular congestion were noted (Fig. 5M, N). Some livers of P. auriculata administrated rats at the dose of 100 mg/kg showed fine strands of fibrous tissue extending from portal areas toward the parenchyma (Fig. 5O) and the hepatic parenchymal cells are near to the normal appearance with mild vacuolation (Fig. 5P). Liver sections of rats treated by P. auriculata (200 mg/kg) signifying that the portal area and surrounding hepatocytes in the parenchyma was histological normal (Fig. 5Q), but the central veins and sinusoid were dilated (Fig. 5R) while preserved liver architecture against damage was observed but small foci of inflammatory cells were noted in some fields at the rats protected by P. auriculata (400 mg/kg) (Fig. 5S, T). ## Discussion In the current study, phytochemical analyses by LC-MS/MS were performed in order to investigate the bioactive compounds present in of P. indica and P. auriculata. LC-MS/MS analysis revealed the presence of numerous compounds that was tentatively identified for the first time in both plants such as genistein 8-C-glucoside, 3-hydroxy-3' ,4' ,5'-trimethoxyflavone, 6-ethoxy-3(4'-hydroxyphenyl)-4-methylcoumarin and luteolin. The findings of LC-MS/MS analyses also in parallel with previous reports that revealed the presence of plumbagin in both plants 6 . The current study also investigated the protective role of P. indica and P. auriculata hydroalcoholic extracts on liver injury induced by TAA-administration in male albino Wistar rats. To the best of our knowledge, this is the first in-vivo assessment for the anti-inflammatory and anti-fibrotic activities of both plants against TAA-induced fibrosis in rats. TAA is a well-established model for induction of hepatic fibrosis by producing free radicals during its metabolism resulting in oxidative stress mediated acute hepatic inflammation and destruction of hepatocytes in the liver . Thioacetamide metabolites covalently bind to the liver macromolecules causing dramatic elevation of serum levels of ALT, AST, GGT, and bilirubin, a decrease in antioxidant capacity as proved by depletion of GSH and CAT levels and an increase in lipid peroxidation as confirmed by elevation of MDA levels and by increasing ROS production. ROS are shown to enhance pro-inflammatory cytokines production (TNF-α and interleukin-6 (IL-6)). There are also solid links between inflammation, oxidative stress, fibrogenesis and angiogenesis 27 . Liver fibrosis induced by TAA is characterized by several amendments occurred at hepatic ECM 28 . The outcomes of the present investigation showed that pretreatment of TAA-induced hepatic fibrotic rats with P. indica extract at 200 mg/kg exhibited a pronounced protection on the levels of serum hepatic enzymes, antioxidant, anti-inflammatory and anti-fibrotic marker: α-SMA, against the toxic and fibrotic changes induced by TAA. The current study also revealed that the hydroalcoholic extract of P. indica aerial parts has a potency similar to the root alcoholic extract of the same plant that was assessed previously against the same hepatotoxic agent, TAA 29 . Moreover, administration of both extracts ahead of TAA reduced the elevated enzyme levels in serum plasma resulting from the stabilization of these biomarkers, clearly showing a preventive effect of the plants on TAA intoxication. Reduced levels of GSH have been associated with TAA induced hepatitis and are closely www.nature.com/scientificreports/ correlated to the lipid peroxidation and disturbance of calcium ions induced by toxic agents 30,31 . In extractstreated groups, there was an observed boost in tissue GSH content showing that both extracts tend to reverse the tissue depletion of GSH in hepatic tissues. The hydroalcoholic extracts of both plants cause a significant increase in hepatic CAT activities and, thus, diminishes the oxidative injury in the liver due to the free radical establishment by the action of TAA. Lipid peroxidation is considered as one of the important characteristics of oxidative stress 32 . Rats treated with TAA showed significantly elevated levels of lipid peroxidation, which is characterized by increase in the levels of MDA resulted in the failure of the antioxidant defense mechanism 33 . There was a decrease in the levels of MDA in plants-treated rats previously to their intoxication with TAA, showing that both extracts may exert a preventive action on hepatic tissue. In case of silymarin, this compound also exerted antioxidant activities indicated by prevention of GSH and CAT depletion besides maintain MDA at normal levels. Consistent with these data, Eldhose et al. 29 was previously reported that silymarin have a clear antioxidant activity at 100 mg/kg against TAAinduced liver injury. Furthermore, a chronic subjection of rats to TAA resulted in an increase in hepatic TNFα and IL-6 levels. This is in parallel with previous studies 26,34 . Silymarin also revealed anti-inflammatory potential where it decreased hepatic TNF-α and IL-6 contents which is in line with a previous study 35 . The anti-inflammatory action of both plants also observed as a reduction in hepatic TNF-α and IL-6 levels compared to TAA group. Here, induction of liver fibrosis via TAA increased liver α-SMA that comes in agreement with several previous studies 36,37 . α-SMA is an actin isoform and a specific marker for smooth muscle cell differentiation 38 . Silymarin, on the other, demonstrated anti-fibrotic activities in rats evidenced by significant www.nature.com/scientificreports/ hepatoprotective properties owing to their antioxidant, antiinflammatory and antifibrotic effects against TAA induced liver toxicity. ## Conclusion The result of the current work highlighted that pretreatment of P. indica or P. auriculata hydroalcoholic extracts have significant hepatoprotective effects on TAA-induced liver oxidative stress and liver fibrosis in rats through their robust antioxidant and anti-inflammatory potentials. Both doses of P. indica and P. auriculata at 200 mg/kg showed promising protection against TAA-induce liver fibrosis with superior hepatoprotective activity accounted for P. indica. The bioactive compounds present at both plants, identified by LC-MS/MS, could be responsible for these activities. However, further clinical studies are needed in order to validate the usage of these extracts as hepatoprotective supplements, to ameliorated fibrosis and thus could be used as antifibrotic agents. Institutional review board statement.

chemsum

{"title": "Phytochemical profiling and anti-fibrotic activities of Plumbago indica L. and Plumbago auriculata Lam. in thioacetamide-induced liver fibrosis in rats", "journal": "Scientific Reports - Nature"}

density_layering_during_gaseous_diffusion_in_carbon_nanotubes:_an_analytic_model

## Abstract: Recently, molecular dynamics simulations have predicted that concentric layers of gaseous carbon dioxide particles will appear in carbon nanotubes. We show in this letter how this effect can be predicted analytically by considering the potential field generated by the pore wall. The layer potential expression thus derived can be used to reproduce the essential features of a particular MD study of gaseous carbon dioxide within a (40, 40) carbon nanotube and confirm, from an energetic point of view, that an outer gaseous layer will be stable. With a closed form expression for the layer potential known, we are able to derive formulas for quantities typically of interest in a Lennard-Jones analysis, such as minimum energy, equilibrium position and the location of zero potential. ## Introduction Carbon nanotubes have been an item of significant research interest lately. These pores, typically with diameters in the nanometer range, are commonly developed within membrane sheets and other more amorphous arrangements . The nature of gaseous flow within these tubes has been vigorously investigated. Also of significant interest is the peculiar fast diffusion phenomenon that has been identified experimentally and confirmed theoretically for such nanopores . Though analytic methods have been developed in a effort to help describe gaseous flow within such tubes , computer simulations, employing molecular dynamics (MD) and Monte Carlo methods, are often the preferred method of investigation . Recently, MD simulations have shown that during gaseous diffusion of CO 2 in cylindrical carbon nanopores, certain distinct CO 2 density layers can develop inside the nanopore . These regions of higher density CO 2 appear as concentric circumferential layers within the pore. The first layer appears beyond an initial exclusion zone of sorts at the pore wall. The MD simulations of Mantzalis et al. indicate that at a gas pressure of 20 bar, and temperature of 300 K, anywhere from one to three concentric high density CO 2 layers form depending upon the radius of the nanopore. They considered carbon nanopores of chiral indices (8,8), (12,12), (16,16) and (20,20). For additional information on how the chiral indices, (n, m), are determined, the reader is referred to the review paper by Qin . Skoulidas et al. performed MD simulations with gaseous CO 2 within a carbon nanopore of indices (40,40) at 298 K and pressures from 1.0 bar to 212.7 bar. They found that at 1.0 bar a lone outer layer appeared, near the pore wall, with the bulk of the gas existing within this layer. With increasing pressures, the CO 2 molecules begin to spill out of the outer layer and eventually the entire center of the pore becomes saturated with the gas and the layering structure becoming less pronounced though the outer and an inner secondary high density band still show slightly higher gas densities than other regions in the nanopore. Though this layering phenomenon in carbon nanopores has been reported for other gases, such as O 2 , Ar, H 2 , N 2 and Br 2 , we focus here on the results for CO 2 and more specifically those reported by Skoulidas et al. . In the above mentioned constant temperature MD studies of gaseous CO 2 , within carbon nanotubes, classical Lennard-Jones (LJ) potentials were used to describe the particle-particle interaction and the particle-wall interaction. The lone two particle LJ interaction is insufficient to restrain a gaseous molecule with transnational kinetic energies at the given constant temperature. However, by using an analytic approach, where the entire energetic influence of the curved pore wall is taken into account, it is shown how the essential features in the density layering phenomenon, as predicted in the simulation studies, can be confirmed. An expression for the net potential acting on a CO 2 molecule, within the pore interior due to the pore wall, is derived here. This result indicates the presence of a circumferential region of low potential the minimum of which exceeds the mean translational kinetic energy for the gaseous particles at the temperature in question. This acts to form the first high density layer in the outermost radial region of the pore. With the character of this net potential known, we are able to derive an expression for estimating the maximum pressure that can be obtained in the first primary layer before a significant amount of particles begin to exit the region. Given a pore of sufficient size, we propose that after the filling of the outermost low potential band this layer then generates another weaker circumferential low potential band further within the pore interior thus creating a secondary layer. This process can then be repeated. The primary result arrived at here, what we term a layer potential, is dealt with numerically but a simplified approximate form is derived whereby the quantities typically of interest in a LJ analysis can be easily computed. These include the energy of the well minimum, the radial coordinate of the minimum and the point of zero potential. This work leads to simple formulas that might be used by future researchers to estimate the energetic nature and location of a boundary layer within a given nanopore. Dimensional results predicted here are consistent with those reported by Skoulidas et al. . The location for the boundaries of the outer gaseous layer in the nanopore system, given by our result, agree with those from the MD study-the curved wall acting to intensify the attractive well and thus move the repulsive region closer to the carbon wall, as compared with Figure 1: Cross section of cylindrical carbon nanopore of radius R o . We seek the net LJ potential for a particle at point P due to the sum of all infinitesimal sections of the outer wall which is taken to have a linear particle density A. a lone particle-particle interaction. Other qualitative features of this layering phenomenon, that have been predicted by MD studies, such as the layering becoming more intense as the pore size decreases, are confirmed by this work as well. ## The Potential Model We invoke the classical LJ expression for the potential energy, U , of interaction between two molecules or atoms with centers at distance d apart . That is, where and σ are the LJ parameters and k B is Boltzman's constant. Henceforth in this study, we let have units Kelvin (K), and assign = /k B . σ will have units nm. We are interested in the potential for a CO 2 molecule due to the wall within the interior of a cylindrical carbon nanopore of circular cross-section. As CO 2 is in the gaseous phase, CO 2 -CO 2 interactions will be ignored. For now, only a two dimensional polar cross-section of the pore, as depicted below in Fig. 1, will be considered. Off plane contributions to the layer potential are evaluated later. We let the linear particle density of the outer edge be given by A. Then, by Fig. 1, the distance d from the element AR o dα to the point P is given by (2) Now, using Eqs. ( 1) and ( 2), we can write the differential for the potential at point P as Due to symmetry, U is independent of θ. We select the line along θ = 0 and multiply by 2 and integrate from 0 to π to arrive at U (r, 0) = U (r): After expanding the first binomial in the denominator of each term in the previous result, and rearranging, one arrives at Though the integral in Eq. ( 5) can be resolved analytically, the complexity of the result makes it worth while to deal with it numerically. Eq. ( 5) was employed to describe results of CO 2 gas within a (40, 40) carbon nanotube as reported by Skoulidas et al. the results of which are discussed below. The LJ parameters from Reference , along with the Lorentz-Berthelot mixing rules, were used to estimate the parameters for the CO 2 -carbon interaction. The linear particle density along the outer ring is estimated by considering the armchair structure of the (40, 40) nanotube. Around the complete circumference, 160 atoms are associated with each ring as depicted in Fig. 2. Therefore, we let A = 160/(2πR o ). The final result is plotted and shown in Fig. 3. Also, shown in the figure for comparison, is the potential for the interaction of one CO 2 molecule and one carbon atom. It is important to note how the effect of the wall acts to deepen the well as compared with the case of two isolated particles. This extended well minimum is around -520 K or around 1.7 times transnational energies at 298 K, whereas the lone CO 2 -C minimum is at around -81 K. This result suggests why, on an energetic basis, a certain number of CO 2 molecules become trapped in this outer layer. The dimensions for this band agree well with the location for the first layer of high density CO 2 within the nanopore as predicted by the MD simulations of Skoulidas et al. at 1.0 bar. This situation is further visualized by the density plot shown in Fig. 4. Here the blue region indicates the low potential region for CO 2 -carbon while the red show the repulsive area. The inner area of near zero potential is given by the yellow region. It stands to reason that at low pressures, say 1.0 bar, the low potential region fills while the yellow and red regions remain vacant. However, as pressure is increased, the zero potential central region easily fills while the outer repulsive/exclusion band remains intact. The derivation for the potential, carried out above, assumes only an influence from the wall atoms in a ring bordering the plane which contains the CO 2 molecule. for improved accuracy, we now consider contributions to the potential from other wall atoms in an infinite tube. We assume the net influence on a CO 2 molecule due to all wall atoms in the tube can be approximated by considering a series of equally spaced identical rings of carbon atoms each of linear density A and with axial spacing 6 √ 2σ c , where σ C denotes the LJ parameter for carbon. In this case, d, originally given by Eq. ( 2), becomes where we have let the plane containing the point P , (see Fig. 1) be at z = 0. We now let z have units of 6 √ 2σ c , and using Eq. ( 5), the sum of the out of plane contributions to the layer potential, U n , from an infinite tube is written as: In the expression above, the Lebesgue Monotone Convergence Theorem holds , so that the order of integration and summation can be reversed and thus, Though the result is somewhat involved, the sum in the above converges and can be written in terms of special functions. With this carried out, the remaining integral was computed numerically. Therefore, now the net layer potential is the sum of U , as given by Eq. ( 5), and U n from Eq. (8). The new adjusted layer potential, U + U n , is compared with the layer potential computed by using solely Eq. ( 5) and curves of both are depicted in Fig. 5. It is seen that the additional term, U n , acts to lower the minimum by only approximately 5.0 K. Therefore, application of the more simplified form of the layer potential, from Eq. ( 5), was considered sufficient for the purposes of this study. Figure 5: Plot of the potential as given by Eq. ( 5), the black curve, and with the addition of the correction term from Eq. ( 8), the red curve, near the well minimum for the case considered in Fig. 3. ## Maximum Pressure in the Outer Layer From Fig. 5 in Skoulidas et al. , it can be seen that at low pressures the CO 2 gas resides essentially completely within the outer potential band. Then, as higher pressures are considered in the simulation, gas begins to leak from this outer band into the interior of the pore. Judging from this figure, this transition occurs between 1.0 and 10.0 bar. It would be useful to be a able to estimate the maximum pressure that the outer band can hold before leakage. The hope is that this result would be general and could then be used to predict the maximum pressure to fill only the outer most band in any similar nanopore. Here, since the volume of the pore is fixed, and conditions are isothermal, we assume that pressure is adjusted by increasing the amount of CO 2 in the pore. Since the predicted well depth exceeds thermal energies at the temperature in question, we hypothesize that CO 2 particles begin to scatter out of the outer potential band when the mean free path for particle-particle scattering reaches some critical value λ c . From elementary kinetic theory, we know that this mean free path is inversely proportional to the gas concentration so that we can write where D is the molecular diameter and N c is the critical number of particles per unit volume. Considering some volume V within the pore and using Avogadro's number N A , we can rewrite this in terms of critical moles n c as Solving Eq. ( 10) for n c and using this in the ideal gas law, we arrive at an expression for the maximum pressure, P m . in the outer band: where R is the gas constant. One might speculate that λ c would be on the order of the dimensions of the pore. Letting D = 6 √ 2σ co 2 , one finds that for λ c = 2R o , Eq. ( 11) yields approximately 15.0 bar while for λ c = 4R o , P m ≈ 7.5 bar. Therefore, the gas at the point of spilling out of the layer is beyond the rarefied range. This leads us to speculate that highly rarefied gases will remain securely within the outer layer in such a nanotube. ## Approximate Analytic Form Though Eq. ( 5) is useful for the general spatial study of the layer potential in the tube, it is convenient to have an approximate form for U (r) from which other relations could be derived. One might assume that the primary contributions to U , would occur for small angles α. In this regime, cos α ≈ 1. As before, considering U along the line of θ = 0 and letting δ 1 and δ 2 be small angles and cos α = 1, Eq. ( 5) can be separated into repulsive and attractive integrals as Evaluating the integrals we are left with To generalize the above, for any armchair carbon nanotube of index (n, n), we write A = 4n/(2πR o ) and then Eq. ( 14) becomes By letting δ 1 and δ 2 be adjustable parameters, a fit of Eq. ( 14), to results from the numerical evaluation of Eq. ( 5) for the case considered in Fig. 3, was obtained. The result of this curve fitting is shown in Fig. 6. Since U , as given by Eq. ( 14), is readily differentiated with respect to r, r at the energy minimum labeled r m , r at the point U = 0, denoted as r o and the minimum energy o , can all be determined. Differentiating Eq. ( 14), setting to zero and solving for r one obtains Setting Eq. ( 14) equal to zero and solving for r one arrives at Finally, using the above result for r m in Eq. ( 14), o is obtained as Figure 6: Plot of the potential given by Eq. ( 14) (solid curve), compared with data points from the numerical solution of Eq. ( 5) for the case considered in Fig. 3. δ 1 = 0.057 and δ 2 = 0.086. For the case considered in this study, the above lead to r m = 2.35 nm, r o = 2.39 nm and o = −535.2 K. Setting U = 0.1 o in Eq. ( 14) and solving for r, we get an estimate for the inner radius of the outer layer: r 1 = 2.12 nm. Both, the values for r o and r 1 computed here are within 1% of the values for the boundary of the inner most layer for CO 2 at 1 bar from Fig. 5 in Reference . ## Secondary Layers The report by Skoulidas et al. seems to predict that only the outermost layer will ever retain CO 2 gas while the remainder of the pore interior is essentially vacant, since when pressures are increased to somewhere between 1 and 10 bar, the particles begin to scatter to the entire region of the pore. However, from the net density level as a function of radius, it can be seen that there is some influence on the gas due to an inner secondary layer and multiple gaseous layers have been reported by other workers who performed MD simulations on carbon nanotubes . Eq. ( 5) can be used to estimate the nature of the potential well that generates a secondary layer if appropriate adjustments are made. First, we get a rough approximation for the number of CO 2 molecules in the outer layer, per particular volume, by using the ideal gas law to determine moles CO 2 in the volume V . We define V to be Eq. ( 18), along with the ideal gas law at temperature 298 K and 1.0 bar, leads to 0.372 molecules per V so that in this case we estimate the linear particle density to be A = 0.372/(2πr o ). Here, we have set the new outer radius to be r o as computed in the previous section. Now using this value for A in Eq. ( 5) setting R o = r o , and using LJ parameter values for a CO 2 -CO 2 interaction, the potential for the secondary band is computed and shown in Fig. 7. Figure 7: Plot of the numerical solution of Eq. ( 5) for the secondary layer potential for CO 2 gas within the carbon nanotube considered in Fig. 3. In this case r o = R o = 2.39 nm and the LJ parameters were set those for CO 2 -CO 2 , taken from Reference . Immediately it can be seen that the potential well for the secondary layer, at 1.0 bar, has a minimum not nearly equal to thermal translational energies at the temperature in question. Therefore, this layer potential could only act to trap the small fraction of low velocity gaseous particles at this temperature. This result is not inconsistent with the density profile predicted in Reference as the secondary layer does not seem to fill as a stand alone region but rather begins to gather particles as the entire inner region fills with increasing pressure. The radial boundaries, predicted here for this secondary layer, are consistent with the result in Reference . ## Conclusion In this report, we demonstrate how a simple analytic model can be used to obtain insight into the nature of the gaseous CO 2 density distribution within a carbon nanotube. Typically, these results must be obtained via computer modeling either by MD or Monte Carlo methods. By considering the LJ potential inside the tube, due to the atoms on the curved wall surface, an expression is arrived at that involves the tube radius and the LJ parameters. This resulting layer potential, valid for any region within the cross-section plane, ends up being only a function of the radial coordinate due to the symmetric nature of a cylindrical nanopore. This analytic result is compared with an MD study of CO 2 gas within a (40, 40) carbon nanotube. The expression for the potential describes well the nature of the outer density layer that appears upon the first insertion of gaseous particles within the tube. Both the beginning and ending radial coordinate for the outer high density layer predicted by this analytic result agree with those reported in the cited MD study. Additionally, results from this work predict that a boundary potential well, which serves to form the outer layer, has a maximum well depth of approximately 1.7 times that of the mean thermal translational energies for the gaseous particles at the temperature in question thus confirming the prediction of the MD study for there being a stable outer gaseous layer. In agreement with the statement of Lee and Sinnott, who studied a similar system, that "smaller nanotubes trap more molecules in the nanotube interiors with stronger moleculemolecule interactions..." , we find from Eq. ( 14) that as the tube radius R o → ∞ the layer potential decays to zero. Additional flexibility is gained via the analytic approach as formulas for the distance to minimum energy from the pore center r m , the point of zero potential r o and the well minimum energy o are all obtained for the outer layer. Unfortunately, these expressions involve the undetermined parameters δ 1 and δ 2 . However, we can reason as to how to best estimate these. It seems logical that the minimum total angle required to obtain the correct contribution from the repulsive term, i.e. 2δ 1 , would be smaller than that of the attractive term, 2δ 2 , as the repulsive term is influential over a shorter range than the attractive. For example, in the case studied here, around 7 o as compared with 10 o . Further, we note that from the values obtained in this work δ 2 /δ 1 ≈ 3/2. These angle values, and/or the three halves ratio, could serve as a standard guide for estimating δ 1 and δ 2 when using Eq. ( 14) to model gaseous behavior within any (n, n) carbon nanotube. However, more comparison with experimental or simulation results are required before conclusions can be established. ## Funding There are no funding sources to acknowledge for this work. The authors report no conflict of interest.

chemsum

{"title": "Density Layering During Gaseous Diffusion in Carbon Nanotubes: An Analytic Model", "journal": "ChemRxiv"}

macrophage-mediated_delivery_of_light_activated_nitric_oxide_prodrugs_with_spatial,_temporal_and_con

## Abstract: Nitric oxide (NO) holds great promise as a treatment for cancer hypoxia, if its concentration and localization can be precisely controlled. Here, we report a "Trojan Horse" strategy to provide the necessary spatial, temporal, and dosage control of such drug-delivery therapies at targeted tissues. Described is a unique package consisting of (1) a manganese-nitrosyl complex, which is a photoactivated NO-releasing moiety (photoNORM), plus Nd 3+ -doped upconverting nanoparticles (Nd-UCNPs) incorporated into (2) biodegradable polymer microparticles that are taken up by (3) bone-marrow derived murine macrophages. Both the photoNORM [Mn(NO)dpaq NO2 ]BPh 4 (dpaq NO2 ¼ 2-[N,N-bis(pyridin-2-yl-methyl)]amino-N 0 -5-nitro-quinolin-8-yl-acetamido) and the Nd-UCNPs are activated by tissue-penetrating nearinfrared (NIR) light at $800 nm. Thus, simultaneous therapeutic NO delivery and photoluminescence (PL) imaging can be achieved with a NIR diode laser source. The loaded microparticles are non-toxic to their macrophage hosts in the absence of light. The microparticle-carrying macrophages deeply penetrate into NIH-3T3/4T1 tumor spheroid models, and when the infiltrated spheroids are irradiated with NIR light, NO is released in quantifiable amounts while emission from the Nd-UCNPs provides images of microparticle location. Furthermore, varying the intensity of the NIR excitation allows photochemical control over NO release. Low doses reduce levels of hypoxia inducible factor 1 alpha (HIF-1a) in the tumor cells, while high doses are cytotoxic. The use of macrophages to carry microparticles with a NIR photo-activated theranostic payload into a tumor overcomes challenges often faced with therapeutic administration of NO and offers the potential of multiple treatment strategies with a single system. ## Introduction Nitric oxide (NO) has exhibited signifcant potential as a cancer therapy, but its effects are highly concentration-and locationdependent. 1 Furthermore, NO has a relatively short lifetime in physiological media and induces signifcant side effects when delivered systemically. 2 Targeting strategies must be a key feature for the delivery of any drug, and this is especially true for a bioregulator such as NO. Photochemical uncaging allows one to defne the location, timing and dosage of such drug delivery, thus this technique has value as an investigative tool and for addressing the progression of specifc disease states. In this context, we and others have developed photo-activated NO releasing moieties (photoNORMs), since triggering with light can provide precise spatial and temporal control of NO delivery. However, this methodology is limited by the strong wavelength dependence of light transmission through tissue. 11,12 Ultraviolet and shorter visible light are much less tissue penetrating than are longer red or (ideally) near-infrared (NIR) wavelengths. The ability of NIR light to transmit deeply into tissue has inspired various approaches to designing pho-toNORM systems including the engineering of molecular compounds that are photoactive at longer wavelengths 13,14 and the use of antennas for multi-photon sensitization of NO release with NIR light. However, delivering these conjugates to the desired physiological targets continues to be a challenge. It is difficult to deliver the desired drug payload specifcally to hypoxic areas of tumors via simple diffusion from the blood stream, owing to poorly developed vascular structures. Instead it would be particularly valuable to utilize the inherent biological mechanisms to facilitate such delivery. Tumor hypoxia generates inflammatory signals that recruit monocytes from the blood via chemotaxis, that once inside the tumor differentiate into macrophages. 22 This behavior has stimulated interest in recruiting macrophages as carriers to localize drugs in tumors at levels difficult to attain with conventional delivery methods. The present study demonstrates the utility of macrophages as carriers of micron-sized, poly(lactic-co-glycolic acid) (PLGA) microparticles in which one can incorporate a theranostic payload. The encapsulated payload here includes a photoNORM that can be triggered for NO release by NIR light together with Nd 3+ doped upconverting nanoparticles (Nd-UCNPs) to provide imaging capabilities. 29,30 The BALB/c bone marrow derived macrophages (BMMs) were shown to undergo phagocytosis of these microcarriers. Such macrophages allow far deeper penetration into large NIH-3T3/4T1 co-cultured tumor spheroids (Scheme 1) than do the microparticles alone. Thus, as previously described by Clare and coworkers, 23 such macrophages have the potential to act as "cellular Trojan Horses" to carry a therapeutic payload into tumors. It is shown here that, once carried inside the spheroid by a BMM, the photoNORM in the microparticles can be activated with NIR light to release NO. At high light intensity, NO concentrations sufficient to cause direct tumor cell cytotoxicity are generated, while at low light intensity, the NO released leads to a signifcant drop in the expression of hypoxia inducible factor HIF-1a in the tumor microenvironment. ## NIR active photoNORM and Nd-UCNPs in microparticles The cell-mediated delivery platform described here consisted of murine BMMs loaded with polymer-based microcarriers into which were incorporated a NIR sensitive photoNORM and NIR active bioimaging Nd-UCNPs. The photoNORM was prepared by metathesis of the water-soluble salt [Mn(dpaq NO 2 )(NO)]ClO 4 (dpaq NO 2 ¼ 2-[N,N-bis(pyridine-2-yl-methyl)]-amino-N 0 -5-nitroquin-olin-8-yl-acetamido) 13 with sodium tetraphenylborate to give the corresponding, hydrophobic photoNORM [Mn(dpaq NO2 )(NO)]BPh 4 (I). The hydrophobicity was needed to minimize leakage of I from the polymer microcarrier into the medium. The BPh 4 salt was further purifed by recrystallization to obtain black needles. Fig. 1 (top) shows the spectrum of I in acetonitrile. Although the l max of the longest wavelength band is $650 nm, this absorbance extends to the NIR region (3: 20.2 M 1 cm 1 at 794 nm) and overlaps with the output from a 794 nm diode laser used as a continuous wave (CW) excitation source in the present study. Direct 794 nm photolysis of I in acetonitrile solution does indeed lead to NO release as measured using the Sievers Nitric Oxide Analyzer (NOA) with a quantum yield of 0.18 (ESI Fig. S1 †). The Nd-UCNPs were high quality core/shell upconverting nanoparticles (Fig. 1, bottom) synthesized using the robotic Workstation for Automated Nanocrystal Discovery and Analysis (WANDA) of the Molecular Foundry at Lawrence Berkeley National Laboratory. The host material for the 10 nm diameter cores was NaY 0.8 Gd 0.2 F 4 . (The Gd 3+ adduct favors the hexagonal structure that typically gives higher upconversion efficiency). § The host material was also doped with the lanthanide ions Yb 3+ (30%), Nd (1.0%) and Tm (0.5%). The 2 nm thick shell was composed of NaGdF 4 doped with Nd 3+ (20%). The shell minimizes surface quenching effects and improves the luminescence efficiency of UCNPs. The Nd 3+ sensitizers in these NaYF 4 :Yb/Gd/Nd/Tm (30/20/1/0.5%)@NaGdF 4 :Nd (20%) nano-particles (Nd-UCNPs) absorb NIR wavelengths near 800 nm allowing these materials to be excited at wavelengths where the transmittance through water is most efficient. Energy absorbed by Nd 3+ ions in the shell migrates via resonant transfer to Nd 3+ in the core and energy transfer to Yb 3+ ions. 36 The energy from multiple photons is transferred sequentially from the excited Yb 3+ dopants to Tm 3+ emitters, thereby producing upconverted photoluminescence (PL) bands in the UV and visible range. The transmission electron microcopy (TEM) and PL spectra of these Nd-UCNPs are shown in Fig. 1 (bottom). The polymer-based microcarriers were prepared from PLGA dissolved in dichloromethane (DCM) by a micro-emulsion technique as described previously 37,38 and in the Experimental methods section (see below). The procedure simultaneously encapsulated the photoNORM I and/or Nd-UCNPs into spherical PLGA particles with ca. 1-micron diameter (0.35-2 mm) (ESI Fig. S2 & S3 †). The average loading of the manganese photo-NORM was 4.36 AE 0.66 wt% as determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES). The spherical micro-particles have the optimal shape to facilitate phagocytosis. 39 The PLGA was acid terminated and the resulting surface carboxylates were modifed with immunoglobulin G (IgG) by amide coupling in 0.1 M pH 5.5 2-(N-morpholino) ethanesulfonic acid (MES) buffer to increase the efficiency of phagocytosis. A bicinchoninic acid (BCA) protein assay indicated there is ca. 44 mg of IgG per mg of PLGA particles. ## NO release from photoNORM loaded polymer microparticles Previous studies in these laboratories 16,40 have shown that UCNPs can serve as photosensitizers that absorb NIR photons and emit visible light to trigger NO release from photoNORMs that were not NIR sensitive. This process requires relatively high intensity irradiation to effect the multi-photon upconversion mechanism of UCNPs and energy transfer from UCNPs to photoNORMs. In the present case, the Mn photoNORM I is photosensitive toward NO release via direct single-photon excitation with 794 nm light, and this allows one to overcome the scattering constraints that may be problematic for multiphoton excitation in deeper tissue. However, since the extinction coefficient for I at this wavelength is low (20.2 M 1 cm 1 ), it was of interest to see whether the visible light generated by upconversion from the Nd-UCNP would enhance NO release owing to the signifcantly higher extinction coefficients of I in the visible spectral region. In order to test this possibility, microparticles of two different compositions were prepared. One contained both I and Nd-UCNPs (PLGA-1), the other contained only the manganese photoNORM I (PLGA-2). Data obtained from dynamic light scattering (DLS) showed these two particle groups to be similar in average size ($1 micron, ESI Fig. S2 † top), while ICP-AES analysis showed the former to have somewhat higher loading of I (4.76 wt% vs. 3.79 wt% respectively). Particles of both types (0.5 mg) were separately suspended in 2.5 mL pH 7.4 phosphate buffered saline (PBS) solution and were irradiated with a 794 nm diode laser while stirring and purging with medical grade air. The purge gas was analyzed for NO using the NOA for 1.0 s irradiation times at different intensities (in W cm 2 ). The NOA signals recorded are shown in ESI Fig. S3 † while Fig. 2 plots the quantity of NO released from these microcarriers in response to the different excitation laser intensities. Notably, both plots appear roughly linear over this narrow range, consistent with a single photon excitation mechanism for NO release. Additionally, the efficiencies of NO release for PLGA-1 and PLGA-2 are 1.63 pmol W 1 cm 2 s 1 and 2.06 pmol W 1 cm 2 s 1 respectively. When differences in loading are taken into account, this represents a lower NO release from the Nd-UCNP loaded microparticles, although this experiment is somewhat qualitative given that these are suspensions of the microparticles. Nonetheless, under these experimental conditions, the Nd-UCNPs have value primarily for imaging purposes. ## Microparticle uptake and compatibility The goal of these experiments was to determine the amount of loaded PLGA microparticles that can undergo phagocytosis into the murine bone marrow macrophages that had been prepared as described in the Experimental section (see ESI †). It has been previously shown that rigid spheres, 1-3 microns in diameter, are readily taken up by this mechanism. 27,41,42 As noted above, said microparticles were also surface modifed with a covalently bound layer of IgG to enhance uptake (ESI Fig. S4 †). An alamarBlue® cell viability assay was used to determine whether the microparticles proved toxic to the BMM cells. Fig. 3 shows that these particles displayed no signifcant acute toxicity for concentrations as high as 100 mg mL 1 for incubation periods of 24 and 48 h with BMMs. Moderate toxicity was observed at higher concentrations. (Notably, a recent review of lanthanide-UCNP toxicity has stated that, while there are numerous reports of negligible or low toxicity, "there is a paucity of knowledge concerning primary and secondary toxicity effects on the environment and humans.") 43 For the present study, the microparticle concentration of 100 mg mL 1 was selected for further incubations. Analysis of BMMs containing NO-donor loaded particles with ICP-AES determined an uptake of 0.667 mg manganese per 1 10 6 cells. This translates into 263 mg particles or 12.1 nano-equivalents of NO as photoNORM I in 10 6 cells. ## Intracellular NO release BMMs loaded with microparticles containing both I and Nd-UCNPs (BMMp + ) were then tested to verify internal release of NO. The reporter was 4-amino-5-methylamino-2 0 ,7 0 -difluoro-fluorescein diacetate (DAF-FM), which reacts to form a fluorescent compound when exposed to intracellular NO. DAF-FM was incubated with the BMMp + macrophages and a control set of BMMs without microparticles. Prior to DAF-FM incubation, both sets were treated with L-N-nitroarginine methyl ester (L-NAME), a nitric oxide synthase inhibitor, to reduce background from biological NO production. 44 Upon 794 nm laser exposure at 13.0 W cm 2 for 90 s, the BMMp + set clearly produced a visible fluorescence response while the BMM set without particles did not (ESI Fig. S5 †). ## Microparticle effects on macrophage chemotaxis The ability of macrophages to target tissue such as tumors is governed by their ability to undergo chemotaxis towards sites of inflammation. 22 Thus, it is crucial that the macrophages retain this function after drug loading. This question was probed using a transendothelial assay (ESI Fig. S6 †). First, monolayers of bEnd.3 blood brain barrier (BBB) endothelial cells were grown on FluoroBlok™ transwell inserts inside of the wells of a 24-well plate as model endothelium barriers. Such monolayers are very tight compared to those from other endothelial cells, thereby making migration through the barrier challenging. 45 After monolayer confluency, media containing 1 10 5 BMMp + cells, media containing a comparable number of BMMs without particles, and media alone were added to the tops of separate transwells and the monocyte chemoattractant protein-1 (MCP-1) was added below. After 24 h, the inserts were stained with NucBlue® allowing for quantifcation of chemotaxis by counting the nuclei of cells that had migrated to the bottom of the insert using a fluorescence microscope (corrected for the signal from wells with a confluent monolayer to which only media had been added). This analysis showed migration of both the native BMM cells and the BMMp + cells across the endothelium layers, although under these conditions, fewer of the BMMp + cells ((10.4 AE 0.8) 10 3 ) migrated relative to the BMM control cells ((17.1 AE 2.4) 10 3 ). Thus, phagocytosis of microparticles did reduce chemotaxis as reported previously, 27 although in the latter case an even sharper drop in chemotaxis was observed. However, there are various strategies that can be used to enhance overall macrophage recruitment and therapeutic efficacy, if needed. 28,46,47 Tumor spheroid penetration In order to model a 3D tumor environment, tumor spheroids were prepared from co-cultured murine NIH/3T3 fbroblast: 4T1 breast cancer cells with a 5 : 1 cell seeding ratio similar to that used previously. 48 Incorporation of fbroblasts into tumor spheroids induces formation of a tumor stroma which enhances spheroid compactness and increases the expression of pro-inflammatory cytokines used in leukocyte chemotaxis. 49 With this seeding ratio, the hanging drop technique produced spheroids with an average diameter of 962 AE 60 mm ($1.35 10 5 cells per spheroid) after 10 days of incubation with the microparticle loaded macrophages introduced on day 7. These spheroids are considerably larger than those produced with 4T1 cells alone (ESI Fig. S7 †) as well as those NIH/3T3: 4T1 spheroids grown without the loaded BMM cells (ESI Fig. S8 †), the latter a likely result of macrophage supported tumor growth. 49 Pimonidazole (PIMO) staining of the spheroids revealed hypoxia throughout the spheroid interior. Considering that hypoxia normally occurs 100-150 mm into a tumor, 50 this is not surprising. Macrophages labeled with Celltracker™ deep red were used to monitor the chemotactic penetration of the BMMp + cells into spheroids. Images of tumor spheroid center slices indicated deep penetration (Fig. 3b), although the macrophages did not reach the center but instead formed a visible ring around the central section. Since oxygen is required for macrophage chemotaxis, this behavior may reflect a central necrotic region with high hypoxia. Flow cytometry of spheroids dissociated with Accumax™ demonstrated a macrophage cell composition of 0.53 AE 0.4% or $715 AE 540 macrophages per spheroid (ESI Fig. S9 †). Assuming that the BMMp + retained their payload at the concentration shown by ICP, these macrophages brought an estimated 8.7 AE 6.6 pico-equivalents of the photoNORM into the spheroid. The value of the macrophages as "Trojan Horses" was demonstrated using particles labeled with fluorescent IgG. Spheroid slices (Fig. 3c) demonstrated that macrophages with internalized fluorescent particles carried their payload into the spheroids (Fig. 3c) but that IgG modifed microparticles alone were unable to penetrate more than a few cell layers when incubated with the spheroids (Fig. 3d). ## Nd-UCNP imaging The Nd-UCNPs introduced to the polymer-based microparticles provide an imaging agent that allows one to track the location of these prodrug carriers and their macrophage hosts using tissuepenetrating NIR excitation wavelengths. In addition, since their chemotactic capacities allow macrophages to hone in on inflammation sites, the PL from UCNPs provides a potential mechanism to detect hidden metastatic sites. Since I is photoactivated by the same NIR wavelength, the combination of this photoNORM and Nd-UCNP in these macrophage-carried microparticles would provide theranostic capability. Fig. 4 shows 3D images of spheroids incubated with BMMp + or with BMMp and recorded with a confocal microscope using a 808 nm laser source. The spheroids were labeled with calcein AM to provide visualization of their periphery. PL signals were noted from the spheroids treated with BMMp + cells while those treated with BMMp cells failed to produce a signal. These luminescent emissions from the Nd-UCNPs provide a proof-of-principle with regard to using UCNPs to track the macrophages, but it is clear that either more efficient emitters or (given that the nonlinear relationship between UCNP PL and excitation intensities) a more intense laser source may be needed for in vivo diagnostic applications. On-going studies will address this issue. ## Photoactivated NO release inside spheroids The next question was whether NO released by NIR excitation of these BMMp + infltrated spheroids can be detected externally using the NOA, which to the best of our knowledge, would be unprecedented. Five spheroids incubated with BMMp + or BMMs with PLGA-only particles (BMMp ) were positioned in the corner of a custom designed cuvette (Fig. 5a) containing 1 mL of Hank's Balanced Salt Solution (HBSS) solution. In typical NOA analysis, the solution is entrained with the carrier gas and stirred to facilitate transfer of NO to the detector. However, such conditions would likely cause spheroid disassembly and cell lysis. Instead, the spheroids were blanketed with unstirred HBSS solution and irradiated with the 794 nm diode laser operating at 13.1 W cm 2 for specifed time periods (Fig. 5a). Subsequent gentle bubbling with medical-grade air above the spheroids released NO from the solution without disrupting the spheroids. The procedure could be repeated several times with each sample (Fig. 5b and c). The NO released photolysis from the fve BMMp + infltrated spheroids after the 6 min (total) was 13.6 pmol ($2.7 pmol per spheroid (avg)). In contrast, spheroids treated with BMMp cells gave no measurable NOA response. Based on volumes of $1 mL, NO steady state concentrations in excess of 1 mM were generated in the spheroids, a concentration exceeding that needed to induce p53 phosphorylation and/or nitrosative stress mediated apoptosis. 1 The effect of generating such high localized NO concentrations was examined by evaluating cell viability after 24 h using a p-nitrophenyl phosphate (PNPP) assay. 51 Fig. 5d illustrates the results of irradiating spheroids containing BMMp + or BMMp cells with 794 nm light for 1 to 3 six-minute periods at 13.1 W cm 2 . In general, the data in Fig. 5d consistently show reduced cell viability for irradiated spheroids infltrated with BMMp + cells relative to those loaded with BMMP cells, the most convincing example being the 26% reduction for spheroids containing photoNORM loaded macrophages after 12 min irradiation compared to 6.6% reduction for similarly treated BMMp loaded spheroids. However, the experimental uncertainties are large and barely statistically signifcant owing no doubt to unavoidable variability in spheroid loadings and the difficulty in preparing, photolyzing and analyzing a statistically large number of loaded spheroids. Notably, the spheroids used in this study were formed from 4T1 breast cancer cells, which are p53 defcient, 52 so the observed damage can largely be attributed to nitrosative stress induced apoptosis. Cancers with p53 pathways are likely to be more sensitive to NO delivery at these concentrations. Low NO concentrations have been shown to produce bene-fcial shifts in tumor microenvironment through reduction of factors such as P-glycoprotein and HIF-1a that are implicated in increased resistance to chemo-and radio-therapy. HIF-1a is a transcriptional factor upregulated in hypoxia that controls the expression of genes correlated with tumor cell survival, metastasis, and angiogenesis. 56 The effect of generating lower NO concentrations on HIF-1a levels in the tumor spheroids was examined by using LED excitation (0.58 mW cm 2 at 735 nm). After 8.5 h exposure, the spheroids were dissociated into a single cell suspension. Cells were then fxed and labeled with anti-HIF-1a and by a fluorescent secondary antibody. Analysis with flow cytometry (Fig. 6) revealed that the low intensity excitation at 735 nm reduced proportion of cells expressing HIF-1a from 96.7 AE 0.5% to 77.7 AE 8.7% in irradiated spheroids containing BMMp + macrophages while spheroids with infltrated with BMMp macrophages showed little change in HIF-1a level upon excitation. These data agree with previous studies demonstrating that low concentrations of NO destabilize HIF-1a in hypoxia due to the inhibition of cytochrome c oxidase, 54 a critical component of mitochondrial respiration. In a second trial, HIF-1a expression in spheroids containing BMMp and BMMp + macrophages was shown by flow cytometry to be 90.3 AE 5.1% and 62.2 AE 2.5%, respectively, after LED excitation for 7 h (ESI Fig. S10 †), thus confrming the impact on this proliferative factor upon delivering a low concentration of NO. ## Summary These studies demonstrate that the timing and dosage of NO release from photoNORM loaded microparticles can be controlled inside tumor models by triggering with tissue penetrating NIR light. Bone marrow derived macrophages serve as Trojan Horse carriers for the polymer-based microparticles incorporating both the photoNORM (I) and Nd 3+ -doped upconverting nanoparticles for therapeutic and imaging applications. In this manner, the carrier macrophages load large quantities of a photoactivated therapeutic agent without signifcant effects on viability. The loaded BMMp + cells maintain the chemotactic ability to traverse an in vitro brain blood barrier transendothelial model and to penetrate tumor spheroids with their photoactive cargos. Such penetration does not occur with the microparticles alone. However, while the tumor spheroids provide a valuable proof-of-principle, in vitro model to demonstrate that macrophages can infltrate and deliver a payload to such tissues, the next stage will be to demonstrate such targeting in vivo. In this context we are encouraged by a recent study 57 where blood monocytes loaded with nano-sized polymeric micelles containing paclitaxel (PTX) were used to treat metastatic breast cancer in mice. Such PTX delivery was signifcantly more efficient than using free PTX or PTX loaded nanoparticles alone. Both I and the Nd-UCNPs are excited with NIR light at $800 nm, the ideal wavelength for tissue transmission. Thus, the Nd-UCNP emission is used to visualize BMMp + cells infltrated into spheroids while NIR photoactivation of I released NO. We demonstrate two types of potentially therapeutic effects by using different light intensities to irradiate the BMMp + cell infltrated tumor spheroids. The micromolar NO concentration generated by high intensity NIR laser excitation leads to increased nitrosative stress and cell mortality. In contrast, lower concentrations of NO served to modify the tumor microenvironment by destabilizing HIF-1a. Although not explored in the current study, it should be noted that targeted NO release in the hypoxic regions of tumors should enhance the effectiveness of cancer chemotherapy 55 and radiotherapy. 58,59 In these contexts, macrophage-mediated delivery of photoNORMs combined with NIR excitation represents a fresh approach that allows one to target tumors or other diseased tissues characterized by inflammation. This strategy will facilitate the spatially and temporally controlled release of NO or of other caged compounds for precise therapeutic applications. ## Materials All cell lines (4T1, NIH-3T3, and bEnd.3) were purchased from ATCC. BALB/c mice (6-8 weeks old) were purchased from Charles River. Sodium tetraphenylborate, sodium tri-fluoroacetate, sodium oleate, ammonium fluoride, lanthanide chlorides (99.9+%), oleic acid (OA) (90%), 1-octadecene (ODE) (90%), poly(vinyl) alcohol (PVA, M w : 13 000-23 000), 4-morpholineethanesulfonic acid (MES, low moisture content, 99+%), immunoglobulin G (IgG from mouse serum), N-(3-dimethylaminopropyl)-N 0 -ethylcarbodiimide hydrochloride (EDC$HCl, commercial grade), Triton™ X-100 (BioXtra), Tween® 20 (Bio-Xtra), agarose (BioReagent), sodium hydroxide (NaOH, ACS reagent grade), sodium acetate (RegentPlus®), purifed human plasma fbronectin (1 mg mL 1 ), and Perfecta3D® 96-well hanging drop plates were purchased from Sigma-Aldrich. 10 PBS (OmniPur® liquid concentrate) was purchased from EMD Millipore. Poly(lactic-co-glycolic acid) (5050 DLG 8A, acid terminated) was purchased from Lakeshore Biomaterials. 4-Amino-5-methylamino-2 0 ,7 0 -difluorofluorescein diacetate (DAF-FM-2DA, Molecular Probes®) was purchased from Life Technologies. N-Hydroxysuccinimide (NHS), fetal bovine serum (FBS), Dulbecco's Modifed Eagle Medium high glucose (DMEM), Dulbecco's Modifed Eagle Medium: Nutrient mixture F12 GlutaMAX™ and high glucose (DMEM/F12), sodium pyruvate solution (100 mM), penicillin streptomycin 10 000 U mL 1 (P/S), trypsin/EDTA (0.25%), Hank's Balanced Salt Solution without calcium and magnesium (HBSS), Dulbecco's Phosphate Buffered Saline without calcium and magnesium (DPBS), NucBlue®, CellTracker™ deep red, goat serum, FITC labeled goat-anti-rabbit, rabbit anti-HIF-1a, p-nitrophenyl phosphate (PNPP) substrate tablets, calcein AM, micro bicinchoninic acid assay kit, and alamarBlue® was purchased from Thermo Fisher Scientifc. All TC and Non-TC treated plasticware for cell culture, and Fluoroblok™ transwell inserts were purchased from Corning. Bambanker™ cell freezing media was purchased from Bulldog Bio. Accumax™ was purchased from Innovative Cell Technologies. L-N-Nitroarginine methyl ester hydrochloride salt (L-NAME) was purchased from Caymen Chemical. Murine monocyte colony stimulating factor 1 (MCSF-1) and murine monocyte chemoattractant protein 1 (MCP-1) were purchased from Peprotech. Hypoxyprobe-Red549 Kit containing pimonidazole HCl and mouse Dylight™ 549-Mab was purchased from Hypoxyprobe. Immunoglobulin G labeled with Cyanine 3.5 (Cy3.5-IgG) was purchased from Jackson Immuno. The [Mn(dpaq NO2 )(NO)]BPh 4 salt (I) was prepared under reduced lighting by anion metathesis of the perchlorate analog that had been synthesized as reported. 13 [Mn(dpaq NO 2 )(NO)] ClO 4 , (119.3 mg, 0.195 mmol) was dissolved in 3 mL solution of 1 : 1 acetonitrile/deionized (DI) water that was then added to a 2 mL volume of acetonitrile (ACN) in which was dissolved NaBPh 4 (66.7 mg, 0.195 mmol). The resulting mixture was sonicated for 3 min after which most of the solvent was removed under reduced pressure. The resulting hydrophobic black solid was suspended in aqueous solution then collected by fltration, washed with DI water and dried under vacuum. The solid was then dissolved in dichloromethane (DCM) and recrystallized by vapor diffusion of ether to produce crystalline black needles of I. ## IgG-modied microcarriers The procedure for forming the polymer micro-emulsions was modifed from the literature. Acid-terminated PLGA (100 mg) and $80 mL of a solution of Nd-UCNPs in hexane ($10 mg UCNPs) were added into 500 mL DCM, and the mixture was sonicated for 45 min at room temperature. A 14 mg sample of I was dissolved in a mixture of ACN (150 mL) and DCM (400 mL) to form a dark purple solution that was then transferred into the PLGA solution, and the mixture was sonicated until homogenous. If the volume of as-prepared solution was lower than 1.1 mL, more DCM was added. The polymer solution was slowly added into 200 mL of 1 wt% polyvinyl alcohol (PVA) aqueous solution contained in a 250 mL half-spherical container while the ultrasonic homogenizer was turned to 350 watts for 30 s. A dark brown colloidal solution formed immediately. The flask containing the colloidal suspension was fully covered with aluminum foil and stirred overnight to evaporate the volatile organic solvents. The milky solution was centrifuged to collect a solid, which was washed with 18 megohm pure water to remove the PVA. Following particle purifcation, the brown pellet was re-suspended in 45 mL pure water and then the 1-micron particles were separated by different centrifuge speed (4000 rpm to remove particles <500 nm and 300 rpm to remove particles size >3 mm). The resulting particles were dried under vacuum and re-suspended in 0.1 M pH 5.5 MES buffer solution with the concentration of 1 mg mL 1 . EDC$HCl (70.94 mmol per 1 mg particles) and NHS (106.87 mmol per 1 mg particles) were added into the colloidal MES solution, which was then sonicated at room temperature for 30 min. Subsequently, 10 mL IgG solution (11.21 mg mL 1 ) per 1 mg microparticles was added and the mixture stirred overnight. The IgG modifed particles were collected by centrifugation and washed with 18 megohm pure water at least three times. IgG concentration was determined with a micro BCA assay according to the manufacturer's instructions. After incubation, particles were removed from solution with centrifugation to avoid light scattering during absorbance measurements. Initially, a beaker was utilized during the emulsion process, however; better yields were obtained using a half-spherical glass flash, since there are no corners in the latter and the sonic energy is distributed equally preventing settling and allowing more of the PLGA to form particles of the correct size. ## Bone marrow macrophage preparation and other cell cultures Bone marrow was harvested from 6-8 week old BALB/c mice in accordance with previously published methods. 60 Experiments were performed in compliance with all United States federal and California state regulations governing the humane care and use of laboratory animals, including the USDA Animal Welfare Act (Registration #: 93-R-0438) and the PHS Policy on Humane Care and Use of Laboratory Animals (PHS Assurance # A3865-01) and were reviewed and approved by the University of California Santa Barbara Institutional Animal Care and Use Committee as part of protocol 6-16-916. After marrow isolation, cells were then suspended in a 90% FBS/10% dimethyl sulfoxide (DMSO) and frozen for later use according to previous reports. 61 Unless otherwise specifed, all culture ware used with macrophages was non-TC treated plastic. Bone marrow macrophages were produced via slight modifcation of previously published methods. 60 Cryo-preserved bone marrow was thawed and added to 10 mL of DMEM/F-12 media supplemented with 10 mM Glu-taMAX™, 10% FBS, 1000 U mL 1 P/S, and 20 ng mL 1 MCSF-1 (media denoted as DMEMb) cells were centrifuged at 400 g for 10 min, resuspended in DMEMb at 2.3 10 6 live cells per mL. DMEMb (24 mL) and 1 mL of bone marrow were added to a flask and put in the incubator for 7-8 days to allow the bone marrow to mature into BMMs. On day 3, 25 mL of additional DMEMb was added to each flask to replenish the MCSF-1 in solution. After cells had matured into BMMs, DMEMb was removed from the flask and cells were washed with DPBS. To dislodge the cells, the flask was treated with ice cold Accumax™ (10-15 mL) for 20-30 min, and thumped once with the palm of the hand. Cells were collected, washed with DMEMb and centrifuged at 400 g for 10 min. Cells were resuspended in Bambanker™ cell cryopreservation media at 5 10 6 cells per mL. BMMs were kept at 80 C overnight and were transferred to liquid nitrogen afterwards. Confrmation of macrophage maturation was done using CD11b staining as described by Zhang et al. 60 For experiments, BMMs were thawed and added to non-TC treated dishes for 24 h. Cells were then detached with Accumax™ and counted. These cells were then replated at 62 500 BMM per cm 2 on plastic ware 18 h prior to the experiment. NIH-3T3 murine fbroblast cells, 4T1 murine breast cancer cells, and bEnd.3 murine brain endothelial cells between passage 3-15 were cultured on TC-treated plastic ware with Dulbecco's modifed eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1000 U mL 1 of penicillin/ streptomycin (P/S) and 1 mM sodium pyruvate (media mixture known as DMEMpy). Cells were passaged with 0.25% trypsin/ EDTA once they reached a confluency of >80%. ## Tumor spheroids Tumor spheroids were grown with a procedure based on previous research. 48 Confluent plates of 4T1 and NIH-3T3 cells were split and resuspended in DMEMpy and then mixed at a ratio of 1 : 5 4T1 : NIH-3T3 at a total cell concentration of 1.11 10 4 cells per mL. To prevent evaporation of the hanging drops, the liquid reservoirs on 96 well hanging drop plates from 3d Biomatrix were flled with hot 0.5% w/v agarose and allowed to cool to room temperature. A 45 mL aliquot of the cell mixture was added to the top of each well. Plates were sealed with Par-aflm® and put in the cell incubator for 7 days. Media was replenished on days 4 and 6 by removing 15 mL of media from each droplet and adding 15 mL of fresh DMEMpy. This procedure was done 2 times in a row during each media change instead of removing 30 mL of media all at once which can compromise the droplet. Spheroids were grown in droplets for 7 days. To incorporate macrophages, spheroids were transferred into 200 mL 96 well PCR plates used to allow for their easy manipulation and inversion on the rotisserie without spilling. Spheroids were washed 3 times with 100 mL of DMEMpy. DMEMpy (50 mL) was added to each spheroid after the fnal wash. A 50 mL aliquot of a solution of BMMs with blank or NOdonor loaded particles at 4 10 5 cells per mL in DMEMpy was added to each spheroid. Plates were put on a rotisserie in a cell incubator for 3 days to assure optimal interaction between macrophages and spheroids. On day 1.5, 100 mL of additional DMEMpy was added to each spheroid to replenish solution nutrients. After 3 days, spheroids were washed 3 times with DMEMpy to remove macrophages that had failed to infltrate the spheroid. Spheroids were then used for various downstream applications. ## Particle loading into BMMs Particles with or without NO-donor and UCNPs in DI water at a concentration of 5 mg mL 1 in a glass vial were sonicated for 10 min to break up particle aggregates. A volume of 340 mL of the microparticle solution was added to macrophages plated at 62 500 cells per cm 2 in a 100 mm diameter non-TC treated Petri dish in 17 mL of DMEMb to make a 100 mg mL 1 solution. Dishes were put in a culture incubator for 2 h to allow for phagocytosis. Cells were washed once with warm DPBS after which, ice cold Accumax™ (5 mL) was added and incubated at 37 C for 15-30 min. Cells were then pipetted up and down gently to release them from the plate. DMEMb (5-7 mL) was added to the cell suspension. Cells were centrifuged once at 400 g for 10 min and resuspended in 2 mL DMEMb. Cells were then centrifuged 3 times at 42 g for 5 min to remove free particles in solution. In cases where macrophage tracking was desired, 2 mM CellTracker™ deep red in serum free DMEMb was added to cells for 45 min before Accumax™ was applied. ## Quantication of particle loading and macrophage uptake Particles of a known concentration containing the NO-donor and UCNPs were digested in 1 : 3 HNO 3 : HCl for 24 h to assure all manganese was completely dissolved. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) was conducted with a Thermo iCAP 6300 and was used to measure the manganese content of each sample. Samples were compared to manganese standard. For quantifcation of particle uptake, 3 10 6 cells with particles were analyzed in the same manner. These cells were compared to cells without particles. ## Macrophage-particle compatibility Macrophages loaded with microparticles as described above with various particle incubation concentrations were plated at 62 500 cells per cm 2 in 96 well plates in 100 mL DMEMb. Cells were put in a cell incubator and left for 24 or 48 h. A mixture containing 100 mL of DMEMb and 20 mL of alamarBlue® was added to the media in each well. The plate was incubated for 3.5 h. A volume of 110 mL of each well was transferred to black bottomed well plates to improve the sensitivity of the fluorescence measurement. Solution fluorescence was measured on a TECAN M220 Infnite Pro plate reader (l ex ¼ 550 nm, l em ¼ 590 nm). Fluorescence from wells without cells was subtracted as baseline. ## Qualitative determination of NO release Human fbronectin at 150 mg mL 1 in DPBS was added to glass confocal dishes and incubated for 2 h at 37 C to improve its cell adhesion properties. Wells were washed 3 times with DPBS. Macrophages with particles were plated in the glass confocal dishes at 62 500 cells per cm 2 and incubated for 8 h in DMEMpy. Cells were incubated with L-NAME at 75 mg mL 1 in DMEMpy for 1 h to inhibit biological NO production from nitric oxide synthases. Wells were aspirated and 9.3 mM 4-amino-5methylamino-2 0 ,7 0 -difluoroflurescein (DAF-FM-2DA), a dye that detects intracellular NO release, with L-NAME in DMEMpy was added to the cells and incubated for 1 h. Wells were washed fve times with phenol red free, FBS free DMEMpy with L-NAME to remove uninternalized DAF-FM. Wells were exposed to a 794 nm laser for 90 s at 13.1 W cm 2 . NucBlue® was added to the wells to label cell nuclei. Wells were imaged 20 min later. Images were taken with an Olympus Fluoview 1000 Spectral Confocal. Care was taken to image cells as quickly as possible to prevent the release of extra NO. Samples were compared to macrophages exposed to the laser without particles. ## Quantication of macrophage chemotaxis Human fbronectin (50 mL, 150 mg mL 1 ) in DPBS was added to the top of 24-well sized FluoroBlok™ transwell inserts with 8 mm pores and incubated in a cell incubator for 1.5-2 h to improve their adhesion properties for endothelial cells. The top of the inserts was washed 3 times with 200 mL of DMEMpy. A 200 mL aliquot of bEnd.3 cells at a concentration of 1 10 5 cells per mL were added to the top of each well. Cells were incubated for 4.5 days with media changes every other day. No media was added to the bottom of the insert until the fnal day to prevent the growth of a monolayer in the bottom of the insert which has been reported previously when media is added to both sides. 62 On day 4.5, media in the top insert was replaced and 600 mL of DMEMpy was added to the bottom of the insert. Transepithelial electrical resistance (TEER) values were taken and all inserts with a value $0.33 U cm 2 , which traditionally represents a fully confluent monolayer, were accepted for the experiment. Inserts were washed 3 times with DMEMpy. 200 mL of either DMEMpy alone, or BMMs with or without particles at 5 10 5 cells per mL in DMEMpy were added to the top of the inserts and 600 mL of DMEMpy with 125 ng mL 1 MCP-1 was added to the bottom of each well to serve as a chemoattractant for the macrophages. Cells were incubated for 24 h in a cell incubator and washed once with DPBS. NucBlue® in HBSS was added to the top and bottom of each insert to visualize cells. After 20 min, images from 3 random locations from the bottom of each insert were taken with an Olympus CKX-41 inverted microscope. Cells were counted with the particle analysis tool on ImageJ. Wells that contained only a bEnd.3 monolayer were used as a baseline and were subtracted out of macrophage wells. ## Determination of macrophage/spheroid penetration Macrophages labeled with CellTracker™ deep red were allowed to penetrate spheroids as described above to allow for visualization. Spheroids were fxed in ice-cold methanol for 30 min. Spheroids were then embedded in Tissue-Tek® O. C. T. compound and sectioned with a 25 mm thickness. Spheroids were mounted with Vectashield® hard set mounting media with 4,6-diamidino-2-phenylindole (DAPI) to preserve samples and stain cell nuclei. For hypoxia detection, some spheroids were incubated with 100 mM PIMO in DMEMpy for 2 h prior to fxation with methanol (MeOH). This allowed PIMO to fully infltrate spheroids and covalently bind to cells in hypoxic regions. After sectioning, these samples were blocked with PBS containing 4% FBS, 1% goat serum, and 0.05% Tween® 20 (blocking solution) for 1 h. Hypoxyprobe Red549 (#HP7-100Kit; Hypoxyprobe) in a 1 : 200 dilution in blocking solution was added to each spheroid for 18 h at 4 C to allow the antibody to bind to the PIMO present in the sectioned spheroid. Sections were then washed three times with PBS with 0.05% Tween® 20. In other experiments to demonstrate retention of particles after spheroid penetration, particles were labeled with Cy3.5®-IgG. Because of the overlap between Hypoxyprobe Red549 and Cy3.5®, they were not used together. Images were acquired with an Olympus Fluoview 1000 spectral confocal microscope and processed with ImageJ. ## NO release measurement The NO measurement followed the modifed procedure from the literature. 16 All spheroids were carefully transferred to 1 mL HBSS solution in a modifed cuvette and closely placed at the corner in order to get full exposure with a 794 nm diode laser. Medical-grade air was purged into the cuvette through plastic tubing but not directly purged into the solution. Laser intensity 13.1 W cm 2 was applied to all spheroid-involved measurement. During the period of laser irradiation on these spheroids, the medical air purging was only above the solution. But once the irradiation was stopped, the purging tubing was manually immersed into the solution carefully in order to avoid agitating these spheroids. The flowing gas conveyed the NO released to the NOA. ## Confocal imaging of UCNPs in live tumor spheroids Spheroids were grown and incubated with macrophages with and without particles as described in a previous section. Calcein AM (10 mM) was incubated with the spheroids for 2 h in DMEMpy and to mark spheroid peripheries. Live spheroids were added to a Petri dish with HBSS. UCNPs were detected using a 3 W 100 fs pulsed laser (37.5 nJ per pulse) operating at 810 nm. Images were acquired with an Olympus Flowview 1000MPE confocal microscope with a 25 objective (numerical aperture ¼ 1.05). Images were recorded over a 508.431 mm 508.431 mm area (1024 1024 pixels) and were acquired with a scan rate of 100 ms per pixel. Bandpass flters (420-460 nm and 495-540) nm were utilized to isolate fluorescence from UCNPs and calcein AM, respectively. Images acquired every 10 mm were combined to form 3D reconstructions with ImageJ. ## Tumoricidal potential of high dose NO therapy Five spheroids containing microparticle loaded macrophages with and without NO-donor were transferred to the corner of a cuvette containing DMEMpy. Spheroids were exposed to a 794 nm laser at 13.1 W cm 2 for 6 min to release NO from particles inside macrophages in the spheroids. This procedure was done to each spheroid 1, 2, or 3 times with 2 min between each exposure. The spheroids were transferred to a 96 TCtreated well plate and incubated for 24 h. After incubation, media was removed and 100 mL of DPBS was added to each well. To measure spheroid viability, a PNPP assay was used as previously described with slight modifcations. 51 Briefly, one 10 mg PNPP substrate tablet was added to 5 mL of a solution containing 0.1 M sodium acetate and 0.1% Triton™ X-100. A 100 mL aliquot of this solution was added to the DPBS already in each well. The plate was incubated at 37 C for 3-3.5 h. NaOH (10 mL, 1 M) was added to each well after which absorbance was measured at 405 nm on a TECAN M220 Infnite Pro plate reader within 10 min of NaOH addition to prevent loss of signal. ## Tumor microenvironment modulation with low dose NO therapy Spheroids with particle loaded macrophages were transferred to a 96 well TC-treated plate with a pipette and were incubated for 12 h to allow them to adhere. Half of the spheroids were then exposed to a 735 nm LED at 0.58 mW cm 2 for 8.5 h. Spheroids were collected in groups of 6 and washed once with DPBS. Accumax™ (300 mL) was added and spheroids were incubated for 30-45 min. Spheroids were broken up with rapid pipetting to form a single cell suspension. Cells were fxed with 4% paraformaldehyde (PFA) for 10 min followed by permeabilization with ice cold methanol (MeOH) for 10 min to allow for antibody labeling. Cells were washed three times with DPBS with 1% BSA and 0.1% Tween 20. Rabbit anti-HIF-1a (1 : 100, #PA1-16601; ThermoFisher Scientifc) in 1% w/v BSA and 0.1% Tween 20 was added to the cells for 1 h at RT. Cells were washed once with DPBS with 1% BSA and 0.1% Tween 20. FITC labeled goat-anti-rabbit IgG (H + L) (1 : 250, #A27034; ThermoFisher Scientifc) was added to the cells for 30 min in DPBS with 1% BSA and 0.1% Tween 20. Cells were washed with DPBS 3 times and analyzed via flow cytometry on a FACSAria (Becton Dickinson) using 488 nm (HIF-1a) or 633 nm (CellTracker Deep Red) excitation. Cells were analyzed for % of the cells with fluorescence signal greater than background by setting the gate to exclude the signal from unstained cells (negative control). Results were analyzed with FCM Express 6 Plus. for synthetic convenience, since the only function Gd 3+ serves is to facilitate synthesis of the desired hexagonal phase of NaYF 4 (ref. 34-36). There are other methods for synthesizing hexagonal NaYF 4 without Gd (ref. 33), and these will be used in future studies.

chemsum

{"title": "Macrophage-mediated delivery of light activated nitric oxide prodrugs with spatial, temporal and concentration control", "journal": "Royal Society of Chemistry (RSC)"}

regioselective_fluoroalkylphosphorylation_of_unactivated_alkenes_via_radical-mediated_alkoxylphospha

## Abstract: A regioselective radical fluoroalkylphosphorylation of unactivated alkenes has been developed by a one-pot twosteps reaction of (bis)homoallylic alcohols, organophosphine chlorides (R2PCl), and fluoroalkyl iodides (RFI) under visible light irradiation. This protocol employs the radical rearrangement of the in situ formed alkoxyphosphane for the first time to regiospecific installing a phosphonyl group onto the inner carbon of terminal olefins in alkene difunctionalization via C-P bond formation and C-O bond homolytic cleavage. Consequently, a series of high valueadded fluoroalkylphosphorylated alkyl iodides and alcohols are easily and efficiently synthesized by subsequent iodination and hydroxylation of the generated carbon-centered radicals. ASSOCIATED CONTENT Supporting InformationExperimental procedures, spectral characterization, crystallographic data, and DFT calculations (PDF). Organophosphoryl compounds have found wide applications in various fields such as medicinal molecules, 1 functional materials, 2 and catalysis, 3 due to their picturesque physical and chemical properties. Thus, many strategies have been developed for the incorporation of phosphoryl group into organic molecules including ionic, 4 radical, 5 and transition metal-catalyzed 6 C-P bond formations. Among them, phosphoryl radical-mediated difunctionalization of olefins represents an extraordinary valuable and versatile toolbox to introduce both phosphoryl as well as other useful functional groups in the molecules simultaneously. 7 However, the regioselectivity of such difunctionalization is always reflected in the formation of the terminal-selective phosphorylation products since phosphonyl radical as a highly active attacking species is intrinsically attached to the end side of the terminal olefins (Scheme 1Aa). In contrast, the assembly of phosphoryl onto the internal side of terminal olefins is hard to obtain. One possible tactic is to use radical replacement reaction between phosphines and the alkyl radicals derived from alkenes to realize C-P bond construction, followed by the conversion of P III to P V by treating with extra oxidants 8 (Scheme 1Ab). However, common phosphines such as alkyl/aryl/alkoxy/ aryloxy phosphines cannot undergo such transformation by forming corresponding phosphoranyl radicals due to their low reactivity 9 (Scheme 1Ac). Thus, a few successful cases have had to use presynthesized highly reactive phosphines Me3X-PPh2 (X = Sn or Si) 8a-8c and Ph2P(O)-PPh2 8d-8e to enable the substitution by delivering sufficient α-scission driven forces. ## Scheme 1. Regioselective Phosphorylation of Alkenes Rearrangement reaction, as one of the most significant transformations, has been widely applied to improve synthetic efficiency and molecular complexity. 10 Recently, radicalmediated functional group rearrangement has proven to be an efficient and elegant mean for olefin difunctionalization. By combing the intermolecular radical addition onto olefin and subsequent radical cyclization and β-fragmentation, this protocol realizes olefin difunctionalization by migrating functional groups bearing π bonds from the distal position of molecules to the medial carbon of terminal olefin 11 (Scheme 1Ba). Unfortunately, this strategy is rarely used for the migration of functional groups with lone pair electrons 12 such as alkoxy phosphine. 13 Despite the intermolecular substitution between alkyl radical and alkoxyphosphine has shown to be fail (Scheme 1Ac), we envisioned its intramolecular mode may be feasible and can be used as the key link to realize the regioselective phosphorylation of olefins. We hypothesized that if a radical is added onto the alkene moiety of alkoxy phosphine adduct formed in situ from homoallylic alcohol and diphenyl phosphine chloride, and the resulting secondary alkyl radical may further cyclize onto the tethered phosphine to produce phosphoranyl radical. If so, by taking advantage of the subsequent β-scission of C-O bond and further functionalization of the formed primary carbon-centered radical, it is possible to achieve the regioselective introduction of phosphoryl at the internal side of terminal alkenes without utilizing special phosphines and extra oxidants; meanwhile, the dehydroxylative trifunctionalization of homoallylic alcohols can be also realized (Scheme 1Bb). Although such rearrangement involves the generation of unstable primary carbon radical and the homolysis of strong C-O bond, the generation of C-P σ bond and P=O π bond may compensate for this endothermic enthalpy and facilitate the reaction. Herein, we demonstrate realization of this goal in the case of regioselective fluoroalkyl phosphorylation of unactivated olefines by a one-pot multi-component reaction of commercially available (bis)homoallylic alcohols, R2PCl, and fluoroalkyl iodides under visible light irradiation (Scheme 1C). Consequently, a series of vicinal fluoroalkylphosphorylated alkyl iodides and alcohols are conveniently synthesized. As the introduction of fluorine into bioactive molecules can significantly enhance their metabolic stability, solubility, permeability, and lipophilicity, 14 incorporating fluorine or fluorine-containing groups into organic phosphonyls are exceptionally meaningful. 15 In this context, our study represents the first example for the synthesis of such compounds by the regioselective incorporation of fluoroalkyl and phosphoryl simultaneously into olefins. Our studies were commenced with a one-pot twosteps model reaction of homoallylic alcohol (a1), Ph2PCl (b1), base, and nC4F9I (c1) in DCM (dichloromethane) under argon atmosphere at room temperature, followed by the irradiation with CFL (compact fluorescence light, 36 W) as shown in Table 1. Delightedly, the desired phosphorous rearrangement took place smoothly and gave the regioselective trifunctionalized product fluoroalkylphosphorylated alkyl iodide d1 in 70% yield when triethylamine was used as base; whereas no product d1 was produced under base-free conditions (Table 1, entries 1 and 2). The yield of d1 increased to 75 % when HNEt2 was employed instead of triethylamine, (Table 1, entry 3). Other organic and inorganic bases such as nBu2NH, DABCO, DBU, TMEDA, DIPEA, NaOAc, and K3PO4 could also promote this conversion but gave d1 in unsatisfactory yields (Table 1, entries 4-10). The results of solvents investigation for CH3CN, THF, toluene, and DMF showed that no better yield was obtained (Table 1, entries 12-15). Light irradiation was essential for an efficient reaction since d1 was only formed in 25% yield without light; replacing the light source from CFL to blue LEDs (lightemitting diodes, 460 nm, 24 W) led to a slightly lower yield (Table 1, entries 16 and 17). With the optimum reaction conditions established (entry 2 in Table 1), we set about to evaluate the generality of this method, and the results are illustrated in Scheme 2. The scope of R2PCl was firstly explored (Scheme 2A). Diphenylchlorophosphines bearing a variety of substituents with different electronical properties such as MeO, Me, Cl, and F on phenyl ring at para-position participated very well in the reaction, providing the desired products d2-d5 in good yields. The structure of trifunctionalized product d4 was further confirmed by X-ray crystallographic analysis. 2-Methyl and 3,5-dimethyl substituted Ar2PCl were also performed well, giving rise to the products d6 and d7 in 40% and 56% yields, respectively. In addition, the reaction was compatible with dinaphthyl chlorophosphines, as demonstrated in the case of d8. Notably, besides Ar2PCl, dialkyl and dialkoxy substituted chlorophosphines were also good candidates for this transformation, furnishing the desired products d9-d11, albeit in slightly lower yields. Next, a variety of fluoroalkyl iodides were examined as depicted in Scheme 2B. A wide array of linear and branched perfluoroalkyl iodides with different length reacted smoothly in this tactic, giving products d12-d17 in excellent yields. Other hybrid fluoroalkyl iodides involving Cl, Br, I, ester, sulfonyl fluoride, and sulfamide were all suitable to the conversion by chemoselective cleavage of C-I bond to give the corresponding products d18-d23. After that, we examined the scope of the homoallylic alcohols as illustrated in Scheme 3B. Primary homoallylic alcohols bearing methyl, ethyl, and benzyl substituents on the terminal alkene moiety were all compatible with this way, affording the corresponding αtertiary phosphoryl products d24-d26 in excellent yields. When inner alkenes such as cyclopentenyl and cyclohexenyl A mixture of a (0.2 mmol, 1.0 equiv.), b (1.05 equiv.), and HNEt2 (1.0 equiv.) in DCM (2 mL) was stirred at rt under Ar till the substitution was complete. Then RFI (2.0 equiv.) was added and the mixture was stirred for additional 5 h under CFL irradiation (36 W) at rt under Ar. b Isolated yields. c For optimal conversion, the adduct of a and b was first isolated and then being subjected to the photoreaction conditions. d Dimethylaminopyridine (0.06 mmol, 30 mol%) was added to promote substitution reaction. e NEt3 (1.5 equiv.) was used instead of HNEt2. f H2O (1.0 equiv.) was added along with RF-I. i Without extra H2O. were merged in primary homoallylic alcohols, the reaction also converted well to provide regio-and stereospecific antifluoroalkylphosphorylation products d28 and d29 in moderate yields. The anti-configuration was confirmed by X-ray crystallographic analysis of its derivative d28' of d28. Terminal alkene incorporated secondary homoallylic alcohols with different structures could also be applied to the agreement and gave products d30-d32 in moderate yields with 4:1-3:1 diastereoselectivity. Significantly, when cyclic secondary homoallylic alcohols like cyclopent-3-en-1-ol was involved in the reaction, trifunctionalized cyclopentanes d33 and d33' were generated as the epimers in a combined yield of 46% with 3:1 stereoselectivity. The structure of major epimer d33 was determined by X-ray crystallographic analysis. Notably, tertiary homoallylic alcohols were also compatible with this regioselective fluoroalkylphosphorylation by affording products in form of aliphatic alcohols rather than alkyl iodides. Apparently, the newly introduced hydroxyl group was derived from the trace amount of water in extra dry DCM. Indeed, the addition of 1.0 equivalent water could further increase the yields of products. As a result, a variety of cyclic tertiary homoallylic alcohols with different cycloalkyl ring sizes were transformed to the fluoroalkylphosphorylated alcohols e1-e5 in good yields. The alcohol structure was also confirmed by X-ray crystallographic analysis of e2. Heteroatoms such as O-/N-atom embedded and gem-dimethyl substituted homoallylic cycloalcohols were converted to the corresponding products e6-e8 in high yields. In addition, the conversion of spiro, fused, and bridged homoallylic cycloalcohols to the desired products e9-e11 was also successful. On the other hand, tertiary alcohols of open chain could also react efficiently under our system, as demonstrated in the cases of e12 and e13. It is noteworthy that this approach worked very well for bishomoallylic alcohols too, yielding the corresponding trifunctionalized products d34-d40 and e14-e17 through a radical 6-membered ring phosphorous intermediate rearrangement. The synthetic practicality of this strategy and the versatility of products are proved as shown in Scheme 3. A gram-scale synthesis of d1 (1.43 g, 68% yield) was successfully proceeded on 3.5 mmol scale. The follow-up derivatization of d1 was performed by the nucleophilic substitution, reduction, and arylation of C-I bond, affording the corresponding derivatives f-l in good yields. ## Scheme 3. Practicality and Follow-up Transformations To account for the mechanism of this reaction, a series of control experiments were carried out as depicted in Scheme 4A. The reaction of a1 with b1 under the conditions of NHEt2 gave the adduct I-1 nearly quantitative yield (Scheme 4Aa). When I-1 was used as the substrate to react with nC4F9I without HNEt2, the reaction took place very well under visible light irradiation and gave d1 in 78% yield (Scheme 4Ab). These results not only confirm that HNEt2 serves as base to promote the reaction of homoallylic alcohols and phosphines to form adduct I-1 but also reveals that I-1 is the key intermediate to react with nC4F9I to take place phosphorous rearrangement-mediated trifuntionalization. When cyclopropyl incorporated adduct I-2 was employed to react with nC4F9I, the cyclopropyl ring-opening product d41 was obtained in 55% yield (Scheme 4Ac). This "radical clock" experiment 16 clearly indicates that a radical-mediated phosphorous process is involved in the reaction. In addition, experiment by utilizing H2 18 O clearly verifies that the newly introduced hydroxyl in products e came from water (Scheme 4Ad). However, e does not come from alkaline hydrolysis of its corresponding tertiary alkyl iodide d since almost no hydrolysis occurred when d was treated under the standard base conditions. Thus, the production of e can be attributed to the fact that tertiary radicals derived from tertiary alcohols are easily oxidized by fluoroalkyl iodides to form carbocations, which incline to react with water to produce alcohols rather than iodide anion to yield aliphatic iodides. On the basis of our experimental observation, a proposed mechanism is delineated by a representative sample in Scheme 4B. A base-promoted nucleophilic substitution of Ph2PCl b1 and homoallylic alcohol a1 occurs first to provide the adduct I-1. Then C4F9 radical II which is initiated by the irradiation of C4F9-I under visible light, 17 adds onto the alkene moiety of adduct I-1 to yield the C-centered radical intermediate III through a 9.9 kcal/mol reaction energy barrier (see the Supporting Information for the details of DFT (density functional theory) calculations). Despite it is not a main process in our reaction, the generation of C4F9 radical can also be obtained from an EDA (electron donor-acceptor) complex of the electron-rich phosphine adduct I-1 and electrondeficient C4F9I. 18 III experiences a fast radical cyclization onto the lone electron pairs of phosphine to yield phosphoranyl radical intermediate IV, which subsequently undergoes C-O bond β-fragmentation across an energy barrier of 13.4 kcal/mol to form the C-radical intermediate V. The calculated Gibbs free energy change of this radical phosphine rearrangement is only -0.1 kcal/mol, which is almost enthalpy neutral. Finally, product d1 is formed by the reaction of V and C4F9I via an iodine-atom transfer (IAT) process. When the reaction involves tertiary alcohol, as in the case of e1, the formed tertiary radical V' tends to undergo a single-electron transfer (SET) process rather than an IAT process to react with C4F9I, 18a,19 providing carbocation VI, C4F9 radical, and iodide anion. The hydroxylation of VI by water affords e1. The calculated spin density map of intermediate IV clearly indicates that the spin is delocalized on P-atom (0.68) and the tethered two phenyl rings (0.33) (Figure 1), manifesting the formation of phosphoranyl radical. EPR (electronic paramagnetic resonance) experiments also detected the signals of PBN (phenyl N-tert-butylnitrone) trapped P V -centered radical 20 IV and C4F9 radical II, respectively, which further confirmed that this reaction involves a radical-mediated phosphine rearrangement process (see the Supporting Information for the details of EPR experiments). In summary, we have successfully developed a novel, facile, and efficient approach for the regioselective fluoroalkyl phosphorylation of unactivated alkenes by a one-pot twosteps reaction of readily accessible (bis)homoallylic alcohols, organophosphine chlorides, and fluoroalkyl iodides under visible light irradiation. The protocol employs radical phosphine rearrangement as the key step to realize the unusual installing phosphoryl in the inner side of the terminal olefins by using commercially available common organic chlorophosphines without extra oxidants. This tactic not only provides remarkable opportunities for the regioselective introducing fluoroalkyl and phosphoryl simultaneously in olefins and the trifunctionalizing (bis)homoallylic alcohols, but also broadens new boundaries of radical rearrangement modes of phosphines and their synthetic application. The further exploration of such radical rearrangement for synthetic purposes are ongoing in our laboratory.

chemsum

{"title": "Regioselective Fluoroalkylphosphorylation of Unactivated Alkenes via Radical-Mediated Alkoxylphosphane Rearrangement", "journal": "ChemRxiv"}

biomimetic_inorganic_camouflage_circumvents_antibody-dependent_enhancement_of_infection

## Abstract: Pre-existing antibodies can aggravate disease during subsequent infection or vaccination via the mechanism of antibody-dependent enhancement (ADE) of infection. Herein, using dengue virus (DENV) as a model, we present a versatile surface-camouflage strategy to obtain a virus core-calcium phosphate shell hybrid by self-templated biomineralization. The shelled DENV stealthily avoids recognition by preexisting antibodies under extracellular conditions, resulting in the efficient abrogation of the ADE of infection both in vitro and in vivo. Moreover, the nanoshell can spontaneously degrade under intracellular conditions to restore the virus activity and immunogenicity due to its pH-sensitive behaviour. This work demonstrates that the biomimetic material shell can significantly improve the administration safety and potency of the DENV vaccine, which provides the promising prospect of chemically designed virus-material hybrids for immune evasion. ## Introduction The efficacy of current biomedicine is highly restricted by unwanted immune responses. In nature, to hide from immune surveillance, a few pathogenic bacteria use host serum albumin to form a stealth surface coating. 4 Inspired by such a phenomenon, nanoparticles that are coated with endogenous cell components can provide a stealth effect to overcome in vivo immune clearance without compromising the activity. Although previous attempts have verifed that the coating of instinct components onto nanoparticles could confer an in vivo immune evasion effect, biomimetic camouflage strategies remain rarely mentioned. We note that some living organisms have the ability to produce biomineral shells, which are biocompatible and endow protective functions. 8 Inspired by the unique characteristics of biominerals, we propose the incorporation of a biomimetic mineral shell onto the outside of a virus could facilitate the camouflage effect to circumvent preexisting immunity. Pre-existing antibodies can increase the severity of viral infectious diseases in humans upon secondary infection or administration, which is designated as antibody-dependent enhancement (ADE) of infection. 9 This effect is critically serious in dengue virus (DENV) infection since 390 million cases per year generate a high baseline of pre-existing anti-DENV antibodies among people worldwide, leading to the risk of fatal ADE of infection due to secondary exploration. 10 Notably, the newly approved dengue vaccine is not recommended for use in children under nine years old due to the risk of ADE. During ADE of infection, a pre-existing antibody recognizes the virus and forms a virus-antibody complex to promote the entry and replication of DENV through ligation of the antibody Fc portion to Fcg receptors on monocytes. 14,15 Theoretically, viral epitopes that bind with pre-existing antibodies can be subtracted or blocked to abrogate the ADE of infection. 16 However, the modifcation of virus surfaces without compromising the native activity of the virus still remains a great challenge. 17 Chemical camouflage is advantageous for virus camouflage owing to its flexibility and low-cost. A key challenge of forming an optimal camouflage is that the coating should be unrecognized by antibodies and be switchable: evading the undesired antibodies under extracellular conditions, meanwhile precisely degrading to ensure the original bioactivity under intracellular conditions. It is noteworthy that abundant endogenous calcium phosphate (CaP) phases are naturally formed in human intestines as a biological selfcomponent; they can avoid body clearance and chaperone antigens to intestinal immune cells. 18 Moreover, negatively charged amino residues on viral surfaces beneft the in situ biomineralization, and the formed CaP shell could afford pH sensitive biodegradation under endosomal pH conditions. Thus, the incorporation of such bio-originated biominerals to viral surfaces could meet the needs for the proposed immune camouflage. In the present study, by using the self-templated biomineralization of DENV, we report that viral particles can be contained within a biodegradable CaP shell, and the resulting DENV-CaP core-shell hybrids can circumvent the ADE of infection as well as maintain the original immunogenicity of DENV. ## DENV-templated biomineralization The approach for DENV-directed mineralization was feasibly achieved by adding CaCl 2 to sodium phosphate monobasic containing Dulbecco's modifed Eagle medium (DMEM) supplemented with DENV particles (Fig. 1A). The abundant glutamic acid (green) and aspartic acid (red) displayed on the outside of the DENV nanoparticles aid in concentrating calcium ions and triggering the in situ nucleation (Fig. 1B). 22 Zeta potential measurement showed that native DENV had a surface charge of 15.6 mV at pH 7.4 due to its anionic carboxylate groups. At an early stage of mineral deposition, we found the electron dense nanoclusters spontaneously assembled outside the viral particle, indicating the presentation of acidic amino acids promoting the mineral nucleation on the viral surface, as evidenced by non-stained transmission electron microscopy (TEM, Fig. 1C). The nano hybrid was identifed by the coexistence of the CaP mineral phase and virus by energy dispersive X-ray spectroscopy (EDS) and Fourier transform infrared spectroscopy (FT-IR, Fig. S2B, † 1J and K). These results indicated that viral surfaces can efficiently induce a heterogeneous CaP deposition. Such DENV-templated mineralization was consistent with the previous understanding that the negatively charged biomolecules can induce biomineral nucleation. 23,24 The prolonged deposition of CaP on the viral surface further produced nanoparticles with a greater electron density (shelled DENV) than the bare ones (Fig. 1D and F). To ensure the presence of the inner viral core, the hybrid was negatively stained. The result depicted hollow spheres contained a white core with an average diameter of around 60 nm, this corresponded to the native virus particle (Fig. 1F and H), because the viral particle has a low electron density and cannot be stained. In contrast, the staining treatment of CaP nanoparticles exhibited a solid sphere structure with an increased electron density compared with that of the phosphotungstic acid background, indicating the enhanced electron density of CaP in the staining TEM (Fig. 1G). In Fig. 1H, the area surrounding the white core is darker than the background, representing a CaP shell. The negative staining TEM revealed a distinct core-shell structure as compared with the pure CaP control. To further confrm the formation of a CaP shell on a single nanoparticle, we used scanning transmission electron microscopy (STEM) to investigate the coexistence of Ca and P elements on shelled DENV (Fig. S3 †). The energy dispersive X-ray spectroscopy (EDS) attached to the STEM confrmed that Ca and P are distributed on the exterior of the shelled DENV (Fig. 1E), indicating that the shelled DENV was similar to endogenous CaP both in phase and size. The virus enclosed in the complex was confrmed by the infectivity of the shelled DENV (Fig. 1I). This material-based modifcation strategy could be generally applied to all four serotypes of DENV with high efficiency (Fig. S1 and S2 †). ## Reversible bioactivity of camouaged DENV In order to investigate whether the biomimetic CaP shell is biodegradable, to release the enclosed virus, and whether the released virus remains active, we examined the capability of the shelled virus to translate and replicate the envelope (E) protein. The indirect immunofluorescence-staining assay (IFA) showed that the viral E protein was expressed efficiently in BHK21 cells after incubating with shelled DENV (Fig. 2A and S4 †). Furthermore, the plaque formation assay revealed that the shelled DENV (81 plaques) is a little more infectious than the native DENV (33 plaques) (Fig. 2B and S4 †). Such phenomena could be explained by the well-established understanding about the receptor-independent cellular uptake pathway of CaP nanoparticles and the pH-switchable degradability of CaP within intracellular microenvironments, which enhances the delivery of virion to BHK-21 cells, resulting in the increasing number of plaques formed. 25 Additionally, the CaP nanoshell exhibited no extra cytotoxic effect on the cells compared with the native DENV due to its similarity to endogenous materials (Fig. S5 †). ## Camouaged DENV circumvents antibody binding Under extracellular conditions, we evaluated the camouflage effect of the CaP shell to inhibit the interaction between the anti-DENV antibodies and DENV. The binding affinities between the shelled DENV and different kinds of antibodies were examined by indirect enzyme-linked immunosorbent assays (ELISAs). Expectedly, all serotypes of bare DENV showed high binding affinities towards the anti-flavivirus monoclonal antibodies, including 4G2 and 2A10G6 26 (Fig. 3B, C and S6 †). In contrast, these antibodies were insufficient to recognize the shelled DENV as the binding affinities were substantially reduced (Fig. 3B, C and S6 †). In addition, the cross-reactive polyclonal human anti-sera were also incapable of recognizing shelled DENV using a native dot-blot assay (Fig. S7 †). As a consequence of insufficient formation of the virus-antibody complex, the CaP shell signifcantly ablated the ligation of DENV to the Fcg receptor (FcgR) on K562 cells (Fig. S8 †). The CaP shell protected the virus from the antibody recognition and receptor ligation (Fig. 3A), showing its camouflage effect under extracellular conditions. ## Camouaged DENV abrogates ADE infection in vitro The camouflage effect of the resulting CaP shell can eliminate the ADE of infection (Fig. 4A). DENV or shelled DENV was opsonized with a serial concentration of a cross-reactive 4G2 antibody to form the virus-antibody complex, which was then allowed to infect K562 cells bearing FcgR. As a result, the intracellular viral load in groups of the native DENV was obviously enhanced by the pre-existing antibody, whilst the shelled DENV completely ablated the enhancement of viral load (Fig. 4B and S9 †). The CaP shell abrogated the augment of the intracellular DENV at all tested concentrations of antibody due to its camouflage effect against pre-existing antibodies (Fig. 4B and S9 †). This result was further confrmed by evaluating the extracellular virion production under the same conditions. The sub-neutralizing pre-existing antibody induced a 30-300 fold enhancement of the DENV production, whereas the shelled DENV completely abolished the antibody-enhanced virion proliferation (Fig. 4C and S9 †). This phenomenon indicated that the superfcial CaP directly inhibits the formation of the virus-antibody complex and consequently reduces its ligation to K562, resulting in the circumvention of the ADE of infection (Fig. 4A). ## Camouaged DENV abrogates ADE of infection in vivo We next investigated whether the CaP shell benefts the abrogation of ADE in vivo by using 1 day-old suckling mice. The preexisting antibody in mice was obtained through the passive transfer of cross-reactive 4G2 monoclonal antibodies, which have been proved to be ADE inductive in mice. 27 Twenty-four hours after intraperitoneal injection (i.p.) with 4G2 or PBS, the mice were challenged intracranially (i.c.) with the shelled DENV or native DENV. The pre-existing antibody signifcantly enhanced the disease severity by decreasing the average survival time (AST) from 13 days to 10 days as compared with the control group, representing an enhancement of neurological infection (Fig. 5A, S10 and Table S1 †). In contrast, in the shelled DENV group, the mice in control group and ADE group both showed identical AST, verifying that the pre-existing antibody failed to enhance the infection of shelled DENV (Fig. 5B, S10 and Table S1 †). Additionally, the shelled DENV recipient induced no extra incidence of death as evidenced by the survival rate, ensuring the safety of the CaP nanoshell (Table S1 †). These fndings demonstrated the feasibility of biomimetic camouflage to evade ADE in vivo. In our experiment, the ADE of infection manifested enhanced neurovirulence and reduced the average survival time, which is in accordance with the previous report. 28 However, CaP camouflage blocked the antibody binding and thereby reduced the FcgR-mediated infection in mice brains, resulting in the elimination of early death. Similar strategies from previous observations suggested that the interference with the interplay between DENV and the antibody was capable of abrogating in vivo ADE. 27 Camouflaged vaccine ablates ADE of infection and preserves immunogenicity ADE of infection has been regarded as one of the biggest challenges in DENV vaccine development. We postulate that such a switchable camouflage would be promising for DENV vaccine engineering. As is well known, ADE occurs when there is coincidence of sub-neutralizing antibodies and DENV, in spite of a virulent DENV or an attenuated DENV vaccine. 10,29 Thus, the administration of a live attenuated DENV vaccine to individuals, who have a waning antibody response due to a previous DENV infection or vaccination, may place the vaccine recipients at risk of enhanced infection. 28 To improve the safety of the DENV vaccine, recommendations have been made to restrict the use of the vaccine to those who are not likely to have had prior exposure to DENV. 30 We hereby examine the possibility of CaP camouflage to evade the ADE of ChinDENV2 infection, a wellcharacterized live-attenuated chimeric dengue vaccine candidate. 31 As shown, the presence of anti-sera recovered from DENV-infected rhesus monkeys promoted the ADE of infection of the vaccine, even though the vaccine was highly attenuated (Fig. 6B). This result highlighted a potential safety issue with the DENV vaccine. In contrast, the CaP camouflaged vaccine restricted the recognition of antibodies, resulting in the efficient evasion of the ADE of infection (Fig. 6A and B). Furthermore, to examine the influence of the CaP shell on vaccine efficacy, the mice were injected subcutaneously with the shelled vaccine. As expected, the administration of the shelled vaccine demonstrated enhanced DENV2-specifc IgG responses (Fig. 6C) and DENV2-specifc neutralizing activities (Fig. 6D). The cytokine-secreting cellular immune responses were also increased compared with those of the native vaccine (Fig. 6E). Moreover, sera collected from vaccine and shelled-vaccine administrated mice did not contribute to ADE induction (Fig. S11 †). These fndings revealed that the stealth CaP shell not only avoids the ADE of infection but also enhances the efficacy of the vaccine. To test if the CaP nanoshell can influence the adaptive immune responses of the vaccine, we evaluated the ability of CaP nanoparticles to stimulate the cytokine secretion of antigen presenting cells. The RAW264.7 cell line, a macrophage derived cell belonging to antigen presenting cells, is incubated with PBS or pure CaP nanoparticles for 1 hour. After stimulation, CaP enhances the secretion of cytokine TNF-a as compared with the control, indicating that CaP can enhance the activation of immune cells. Therefore, the CaP shell could interact with the immune cells and stimulate the immune process (Fig. S12 †). Together, these results verifed the improved safety and potency for vaccine administration. The balance between the induction of protective immunity and the induction of ADE should be of serious consideration in DENV vaccine development. The attenuated vaccine is the only optimal strategy that could be effective and relatively safe. On one hand, the progeny attenuated vaccine would not have the opportunity to encounter antibodies induced by themselves, because the live attenuated vaccine only replicates in limited cycles in cells and is rapidly cleared by the immune system. 31 On the other hand, when vaccinated people are subsequently exposed to epidemic DENV, the chance of ADE induction is dependent on the efficacy of vaccine. Either neutralization or enhanced virus infection is determined by the threshold concentration of dengue antibodies. When the antibody concentration remains above the neutralization threshold, it will lead to a strong neutralization capability but not ADE infection (Fig. S9B †). Previous reports have demonstrated that the antibodies with strong and long-lasting neutralization capability can minimize the risk of the ADE of infection. 32 However, scientists are still facing hurdles to developing effective vaccines that can safely protect populations due to the poor understanding of DENV immunity. 33 Even the most successful vaccine, Dengvaxia (Sanof Pasteur), still elicits an incomplete immune response to DENV that contributes to the immune enhancement of disease in certain populations. The question of whether the dengue vaccine could be safely administered to pre-infected individuals remains unresolved. The development of methods to mask epitopes that may evade "unwanted" interactions is proposed in the development of a respiratory syncytial virus (RSV) vaccine, which also faces the problem of ADE of infection. 34 But the challenge is to apply the masking strategy to solve the dengue vaccine conundrum. We for the frst time propose the use of chemistry based engineering to evade the "unwanted" interactions, by introducing bioinspired and biodegradeable CaP nano camouflage on virus surfaces. The tunable characteristics of CaP nanoshells can cover the virus epitope and preserve its activity and immunogenicity. The camouflage strategy is a proof-of-concept that improves the safety of vaccinations in individuals with preexisting or waning antibodies. ## Virus-templated biomineralization Four serotypes of DENV nanoparticles and ChinDENV2 (diluted to 10 6 PFU per mL) in 500 mL of serum-free DMEM buffer were incubated with 10 mM calcium chloride (CaCl 2 ) at 37 C for 2 h to induce the site-specifc nucleation of calcium phosphate (CaP) on the viral surface. The complex designated as shelled DENV was collected by centrifugation at 8000 g. ## Physiochemical characterization The morphology of shelled DENV nanoparticles was investigated using TEM (JEM-1400, JEOL, Tokyo, Japan) with or without staining. Briefly, the shelled DENV was drop-cast from dispersions onto carbon-coated TEM grids and the solvent was blown away to keep it dry. For negative staining, native virus and shelled DENV were cast onto a carbon-coated TEM grid and stained with phosphotungstic acid before observation. Energydispersive X-ray spectroscopy (EDS) analysis of shelled DENV and the CaP control were conducted using TEM. ## Biological characterization Expression of the viral envelope (E) protein was explored by indirect immunofluorescence assays (IFAs). The confluent BHK-21 cells grown on a glass slide were infected with viruses at a multiplicity of infection (MOI) of 0.01. At 72 h post infection, cells seeded on the coverslip were fxed with an ice-cold acetone/ methanol mixture with a ratio of 7 to 3. After drying at room temperature, fxed cells were incubated with cross-reactive monoclonal antibody 2A10G6 to detect the DENV E protein. Cells were washed three times with phosphate-buffered saline (PBS) and then incubated with secondary antibodies conjugated to Alexa Fluor 488 (Invitrogen) in PBS for 30 min at 37 C. Fluorescent cells were examined using a fluorescence microscope (Olympus). For the plaque formation assay, BHK-21 cells were used to determine infectivity and viral titer. The samples were incubated with confluent cells in the 12-well plate. The virus containing suspensions were absorbed for 1 h and the cell monolayer was overlaid with DMEM containing 2% FBS and low melting temperature agarose. The cells were incubated for 3 days and overlaid with 4% formalin and crystal violet. The plaques were counted to calculate the number of PFU per mL. ## In vitro ADE assay The ADE assay performed using K562 cells. Briefly, for the detection of antibody dependent enhancement of DENV infection, DENV or shelled DENV was frst pre-incubated with 10-fold serial dilutions of 4G2 under concentrations ranging from 50 to 0.005 mg at 37 C in 5% CO 2 , allowing the formation of the immune complex. The obtained complex was placed into separated tubes further inoculated with 2 10 5 K562 cells at a MOI of 1 for 2 h. Then the inoculated cells were washed 3 times with serum free DMEM to remove excess and unbound immune complexes. The K562 cells were subsequently resuspended with 2% maintenance DMEM and cultured for an additional 72 h. We next used a flow cytometry assay to detect the intracellular expression of the DENV E protein. The K562 cells were washed once with PBS and subsequently fxed in 2% paraformaldehyde for 15 minutes at room temperature. Cells were then rinsed with a PBS buffer containing 1% FBS to remove residual fx solution. For staining, cells were permeabilized with 0.1% saponin (Sigma Aldrich) and stained with an Alexa-488conjugated (Invitrogen) 4G2 that probed the DENV E protein. The cells were washed twice, and the number of E-protein positive cells was monitored by a flow cytometry instrument (BD Biosciences). The results are expressed as percentage of infected cells versus antibody concentration. In a viral yields assay, the inoculated suspension from K562 cells in the ADE assay were collected and then used to conduct a plaque assay on BHK-21 cells. For the indirect enzyme-linked immunosorbent assays (see ESI †), the DENV-positive human sera were obtained and provided by Guangzhou 8th People's Hospital. The use of DENV-positive sera was approved by the Ethics Review Committee of Guangzhou 8th People's Hospital (GPH). ## In vivo ADE assessments All experimental procedures involving animal testing were performed in compliance with the guidelines and protocols approved by the Institutional Animal Care and Use Committee at Beijing Institute of Microbiology and Epidemiology. The in vivo ADE model was established using sucking mice. The enhancement of neurovirulence was observed by the measurement of survival time in the absence or in the presence of 4G2. Groups of 1 day-old sucking mice were injected intraperitoneally with 4G2 or PBS in a dosage of 1 mg per mouse one day before infection. Then, 24 hours after passive injection of antibody, DENV3 was administrated with 20 PFU via an intracranial route (i.c.). Animals found in a moribund condition were euthanized and scored as dead within 3 weeks of monitoring. Average survival times (AST) were calculated for animals that succumbed to infection. ## Immunization with shelled vaccines Immunogenicity was assessed by subcutaneous (s.c.) inoculation with 10 5 PFU of the ChinDENV2 vaccine or shelled Chin-DENV2 into 4 week-old female BALB/c mice. The mice were bled by tail veins after 2 weeks post-immunization. The sera were isolated for the determination of IgG antibodies by an ELISA assay. The neutralizing antibodies against a wild type strain of DENV2 were detected by a 50% plaque-reduction neutralization test (PRNT50). Splenocytes were isolated from the immunized mice at 2 weeks post-immunization and subjected to enzymelinked immunospot assay. ## Conclusions We have developed a biomimetic strategy to modify DENV nanoparticles with a CaP camouflage, which is benefcial as it precludes the recognition of different types of pre-existing antibodies, verifying the stealth effect of CaP nanoshells. Both in vitro and in vivo results have demonstrated that the switchable nature of the CaP shell abrogates the ADE of infection under extracellular conditions and ensures the intracellular immunogenicity, presenting improved administration safety and efficacy. Our results indicate biomimetic, biomineral engineered viruses would be benefcial to reduce the risk of enhanced infection in recipients with a prevalence of subneutralizing anti-DENV antibodies, providing a translational camouflage tactic to confer immune evasion and enhance the vaccine potency by using chemical materials.

chemsum

{"title": "Biomimetic inorganic camouflage circumvents antibody-dependent enhancement of infection", "journal": "Royal Society of Chemistry (RSC)"}

dsi-rna_knockdown_of_genes_regulated_by_foxo_reduces_glycogen_and_lipid_accumulations_in_diapausing_

## Abstract: Culex pipiens is a major carrier of the West Nile Virus, the leading cause of mosquito-borne disease in the continental United States. Cx. pipiens survive overwinter through diapause which is an important survival strategy that is under the control of insulin signaling and Foxo by regulating energy metabolism. Three homologous candidate genes, glycogen synthase (glys), atp-binding cassette transporter (atp), and low-density lipoprotein receptor chaperone (ldlr), that are under the regulation of Foxo transcription factor were identified in Cx. pipiens. To validate the gene functions, each candidate gene was silenced by injecting the target dsi-RNA to female Cx. pipiens during the early phase of diapause. The dsi-RNA injected diapause-destined female post-adult eclosion were fed for 7 days with 10% glucose containing 1% d-[ 13 C 6 ]glucose. The effects of dsi-RNA knockdown on glucose metabolism in intact mosquitoes were monitored using 13 C solid-state NMR and ATR-FTIR. Our finding shows that the dsi-RNA knockdown of all three candidate genes suppressed glycogen and lipid biosyntheses resulting in inhibition of long-term carbon energy storage in diapausing females. Abbreviations ATR-FTIR Attenuated total reflection Fourier-transform infrared spectroscopy CPMAS Cross-polarization magic-angle spinning Glc Glucose dsi-RNA Synthetic dicer-substrate siRNA Foxo Forkhead of transcription factors GlcNAc N-Acetylglucosamine NMR Nuclear magnetic resonance TAG TriacylglycerideCulex pipiens is the mosquito that vectors the West Nile Virus, St Louis and equine encephalitis viruses, bird malaria, and dog heartworms in North America. In late summer and early winter, this mosquito, like some other insects, initiates a diapause program in response to short photoperiods and decreasing temperatures 1 The diapausing females do not seek blood but instead seek flower or fruit nectars, primarily consisting of glucose, fructose, and sucrose 2 , to convert mono and disaccharides into glycogen and fat. This metabolic shift doubles its fat and glycogen reserves, which are used as an energy source for survival during the long winter season. During the first month of diapause glycogen reserve is consumed only as an energy source, and after glycogen is used up the energy demands are fueled from fat in the fat body. Thus, fat and glycogen storage in early diapause and regulation of their timely consumption are critical features for successful overwintering. Insulin/Foxo signaling is a key component of the cascade regulating early diapause energetics in Cx. pipiens 3 . The insulin signal is not activated in the early diapause program and subsequently lifting suppression of a forkhead of transcriptional factor (Foxo). The inactivation of the Foxo results in the reduction of nutrient storage, which suggests that the downstream genes targeted by the Foxo are crucial for glycogen and fat homeostasis in diapausing females 3 . To understand the key genetic components that are involved in the energy storage, ChIP-seq analysis was used to identify three target genes under the control of Foxo, which are relevant to fat and glycogen transport and utilization 4 . Three candidate genes are glys, atp, and ldlr. The glys gene encodes for glycogen synthase which is a key enzyme in glycogenesis that catalyzes the conversion of monosaccharide glucose into a polymeric chain of glycogen for storage. In Cx. pipiens, glycogen is rapidly accumulated during the early diapause (7-10 days after adult eclosion) and is used to maintain energy homeostasis during the first month of diapause 5 . Although the mechanism that governs differential carbohydrate utilization is unknown, we speculate that the regulation of glys is closely involved in glucose metabolism during the early nectar feed. In addition, two genes atp and ldlr have potential roles in intracellular lipid transport processes and cellular lipid homeostasis 6,7 . These two genes may be linked to the transport of various lipids across membranes to store and distribute nutrients in the diapausing fat body cells. To validate the gene functions, synthetic dicer-substrate siRNAs (dsi-RNA) were used to silence the candidate genes by injecting the target dsi-RNAs to female Cx. pipiens during the early phase of diapause. In this study, the effects of three candidate genes knockdown on glucose metabolism were characterized using a combined 13 C solid-state nuclear magnetic resonance (NMR) and attenuated total reflection Fouriertransform infrared spectroscopy (ATR-FTIR) methods. Solid-state NMR is a powerful technique that enables a direct measure of chemical compositions of intact whole cells and entire organisms 11,12 . Solid-state NMR measurements were carried out on the lyophilized dsi-RNA injected diapause-destined females of Cx. pipiens that were fed for 7 days with uniformly 13 C-labeled d-[ 13 C 6 ]glucose (Fig. 1a). Since only the 13 C isotope is NMR active, solid-state NMR was used to measure the newly synthesized 13 C-labeled glycogen and lipids that were metabolized from the provisioned d-[ 13 C 6 ]glucose. Then, ATR-FTIR was used to measure the total glycogen and lipid accumulations in mosquitoes (Fig. 1b). ATR-FTIR is a highly sensitive technique that can rapidly quantify protein, lipid, and glycogen compositions of individual mosquitoes (Fig. 1c). Recently, ATR-FTIR has been developed as a portable instrument for the identification and screening of individual Aedes aegypti mosquitoes with Wolbachia infection 13 . Unlike 13 C solid-state NMR which measures the d-[ 13 C 6 ]glucose utilization, FTIR is independent of the 13 C-labeling. Thus, FTIR was used to determine the total glycogen and lipid compositions in mosquitoes. By combining FTIR and solid-state NMR measurements, the effects of the candidate gene knockdown on the glucose utilization for the newly synthesized 13 C-labeled glycogen and lipids, and the changes in the total accumulations of glycogen and lipids in dsi-RNA injected mosquitoes were determined. ## Results and discussion Dsi-RNA injection into adult female Cx. pipiens and dsi-RNA efficiency by qRT-PCR. mRNA expression levels of the genes glys, atp, and ldlr in nondiapausing (ND) and diapausing (D) Cx. pipiens were obtained using quantitative real-time PCR (qRT-PCR). The mRNA expression patterns of these genes showed that all three genes were more than twofold upregulated in diapausing mosquitoes 7-10 days after eclosion compared to nondiapausing mosquitoes reared at short day-length (Fig. 1d). Increased expression of the glys, atp, and ldlr genes in mosquitoes in response to short day-length (destined to diapause) strongly suggested that these genes may be involved in the nutrient storage for overwintering. The dsi-RNA efficiency was assessed by qRT-PCR. Compared to the relatively high induction of glys, atp and ldlr in dsi-control injected mosquitoes, less than 30% of glys, atp, and ldlr mRNA were detected in dsi-RNAinjected diapausing mosquitoes (Fig. 1e, P < 0.05, t test). This confirmed that the injection of dsi-glys, dsi-atp, and dsi-ldlr successfully inhibited the induction of the glys, atp, and ldlr genes, respectively. The absence of changes in the basal expression of ribosomal protein large subunit 19 in mosquitoes after 7 days post-injection indicated that the observed low levels of glys, atp, and ldlr genes were related to the knock-down effect of dsi-RNA treatments rather than variation in sample loading. 13 C-labeled glycogen by solid-state NMR. The effects of glys, atp, and ldlr knockdown on the uptake and utilization of d-[ 13 C 6 ]glucose for glycogen and lipid biosyntheses were determined using solid-state NMR (Fig. 1a). 125-MHz 13 C-CPMAS spectra of dsi-RNA-injected mosquitoes are shown in Fig. 2(b-e, bottom black). Each spectrum is shown with overlapping 13 C-natural abundance spectrum (red) of diapause-destined females without the dsi-RNA injection that was fed for 7 days with 10% sucrose. The carbon chemical shift assignments for the observed resonances are as follow: 175 ppm for peptidyl-carbonyl carbons in proteins, 130 ppm for aromatic and ethylene carbons in nucleic acids and lipid head groups, 60-105 ppm for the O-alkyl carbons in carbohydrates, and 10-40 ppm for the aliphatic carbons of lipids. The spectra were normalized to the 175-ppm intensity for comparison. ## Direct quantification of All 13 C-CPMAS spectra of d-[ 13 C 6 ]glucose-fed diapausing mosquitoes show intense O-alkyl carbon resonances at 61, 73, 82, 93, 99, and 104 ppm (Fig. 2). The 61-ppm peak is assigned to the C6 of glucose, 73 ppm to the C2, C3, and C5 carbons, and 84 ppm to C4 of glucose. The anomeric carbon at the C1 position has chemical shifts in the range of 90-110 ppm depending on the glycosidic linkages. The CPMAS spectra are consistent with the routing of d-[ 13 C 6 ]glucose in diapausing mosquitoes for glycogen biosynthesis, but not to chitin biosynthesis. In the event of chitin biosynthesis, the d-[ 13 C 6 ]glucose needs to be converted to N-acetylglucosamine (GlcNAc). This results in the change of C2 chemical shift at 73 ppm in glucose to 55 ppm in GlcNAc 14 . The absence of an increase in 55-ppm peak intensity (Fig. 2b) strongly supports that the uptake of glucose during diapause is exclusively routed to glycogen biosynthesis 15 . Thus, the changes in the 73-ppm and 103-ppm peak intensities are directly proportional to the amount of 13 C-labeled glycogen accumulations in mosquitoes (Fig. 3a). The largest glycogen accumulation was observed for the control mosquitoes injected with β-gal dsi-RNA, referred to as dsicontrol, followed by the dsi-ldlr, dsi-glys, and dsi-atp injected mosquitoes. To accurately quantify the 13 C-labeled glycogen accumulation in mosquitoes, the 13 C-natural abundance contribution was removed by spectral subtraction of the natural abundance 13 C-CPMAS spectrum of dsi-RNA treated mosquitoes. In the difference spectrum (Fig. 2b-e, top) intense O-alkyl carbon resonances of glycogen are visible. The maximum glycogen accumulation was observed for the dsi-control, followed by the dsi-ldlr with 31% reduction, dsi-glys with 59% reduction, and dsi-atp injected mosquitoes with 86% reduction with respect to the dsi-control (Fig. 2c-e). Since the dsi-RNA knockdown of the genes glys, atp, and ldlr have all negatively impacted the d-[ 13 C 6 ]glucose utilization for the biosynthesis of glycogen, this strongly indicates that these genes are directly involved in the long-term carbon storage for the energy homeostasis in diapause programmed female mosquitoes. Figure 1. 13 C-Isotope labeling of Culex pipiens for solid-state NMR and ATR-FTIR analysis. (a) Nondiapausing female Cx. pipiens within 1 day after eclosion were injected (0.5 µg/♀) with dsi-glys or with dsi-control into the thorax of cold-anesthetized mosquitoes. Then the mosquitoes were fed with 10% glucose containing 1% 13 C-isotope labeled D-[ 13 C 6 ]glucose (uniformly 13 C labeled with 99% isotopic enrichment). After 7 days of feeding, the mosquitoes were frozen then lyophilized. (b) The lyophilized 13 C-labeled intact mosquitoes were placed into a 3.2-mm zirconia rotor and spun at 12 kHz magic-angle spinning for 13 glucose for lipid biosynthesis. However, the 34-ppm peak in the CPMAS spectra of dsi-RNA injected mosquitoes targeting Foxo downstream elements is absent (Fig. 3b) and appears as a negative intensity in the difference spectra (Fig. 3c). The negative aliphatic carbon intensity indicates that the dsi-RNA knockdown of ldlr, atp, and glys inhibited the routing of d-[ 13 C 6 ]glucose for de novo lipid biosynthesis while an increase in the metabolism of the stored lipids in mosquitoes. Interestingly, the reduction of 34-ppm intensity was compensated by increases in 30 and 25-ppm intensities (Fig. 3c), which suggests a complex change in lipid composition that is associated with the knockdowns. FTIR absorption band assignments. Total lipid, protein, and glycogen accumulation in lyophilized mosquitoes were monitored using ATR-FTIR. IR absorption spectrum of lyophilized dsi-control injected diapausing Cx. pipiens is shown in Fig. 4a, along with the IR spectra for the representative cellular components (Fig. 4b): lipid, glycogen, protein, DNA, and chitin. The characteristic IR absorption bands for each lipid, glycogen, protein (BSA), DNA, and chitin were identified by taking a second-order derivative of the absorption spectra (Fig. 5). A second-order derivative (Fig. 5) removes the broad component, revealing the narrow absorption bands from the spectrum. The characteristic absorption bands for lipids are 2956 cm −1 , 2850 cm −1 , and www.nature.com/scientificreports/ 1745 cm −1 . The IR assignments for 2956 cm −1 is for CH 3 antisymmetric stretch, 2920 cm −1 for CH 2 antisymmetric stretch, 2850 cm −1 for CH 2 symmetric, 1745 cm −1 for C=O stretch, and 1468 cm −1 for CH 2 scissoring 16 . These absorption bands are predominant in vegetable oil but absent from other representative cellular components. For protein (bovine serum albumin), Amide I at 1651 and 1631 cm −1 , which appear as the most intense absorption bands are assigned to C=O, C-C, and C-N stretching of the protein-peptide backbone. The Amide II, positioned at 1510-1580 cm −1 , are assigned to in-plane N-H bending vibrations, and C-N and C-C stretching vibrations. Amide I and II bands (1651, 1631, and 1515 cm −1 ) are visible in the second-order derivative spectrum of the dsi-control treated mosquitoes but are absent from the derivative spectra of lipid, glycogen, DNA, and chitin. The IR absorption band assignments for the second-order derivative spectrum of glycogen 17 are as follow: (i) 991 cm −1 is assigned to in-plane bendings of CH 2 , and COH, and CO and COC stretching of glycosidic linkage, (ii) 1017 cm −1 for COH stretching, (iii) 1078 cm −1 for in-plane bending of COH, and (iv) 1149 cm −1 for COC and CC stretching modes of glycosidic linkage and asymmetric ring stretching. Despite the broad overlapping IR absorption bands of glycogen and chitin (Fig. 4b), the second-order derivative spectra of dsi-control treated mosquitoes at 950-1800 cm −1 range (Fig. 7d) closely resembles glycogen (Fig. 5, gray boxes) and not chitin. Total lipid quantification by FTIR. FTIR spectra of lyophilized diapausing females treated with dsi-RNAs (dsi-control, dsi-glys, dsi-atp, and dsi-ldlr) and its second-order derivative spectra are shown in Figs. 6a and 7a, respectively, with the characteristic IR absorption bands representing the total protein, lipid, and glycogen highlighted in blue boxes. Both FTIR spectra and the second-order derivative are normalized to Amide I intensity at 1631 cm −1 (Figs. 6b and 7a), which is almost entirely due to C=O stretching vibrations. The normalization assumes that the molar absorptivity of C=O stretching vibration for each protein secondary structural element www.nature.com/scientificreports/ is essentially the same 18 . The Amide I normalization of FTIR spectra is comparable to the normalization of 175-ppm peak intensity in the 13 C-CPMAS spectra of Fig. 2. Since the uptake of glucose by mosquitoes during the diapause is not routed to protein or chitin biosynthesis, the total amount of protein in individual mosquito during the early window of diapause remains constant. Enlarged IR spectra of dsi-RNA injected diapausing females for the characteristic IR absorption bands for lipids at 2920 cm −1 and 2850 cm −1 are shown in Fig. 6c and its corresponding second-order derivative spectra in Fig. 7b. The total lipids in mosquitoes were estimated based on the IR absorption intensity of CH 2 at 2920 cm −1 . The maximum lipid accumulation was observed for dsi-control mosquitoes followed by dsi-ldlr (33% reduction), dsi-atp (36% reduction), and dsi-glys injected mosquitoes (44% reduction). The total triacylglyceride (TAG) accumulation in mosquitoes was also determined by comparing the intensity of absorption at 1745 cm −1 . The IR band at 1745 cm −1 which is also found in the IR spectrum of vegetable oil (Fig. 4b) is assigned to the C=O stretch of the ester-linked carbons found in glycerol of TAG. Since intense Amide I at 1651 cm −1 overlaps significantly with the 1745 cm −1 , the second-order derivative of the spectra was used to remove the broad component. The resolved narrow C=O stretch band at 1745 cm −1 (Fig. 7c) was used for comparison. The maximum TAG accumulation was observed for dsi-control followed by dsi-ldlr, dsi-atp, and dsi-glys injected mosquitoes. Hence, the dsi-RNA knockdown of the genes glys, ldlr, and atp resulted in the reduction of TAG (lipids) in diapausing mosquitoes. It is important to point out that the FTIR measurements were performed on powdered lyophilized mosquitoes, a group that consisted of 40-60 individuals. The use of a powdered lyophilized sample was critical because of the high sensitivity of the FTIR instrument which can readily differentiate the chemical composition based on the different body segments. Thus, the FTIR spectra of an intact individual mosquito can vary significantly based on the position of the body segment where the measurements were taken. The use of the powdered mosquitoes eliminated the spectral variations associated with the sampling. Total glycogen quantification by FTIR. For glycogen analysis, the overlapping IR absorption bands between 940 and 1180 cm −1 (Fig. 6d) are assigned to the stretching and bending vibrational modes of COH, and CO and COC found in polysaccharides. Although the IR absorption bands found in glycogen and chitin overlap significantly (Fig. 4), the second-order derivative spectra (Fig. 7d) clearly resolves 1078 cm −1 (COH www.nature.com/scientificreports/ plane bending) and 991 cm −1 (COH, COC, and CH 2 stretching) bands unique to glycogen. Also, since glucose uptake during the diapause is not routed to chitin biosynthesis, as determined by solid-state NMR (Fig. 2b), the IR absorption intensity at 1017 cm −1 (Fig. 6d) is directly proportional to the total glycogen accumulation in mosquitoes. The maximum glycogen accumulation was observed for the dsi-control, followed by dsi-glys, dsildlr, and dsi-atp. Thus, the dsi-RNA knockdown of the genes glys, ldlr, and atp showed a significant reduction in total glycogen accumulation in diapausing mosquitoes. The order of total glycogen accumulations from the most to least as measured by FTIR was dsi-control, followed by dsi-glys, dsi-ldlr, and dsi-atp injected mosquitoes (Fig. 6d). This differed from solid-state NMR which measured the maximum 13 C-labeled glycogen accumulation was dsi-control, followed by dsi-ldlr, dsi-glys, and dsi-atp injected mosquitoes (Figs. 2 and 4a). The solid-state NMR data show that the newly synthesized 13 C-labeled glycogen accumulation in dsi-glys was approximately 41% less than the dsi-ldlr injected mosquitoes, but the total glycogen in dsi-glys as measured by the FTIR was greater than dsi-ldlr injected mosquitoes. This indicates that the dsi-glys knockdown primarily suppressed the d-[ 13 C 6 ]glucose utilization for the de novo glycogen biosynthesis without a significant impact on the total glycogen (Fig. 6d, green line). In contrast, the www.nature.com/scientificreports/ dsi-ldlr injected mosquitoes show increased accumulation of 13 C-labeled glycogen but a decrease in total glycogen (Fig. 6d, red line). Hence the ldlr knockdown exhibits increased utilization of the natural-abundance glycogen that was synthesized prior to the d-[ 13 C 6 ]glucose feeding). Glycogen quantification by combined solid-state and FTIR. The solid-state NMR and FTIR quantifications of glycogen accumulations in dis-RNA knockdown mosquitoes are combined and shown in Fig. 8. To combine solid-state NMR and FTIR analyses, the 73-ppm peak intensity of 13 C-natural abundance CPMAS spectrum, shown in Fig. 2b (red), was scaled using a normalization factor to equal to the 1078 cm −1 absorption intensity of the FTIR spectrum of dsi-control mosquitoes (Fig. 6d). The same normalization factor was used to scale the 73-ppm peak intensity of the difference spectra of 13 C-glucose fed mosquitoes (Fig. 2b, top). In addition, a scaling factor of 1/10 was applied to the 73-ppm peak intensity of 13 C-glucose fed mosquitoes to compensate for the 13 C-isotope dilution (mosquitoes were fed on 10% glucose solution containing 1% 13 C-labeled d-[ 13 C 6 ]glucose). The resulting intensity, shown as black bars, is directly proportional to the amount of glycogen that was synthesized from the feeding of provisioned d-[ 13 C 6 ]glucose in diapause-destined females. Subtracting the 13 C-labeled glycogen (black bar) from the total glycogen (intensity of 1078 cm −1 absorption band) results in a gray bar that represents the unlabeled glycogen that was synthesized prior to the d-[ 13 C 6 ]glucose feeding. ## Lipid and glycogen biosyntheses have strong co-dependence. The dsi-RNA injection of glys, atp, and ldlr genes showed varying degrees of de novo glycogen biosynthesis inhibition in diapause-destined female mosquitoes. The dsi-atp injection which targeted the ATP synthase showed the maximum inhibition of d-[ 13 C 6 ] glucose utilization for glycogen biosynthesis, 86% reduction compared to the dsi-control. This was followed by the glycogen synthase knockdown 19 that resulted in a 59% reduction in glycogen accumulation. Finally, the LDLR knockdown which would have interfered with the uptake and trafficking of LDL showed a surprising 31% reduction in glycogen accumulation (Fig. 2c, top). In addition to the inhibition of glycogen biosynthesis, www.nature.com/scientificreports/ solid-state NMR measurements (Figs. 2 and 3c) determined that the dsi-RNA injected mosquitoes showed the reduced 34-ppm intensity which indicated a decrease in the amount of the stored lipids. The reduced lipid accumulation was confirmed by FTIR by the reduced CH 2 vibrational intensities at 2956 cm −1 , 2920 cm −1 , and 2850 cm −1 (Figs. 6c and 7b). This indicated that the dsi-RNA injection not only inhibited de novo lipid biosynthesis but also increased utilization of the stored triacylglyceride that was synthesized before the feeding of d-[ 13 C 6 ]glucose. Hence, the knockdown of Foxo downstream genes has a direct impact on the lipid homeostasis resulting in an undesirable increase in lipid expenditure in diapausing mosquitoes. This suggests that lipid and glycogen biosyntheses have strong co-dependence where the intracellular concentration of one may affect the biosynthesis of the other. One possibility for this co-dependence between lipid and glycogen biosyntheses is that LDLR acts to deliver a signaling lipid to the mosquito brain and regulate the secretion of hormones related to glucose metabolism. For example, in Drosophila, fat-containing LDLR convey information about circulating lipid composition to sensory cells in the brain to regulate insulin signaling 20 . Blocking fat-containing molecules reduced insulin signaling and subsequently modulated glucose metabolism. Although in the early diapause period, the insulin-like peptide is suppressed, its role has not been studied like other 7 or 8 insulin-like peptides and their functional role in diapausing mosquitoes 21 . It may be interesting to examine whether this factor plays a role in mosquito diapause at low temperatures. ## Concluding remarks. A study 19 shows that suppression of the expression of the gene encoding glys during the early diapause period inhibited the synthesis of glycogen and lipids accumulation in mosquitoes. Moreover, suppression of the expression of this gene reduced the survival rate of diapausing females by more than 80% www.nature.com/scientificreports/ in 1 month. Interestingly, in the first month of the diapause program, it was found that the lipid was not used as the main energy source, but the glycogen in the fat body was used first. It is assumed that the use of lipids is determined by sensing the remaining amount of glycogen or glucose in the fat body by a mechanism that is not yet known 22 . Inhibiting the function of the glys gene is another evidence that supports this hypothesis, given the same consequences of inhibiting the storage of the lipid levels in this study. ABC transporters (atp) are ubiquitous in all domains of life. They share the same core structure consisting of two transmembrane regions and two soluble nucleotide-binding domains. The former domains bind and translocate substrates across lipid bilayers while the later domains bind and hydrolyze ATP 23 . This protein plays an important role in the metabolism of lipids in cells by transferring of lipids to peroxisomes. Mutations in the atp gene also cause cytoplasmic accumulation in unbranched and saturated long fatty acids. Accumulation indicates that the mitochondria are unable to metabolize the lipids and thereby inhibiting energy production through beta-oxidation 24 . In a previous study, different activations of genes related to lipid metabolism were found in early diapause. The genes in beta-oxidation metabolism are strongly inhibited in this period, where the diapausing females need to store lipids. In contrast, after exhausting glycogen, the genes in beta-oxidation were up-regulated which include acetyl-coA synthetase, carnitine o-octanoyltransferase, acyl-coA dehydrogenase, 3-hydroxyacyl-coA dehydrogenase, and β-ketoacyl-coA thiolase. The female began to use the stored lipids, following the glycogen exhaustion, by converting the lipids into free fatty acids and then transferring them to mitochondria and peroxisomes for necessary energy metabolism 21 . In Drosophila, the low-density lipoprotein receptor chaperone is an evolutionary conserved endoplasmic reticulum protein involved in the folding, trafficking, and quality control of LDLR proteins 6 . Since the LDLRs mediate the endocytosis of low-density lipids and the recycling of LDLRs is an essential mechanism for delivery of lipids to cells, it is speculated that RNAi suppression of the activity of the ldlr gene will inhibit the metabolism of the lipids in the cell. This speculation was confirmed in this study, and it was shown that suppression of the activity of the ldlr gene inhibits not only fat but also the metabolism of glycogen in the diapausing females (Figs. 3 and 6). Since the diapausing mosquitoes do not feed on other nutrients during the winter, efficient consumption of the stored nutrient is important for survival during the cold winter. In addition, the stored nutrient is also used as an important resource for egg production at the end of diapause. Thus, genetic and biochemical mechanisms to control the nutrient stores and efficient use of the energy resources is important for the overwintering survival in the mosquitoes. To this end, a model has been proposed for genetic mechanisms related to efficient use and storing energy in diapause program 3,25 . The suppression of the insulin signal causes the Foxo to become active in the fat body cells and activate the downstream genes, which initiates the alternative development program (diapause) in the adult females of Cx. pipiens. Three of the downstream genes of the Foxo signaling are proposed in the regulation of diapause energetics and are an essential factor in synthesis, storage, and transfer of glycogens and lipids. In this study, we have shown that the genes encoding glycogen synthase (glys), ATP-binding cassette transporter (atp), and the low-density lipoprotein receptor chaperone (ldlr), that are the targets of Foxo transcription factor, play an important role in the energy homeosis in the mosquitoes destined for overwintering diapause. ## Materials and methods Insect rearing. The detailed rearing of a stock colony of Cx. pipiens can be found elsewhere 3 , but briefly, the Cx. pipiens colony was established in September 2000 from larvae collected in Columbus, OH, and additional field-collected mosquitoes were added to the laboratory colony in 2009. The colony was reared at 25 °C and 75% relative humidity under a 15-h light:9-h dark (L:D) photoperiod as previously described. When larvae reached www.nature.com/scientificreports/ the second instar, rearing containers were placed under one of two environmental conditions: nondiapausing females were generated by rearing at 18 °C, 75% relative humidity, and 15:9 L:D. To induce diapause, mosquitoes were reared at 18 °C, 75% relative humidity, and 9:15 L:D. To confirm diapause status, primary follicle and germarium lengths were measured, and the stage of ovarian development was determined according to the methods described by Christophers 26 . Synthetic dicer-substrate siRNA (dsi-RNA) injection into adult female mosquitoes. Targeting of the genes encoding glycogen synthase (glys, vectorbase gene i.d.: CPIJ005086), ATP-binding cassette transporter (atp, CPIJ012364) and low-density lipoprotein receptor chaperone (ldlr, CPIJ010816) was performed as described previously 3,4 . Briefly, the DsiRNAs were used in silencing experiments against these genes include dsi-glys, dsi-atp and dsi-ldlr that correspond to target sequences. The sequences of these siRNA duplexes, which were purchased from Integrated DNA Technology (IDT, Coralville, IA) and confirmed through BLAST searches to have no significant homology to Cx. pipiens genes other than glys, atp and ldlr, are as follows: dsi-glys: 5′-rArG-rCrGrArCrUrCrCrArCrGrUrUrGrArArGrUrUrGrUrUrGrGrUrU-3′/5′-rCrCrArArCrArArCrUrUrCrArArC rGrUrGrGrArGrUrCrGCT-3′, dsi-atp: rGrArGrGrArArArUrCr CrGrArCrGrGrArCrUrGrUrArGrUrGrCrU rC-3′/5′-rGrCrArCrUrArCrArGrUrCrCrGrUrCrGrGrArUrUrUrCrCTC-3′, dsi-ldlr: 5′-rCrArUrArArArGr-ArUrGrGrCrUrCrGrArUrUrGrUrCrArUrCrGrArU-3′/5′-rCrGrArUrGrArCrArArUrCrGr ArGrCrCrArUr-CrUrUrUrATG-3′. All data were confirmed following knockdown with dsi-glys, dsi-atp and dsi-ldlr, suggesting that none of the phenotypes reported herein were the result of off-site targeting by the dsiRNA. A scrambled negative control dsiRNA, a dsi-control duplexes lacking significant sequence homology to any genes in the Cx. pipiens genome, was used for control experiments, dsi-control: 5′-rGrArArGrArGrCrArCrUrGrArUrAr-GrArUrGrUrUrArGrCGT-3′ /5′-rArCrGrCrUrArArCrArUrCrUrArUrCrArGrUrGrCrUrCrUrUrCrCrG-3′. None of the results reported were observed in control-injected females, which were not significantly different than wild type females for any of the phenotypes assessed. Dsi-control injected and WT females in the results represent the natural abundance of the RNA transcript levels and nutrient composition. Dsi-RNA efficiency evaluation by qRT-PCR. Total RNA samples were extracted with TRIzol (Invitrogen) from three batches of 8-12 adult female mosquitoes on 7-10 days after dsi-RNA injection. Only adult females were kept in 12″ × 12″ × 12″ mesh cages with unrestricted access to clean water and a 10% glucose solution with or without 1% 13 C-labeled d-[ 13 C 6 ]glucose. cDNA was synthesized using a High-Capacity cDNA Archive Kit (Applied Biosystems). Real-time PCR reactions were performed on a 7300 Real-Time PCR System (Applied Biosystems) using SYBR green supermix (BioRad). Reactions were run in triplicate using three independent biological replicates of each sample. Primer pairs to glys, atp, ldlr, and ribosomal protein large subunit 19 (rpl19, a loading control) yield single peaks in the dissociation curve as described in previous study 27,28 . mRNA expression levels of glys, atp, and ldlr were determined relative to rpl19 expression by relative quantification. The following qRT-PCR primers sets were used: qRT-glys, CGA TCC ACG AGT TCC AGA AT, and GCG TCT TCT CCA GGT CAA AG; qRT-atp, ATC CGA CGG ACT GTA GTG CT, and AGG GTC AGT GCA TTT TCA CC; qRT-ldlr, AAT GTT TTC CGG ATT GCT TG, and CTG GAT CAA TGG GAG GAA GA; qRT-rpl19, CGC TTT GTT TGA TCG TGT GT, and CCA ATC CAG GAG TGC TTT TG. Statistical significance of differences in transcript levels was determined using a Student's t test between the relative transcript values of dsi-glys, dsi-atp, or dsi-ldlr injected vs. control samples (dsi-control injected), using three biologically independent replicates for each gene (three groups of 8-12 RNAi-injected females). A P value of less than 0.05 was considered a significant transcriptlevel change. 13 C-isotope labeling of Cx. Pipiens. Dsi-RNA injected female Cx. pipiens after adult eclosion were fed for 7 days on sponges soaked with 10% glucose solution containing 1% 13 C-labeled d-[ 13 C 6 ]glucose. Uniformly 13 C-isotope labeled d-[ 13 C 6 ]glucose (isotopic enrichment of 99%) was purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA). After the 7-days feeding, the mosquitoes were frozen at − 80 °C then lyophilized for 3 days. Lyophilized intact mosquitoes were weighed and packed into a 3.2-mm zirconia rotor for solid-state NMR analysis 12,19 . Solid-state NMR. Solid-state 13 C cross-polarization magic-angle spinning (CPMAS) NMR 29 of diapausedestined female Cx. pipiens injected with dsi-RNA and then fed for 7 days with 10% glucose solution containing 1% 13 C-labeled d-[ 13 C 6 ]glucose was collected on 11.75-T (proton radio frequency of 500 MHz) Bruker Avance NEO with a double resonance HX probe. 13 C-CPMAS NMR was performed as described earlier 12,19 . Briefly, lyophilized mosquitoes were contained in a 3.2-mm outer diameter zirconia rotor with Kel-F endcap spinning at 10 kHz. Proton-carbon matched cross-polarization ramp was at 50 kHz with 2-ms contact time. The proton dipolar decoupling was achieved by applying continuous-wave spinal64 30 on the 1 H channel during acquisition. The π pulse length was 2.5 µs for 1 H and the recycle delay was 5 s. The line broadening for the spectrum was 50 Hz. The spectra were normalized to equal 175-ppm intensity of peptidyl-carbonyl carbons in proteins. FTIR. FTIR spectra were obtained using Thermo Scientific Nicolet iS50 Spectrometer with iS50 ATR module containing a diamond crystal (Madison, WI, USA), equipped with deuterated triglycine sulfate detector working at room temperature. Spectra acquisition was performed in the 500-4000 cm −1 range with 4 cm −1 resolution, after 100 scans, using the N-B Strong as apodization and 2-levels zero fillings of the interferogram giving data spacing of 0.482 cm −1 ; these were converted into absorbance using Thermo Scientific OMNIC software version 9.2 (https ://www.therm ofish er.com/order /catal og/produ ct/INQSO F018#/INQSO F018). Powdered lyophilized mosquito samples, following the solid-state NMR measurements, were removed from the zirconia rotor, www.nature.com/scientificreports/ and then measured directly without any further preparation. Background measurement of air was taken and automatically subtracted from the sample measurements. Spectra were acquired from five randomly selected locations across the sample to minimize sampling bias. Between measurements, the ATR crystal was carefully cleaned using ethanol (Sigma-Aldrich, analytical standard) and dried with light-duty tissue wipers. Using the same method of analysis, FTIR spectra of canola oil, glycogen from bovine liver (Sigma-Aldrich), bovine serum albumin (BSA) (lyophilized powder, Sigma-Aldrich), deoxyribonucleic acid sodium salt from salmon testes (Sigma-Aldrich), and chitin from shrimp (Sigma-Aldrich) were obtained. These were used as standards for the identification of absorption peaks for lipids, glycogen, protein, DNA, and chitin, respectively. Received: 10 March 2020; Accepted: 25 September 2020

chemsum

{"title": "Dsi-RNA knockdown of genes regulated by Foxo reduces glycogen and lipid accumulations in diapausing Culex pipiens", "journal": "Scientific Reports - Nature"}

<i>p</i>-<i>tert</i>-butylthiacalix[4]arenes_functionalized_by_<i>n</i>-(4’-nitrophenyl)acetamide_an

## Abstract: New p-tert-butylthiacalix [4]arenes, which are mono-, 1,2-di-and tetrasubstituted at the lower rim containing N-( 4'nitrophenyl)acetamide and N,N-diethylacetamide groups in cone and partial cone conformations have been synthesized. Their complexation ability towards a number of tetrabutylammonium salts n-Bu 4 NX (X = F − , Cl − , Br − , I − , CH 3 CO 2 − , H 2 PO 4 − , NO 3 − ) was studied by UV spectroscopy. The effective receptor for the anions studied as well as selective receptors for F − , CH 3 CO 2 − and H 2 PO 4 − ions, which based on the synthesized thiacalix[4]arenes, have been obtained. It was shown that p-tertbutylthiacalix[4]arene tetrasubstituted at the lower rim by N-(4'-nitrophenyl)acetamide moieties bonded to the anions studied with association constants within the range of 3.55 × 10 3 -7.94 × 10 5 M −1 . Besides, the binding selectivity for F − , Cl − , CH 3 CO 2 − , and H 2 PO 4 − anions against other anions was in the range of 4.1-223.9. Substituting one or two fragments in the macrocycle with N,Ndiethylacetamide groups significantly reduces the complexation ability of the receptor. In contrast to the 1,3-disubstituted macrocycle containing two N-(4'-nitrophenyl)acetamide moieties, the 1,2-disubstituted thiacalix[4]arene, which contains only one such fragment and a N,N-diethylacetamide moiety, selectively binds F − anions. ## Introduction Anions play a key role in many biochemical processes as substrates and/or cofactors in enzymatic reactions , in the environment (phosphate and nitrate in the ponds provoking their eutrophication) , and in phase-transfer catalysis . The dysfunction in the regulation of anions is a cause of many diseases . Thus, selective synthetic receptors can be used in medicine as medicinal and diagnostic agents. However, the design and synthesis of the systems for recognizing anions remains one of the important scientific challenges in organic chemistry . Anion-receptor biomimetics aimed at devel-Scheme 1: Synthesis of 1,3-di-and tetrasubstituted thiacalix arenes 2 and 3. Conditions: (i) macrocycle 1 (1 equiv), 2-bromo-N-(4'-nitrophenyl)acetamide (4 equiv), Na 2 CO 3 (4 equiv), CH 3 CN, reflux. oping synthetic analogs of the natural compounds that offer a deeper understanding of a number of biological processes. The design of anion receptors is quite a challenge for several reasons. The anions are larger in size than cations and therefore have a smaller charge in relation to their radius. It means that the electrostatic interactions for anions binding are weaker than for cations. Anions have various shape and geometry, and the design of receptors that are complementary to a specific type of anion is needed. (Thia)calix arene derivatives are a favorable platform for the design of such structures . Due to their macrocyclic nature and the possibility to modify them in three different ways (upper and lower rims and bridge fragments), they show the ability to selectively recognize and bind different types of substrates . Previously, our scientific group showed that 2-bromo-N-(4'nitrophenyl)acetamide is the regioselective alkylating reagent for p-tert-butylthiacalix arene. In these reactions, derivatives of thiacalix arene variously substituted at the lower rim form 1,2-, 1,3-di-and trisubstituted macrocycles , depending on the nature of the alkali metal carbonates and solvent. In this paper, we describe the regioselective synthesis of p-tert-butylthiacalix arene monosubstituted at the lower rim by N,Ndiethylacetamide fragment and its further functionalization with the N-(4'-nitrophenyl)acetamide moiety. We also calculated the proposed model of anion binding for the new and previously synthesized thiacalix arenes and compared their complexation properties toward number of singly charged anions (F − , Cl − , Br − , I − , CH 3 CO 2 − , H 2 PO 4 − , NO 3 − ). The obtained novel synthetic anion receptors hold promise for their potential application in the development of more sophisticated biomimetic materials and diagnostic agents. ## Results and Discussion Synthesis of p-tert-butylthiacalix arene derivatives containing N-(4'-nitrophenyl)acetamide and N,N-diethylacetamide moieties at the lower rim The regioselective synthesis of p-tert-butylthiacalix arenes partially substituted at the lower rim, is an important challenge because it allows the sequential functionalization of the macrocyclic platform with the necessary substituents. It was shown in acetonitrile using the weak base Na 2 CO 3 that p-tert-butylthiacalix arene 2, 1,3-disubstituted at the lower rim, is formed with a low yield of 50% . We estimated the effect of various reagent ratios for increasing the yield of the major product. The increase of the ratio to 1:4:4 = macrocycle 1/Na 2 CO 3 / 2-bromo-N-(4'-nitrophenyl)acetamide resulted in an increase of the yield of the 1,3-disubstituted thiacalix arene to 60%. In addition, p-tert-butylthiacalix arene 3 tetrasubstituted at the lower rim by N-(4'-nitrophenyl)acetamide fragments in cone conformation was obtained with 10% yield (Scheme 1). 1 Н, 13 С, 2D NOESY NMR, IR spectroscopy, mass spectrometry (MALDI-TOF) confirmed the macrocycle 3 structure. The elemental analysis gives us the composition of 3. In contrast to the 1 H NMR spectrum of 1,3-disubstituted thiacalix arene 2 , the signals of tert-butyl, oxymethylene and aryl protons are observed in the 1 H NMR spectrum of tetrasubstituted macrocycle 3 as a single singlet. There are no signals of the hydroxy protons and only one singlet corresponding to amide protons is observed (see Supporting Information File 1, Figure S1). This indicates complete substitution of the initial macrocycle 1. Scheme 3: Synthesis of tetra-and 1,2-disubstituted thiacalix arenes 5 and 6, respectively. Conditions: (i) macrocycle 4 (1 equiv), 2-bromo-N-(4'nitrophenyl)acetamide (6 equiv), Na 2 CO 3 (6 equiv), acetone, reflux; (ii) macrocycle 4 (1 equiv), 2-bromo-N-(4'-nitrophenyl)acetamide (6 equiv), Cs 2 CO 3 (6 equiv), acetone, reflux. In the MALDI-TOF mass spectrum of the compound 3 (M(C 72 H 72 N 8 O 16 S 4 ) = 1432.4), a molecular ion peak with Na + cation (m/z (M + Na + ) = 1455.5) was observed (see Supporting Information File 1, Figure S4). p-tert-Butylthiacalix arenes monosubstituted at the lower rim are also valuable precursors for a further functionalization of the macrocycle. However, their preparation is significantly more difficult than the synthesis of 1,3-disubstituted thiacalix arenes. Bulky functional groups able to form hydrogen bonds with free phenolic hydroxy groups and therefore block three residual hydroxy groups are usually applied in the synthesis of the monosubstituted derivatives of p-tert-butylthiacalix arene . For the synthesis of the monosubstituted derivative, the bulky N,N-diethylacetamide group able to participate in the formation of hydrogen bonds was used. The monosubstituted derivative 4 was obtained by the alkylation of p-tert-butylthiacalix The introduction of the N-(4'-nitrophenyl)acetamide group into the lower rim of p-tert-butylthiacalix arene 4 is also of interest because this fragment contains a polar NH group able to interact with anionic substrates and a chromophore fragment necessary for the spectrophotometric detection of the complex formation. In this regard, the alkylation of monosubstituted derivative 4 with 2-bromo-N-(4'-nitrophenyl)acetamide was further studied in the presence of Na 2 CO 3 , K 2 CO 3 and Cs 2 CO 3 . It was found that tetrasubstituted p-tert-butylthiacalix arene 5 is formed with 43% yield in the presence of Na 2 CO 3 . Using Cs 2 CO 3 , the 1,2-disubstituted thiacalix arene 6 was isolated as partial cone with 42% yield (Scheme 3). In the case of K 2 CO 3 , a hardly separable mixture of differently substituted derivatives formed. The presence of two singlets (with an integral intensities of 3:1) of proton signals of t-Bu groups, one signal of oxymethylene protons and two singlets of OH groups in the 1 H NMR spectrum of 4 confirms the formation of the monosubstituted product in cone conformation (see Supporting Information File 1, Figure S5). The absence of signals of OH groups in the 1 H NMR spectrum of 5 indicates the complete substitution of the lower rim of initial macrocycle 4 (see Supporting Information File 1, Figure S9). Four singlets (with equal integrated intensities) of t-Bu groups in the 1 H NMR spectrum show the asymmetric structure of 6 (see Supporting Information File 1, Figure S13). Also the presence of two singlets of OH groups and one singlet of NH group in the 1 H NMR spectrum of 6 confirms the formation of disubstituted product. The spatial structure of the compounds 4-6 was studied by 1 H-1 H NOESY NMR spectroscopy. In the NOESY NMR spectrum of the compound 4 (see Supporting Information File 1, Figure S7), the presence of cross peaks due to the dipole-dipole interaction of protons of the oxymethylene group with the hydroxy protons, as well as the cross peaks between the aromatic protons of the macrocycle, confirm that the macrocycle 4 is in the cone conformation. The cross peaks between the protons of substituents and the oxymethylene groups, the cross peaks between the protons of the amide groups and the N,N-diethylacetamide group, and also those between the aromatic protons of the macrocycle in the 2D NOESY NMR spectrum of 5 (see Supporting Information File 1, Figure S11) indicate that the macrocycle 5 is in the cone conformation. The presence of cross-peaks between the amide proton and protons of tert-butyl groups, the aromatic protons of the p-nitrophenyl substituent and protons of tert-butyl groups in 2D NOESY NMR spectrum of 1,2-disubstituted compound 6 (see Supporting Information File 1, Figure S15), as well as the cross peaks between the aromatic protons of the macrocycle and the oxymethylene protons of N,N-diethylacetamide moiety indicate the presence of substituents on opposite sides of the macrocyclic ring. This fact, as well as the cross-peak due to the dipole-dipole interaction of protons of the hydroxy groups confirms the presence of the macrocycle 6 in the partial cone conformation. In the IR spectra of the compounds 4 and 6, the absorption bands for the stretching vibrations of the hydroxy groups (ν 3330 and 3745 cm −1 , respectively) were observed. However this absorption band is absent in the IR spectra of compounds 3 and 5. It confirms complete substitution of the initial thiacalix arenes. It is obvious that the cone formation of the thiacalix arenes 3 and 5 is stabilized by intramolecular hydrogen bonding observed in the IR spectra. The bands of stretching vibrations for the NH bonds of N-(4'-nitrophenyl)acetamide fragment are observed in the IR spectra of the macrocycles 3 and 5 as a narrow and broadened bands in the region of 3383-3278 cm −1 (see Supporting Information File 1, Table S1). In the IR spectra of the compounds 4, 5 and 6, absorption bands of valence vibrations for the N,N-diethylacetamide group (ν 1658, 1648 and 1665 cm −1 , respectively) were observed, which were absent in the IR spectrum of the tetrasubstituted macrocycle 3 (see Supporting Information File 1, Table S1). Complexation properties of the synthesized p-tert-butylthiacalix arenes towards some single-charged anions There are several ways of binding negatively charged substrates. Usually, receptors for anions are charged systems capable of electrostatic interaction with anions or neutral systems using other types of interactions, such as donor-acceptor interactions, hydrogen bonds, hydrophobic effects, etc. . Modified macrocycles with the amide fragments at the lower rim can form complexes with the anions by hydrogen bonds of the amide group with the guest. It was shown that (thia)calix arene derivatives with urea and thiourea fragments at the lower rim can bind anions through hydrogen bonds [49, . In this regard, it was suggested that the synthesized thiacalix arenes 3, 5 and 6 could be potential receptors for anions because they contain proton donor groups (amide, hydroxy) at the lower rim. Also, the comparison of the complexation properties of early synthesized thiacalix arenes 2, 7-10 (Figure 1) with N-(4'-nitrophenyl)acetamide and N,N-diethylacetamide fragments was of interest. Initially, molecular modeling of the host-guest complexes of the above-mentioned thiacalix arenes 2, 3, 5-10 with a number of single-charged anions (F − , Cl − , Br − , I − , CH 3 CO 2 − , H 2 PO 4 − , NO 3 − ) was carried out at a semi-empirical level using the quantum-mechanical method, PM3 (HyperChem 7.0). The proposed model for the binding of anions by p-tert-butylthiacalix arenes containing proton-donor (amide, hydroxy) groups was studied in order to identify steric and/or electronic hindrances to the complex formation. Review of literature data indicate a good agreement between the results of calculations involving the complexes geometry determined by this method and experimental data. The PM3 method is used quite productively for molecular design and modeling of the receptor structures . Unfortunately, this calculation method does not provide adequate absolute energy values for the molecules calculated. Therefore, we further discuss the comparative values of the energies of different complexes . As an example, Table 1 shows the difference in the energy gain of the resulting complexes of the macrocycles 3 and 9 with a number of single-charged anions. It can be clearly seen that a linear dependence of the anion size on the energy difference is observed for halide anions. That is, the larger the anion size, the lower the gain in the energy of the complex formed. Thus, the binding efficiency should decrease in the range from the F − ion to the larger I − ion. The receptor properties of the synthesized compounds 3, 5, 6 toward the tetrabutylammonium salts n-Bu 4 NX (X = F − , Cl − , Br − , I − , CH 3 CO 2 − , H 2 PO 4 − , NO 3 − ) were studied by UV spectroscopy for investigation the influence of structural factors (the macrocycle conformation, the number and the nature of the substituents) on the complexation properties of p-tert-butyl-thiacalix arenes. In these thiacalix arenes, both proton donating secondary amide (N-(4'-nitrophenyl)acetamide fragment) and hydroxy (phenolic fragments) groups can participate in anion binding. To determine the possibility of the tetrabutylammonium cation binding by the synthesized thiacalix arenes, solutions of the compounds 3, 5 and 6 in the presence of a 10-fold excess of were studied in CDCl 3 by 1 H NMR spectroscopy. In the 1 H NMR spectra, the chemical shifts of the n-Bu 4 N + protons did not change indicating the absence of interaction between the thiacalix arenes obtained and tetrabutylammonium cation. The complexation of anions by compounds 3, 5 and 6 was studied in the presence of a 200-fold excess of n-Bu 4 NX in chloroform. The most significant changes in the absorption spectra of p-tert-butylthiacalix In the absorption spectra of the thiacalix arene 5 with n-Bu 4 NX, the changes are observed only in the interaction with F − and H 2 PO 4 − anions. A strong hypochromic effect on the absorption band maximum in both cases and a weak bathochromic shift of the absorption band in the interaction of the macrocycle 5 with F − anions were observed (see Supporting Information File 1, Figure S17). Interaction of thiacalix arene 6 with the F − , CH 3 CO 2 − , H 2 PO 4 − anions have a complicated character in the spectra. There is a bathochromic shift and a hypochromic effect on the maximum of the absorption band at 305 nm. In addition, a new absorption maximum appears in the region of 335 nm (see Supporting Information File 1, Figure S18). The complexation ability of the compounds 3, 5 and 6 in relation to the anions (F − , Cl − , Br was quantitatively evaluated by the stoichiometry and the association constants (Table 2) of the formed complexes. They were determined by the isomolar series method, that all the thiacalix arenes studied form in CHСl 3 the 1:1 complexes with selected n-Bu 4 NX (see Supporting Information File 1, Figure S19). The association constants of the complexes studied were estimated by the dilution method (Figure 2). The appropriate complexation constants (Table 2) were determined by the Benesi-Hildebrand method . Trisubstituted thiacalix arene 8 in cone conformation binds the F − anion with high selectivity against other anions studied. Apparently, the presence of a bulky diethylamide group makes it difficult to bind other anions. However, trisubstituted thiacalix arene 9 in partial cone conformation effectively binds almost all the anions studied except the large I − ion. Among all the macrocycles studied tetrasubstituted at the lower rim p-tert-butylthiacalix arene 3 containing four N-(4'-nitrophenyl)acetamide fragments binds anions most effectively. In some cases, the selectivity of the binding of F − , CH 3 CO 2 − and H 2 PO 4 − anions by the macrocycle 3 with respect to I − and NO 3 − anions is significant according to Table 3. Replacement of one N-(4'-nitrophenyl)acetamide fragment in the compound 3 by N,N-diethylacetamide group leads to a significant decrease in the complexation ability of the macrocycle 5, while the replacement of two fragments leads to the absence of complexation between the thiacalix arene 10 and the anions studied . Compared to the macrocycle 3, the complexation properties of other thiacalix The comparison of the experimental data obtained (logarithms of the association constants of the complexes of the studied thiacalix arenes with a series of anions) and the results of quantum mechanical calculations (energy change in the formation of guest-host complexes calculated by the PM3 method) (HyperChem 7.0) was the final stage of the work (Table 4). The results of the semi-empirical modeling are in a good agreement with spectrophotometric titration data in the case of halide ions: the efficiency of the anion binding by thiacalix arenes 3 and 9 decreases in the range from F − ion to the larger I − ion (Table 4). In the case of the macrocycles 3 and 9, the logarithm of the association constant of the resulting complexes increases from the energy gain difference calculated using the semiempirical PM3 method (Table 4). However, an unambiguous tendency is observed only for spherical substrates (halide ions agreement with the logarithms of the association constant (Table 4). An increase in the logarithm of the association constant correlates with an increase in the energy change in the formation of the host-guest complexes in the case of the macrocycles 3, 7. For the compounds 2, 6, 9, the efficiency of complexation is close. ## Conclusion New mono-, 1,2-di-and tetrasubstituted at the lower rim p-tertbutylthiacalix arenes containing N-(4'-nitrophenyl)acetamide and N,N-diethylacetamide groups in cone and partial cone conformations were synthesized. The calculations of the proposed model for the anion binding by the synthesized thiacalix arenes were carried out using molecular modeling (the quantum mechanical method PM3). Also, their complexation ability toward a number of tetrabutylammonium salts n-Bu

chemsum

{"title": "<i>p</i>-<i>tert</i>-Butylthiacalix[4]arenes functionalized by <i>N</i>-(4\u2019-nitrophenyl)acetamide and <i>N</i>,<i>N</i>-diethylacetamide fragments: synthesis and binding of anionic guests", "journal": "Beilstein"}

fast_and_quantitative_phospholipidomic_analysis_of_sh-sy5y_neuroblastoma_cell_cultures_using_lc-ms/m

## Abstract: Global lipid analysis still lags behind proteomics with respect to the availability of databases, experimental protocols, and specialized software. Determining the lipidome of cellular model systems in common use is of particular importance, especially when research questions involve lipids directly. In Parkinson's Disease research there is a growing awareness for the role of the biological membrane, where individual lipids may contribute to provoking -Synuclein oligomerisation and fibrillation. We present an analysis of the whole cell and plasma membrane lipid isolates of a neuroblastoma cell line, SH-SY5Y, a commonly used model system for research on this and other neurodegenerative diseases.We have used two complementary lipidomics methods. The relative quantities of PC, PE, SMs, CL, PI, PG, and PS were determined by 31 P NMR. Fatty acid chain composition, and their relative abundances within each phospholipid group were evaluated by LC-MS/MS. For this part of the analysis, we have developed and made available a set of Matlab scripts, LipMat. Our approach allowed us to observe several deviations of lipid abundances when compared to published reports regarding phospholipid analysis of cell culture or brain matter. Most striking was the high abundance of PC (54.7 ± 1.9 %) and low abundance of PE (17.8 ± 4.8%) and SMs (2.7 ± 1.2%). In addition, the observed abundance of PS was smaller than expected (4.7 ± 2.7 %), similar to the observed abundance of PG (4.5 ± 1.8 %). The observed fatty acid chain distribution was similar to whole brain content with some notable differences: ## Introduction Lipids play a critical role in many cell processes including regulation of transcription 1 , protein and metabolites distribution 2 , energy metabolism 3 , cell apoptosis induction 4 , as well as protein folding and misfolding 5,6 . However, in many cases we still lack basic knowledge of the cell lipid composition, and how it varies as a function of the sub-cellular compartment or cell state. This deficiency is in part caused by the great diversity displayed by lipid species in cells. The combination of headgroups and fatty acid chains (FA), each with different length and level of saturation, give rise to thousands of possible lipid species in the phospholipid class alone. The analytical problem is exacerbated when accurate quantitative data is also sought 7,8 . The lipidomic gap in knowledge is evident for the neuroblastoma cell line SH-SY5Y. This cell line is of human origin, catecholaminergic, easy to maintain and reported to be differentiable into a neuron-like a phenotype 9 . These properties have led SH-SY5Y to be the cellular model of choice when studying neurodegenerative diseases like Alzheimer's disease, Amyotrophic Lateral Sclerosis, and, in particular, Parkinson disease (PD) 9,10 . Lipids are implicated in all these diseases 11,12 , and for PD this connection is particularly strong 13 . α-Synuclein, a key protein of PD pathology, is thought to be a regulator of vesicle recycling in presynaptic terminals through reversibly contacting the lipid membrane in a carefully timed fashion 14 . Moreover, its misfolding is influenced by particular lipids and the physical state of the membrane which in turn affects its rate of oligomerization 15 . Given the diverse and robust evidence for the role of lipids in -Synuclein pathology, it is then disconcerting to realize that lipid the composition of one of the vital cell models used in PD research -SHSY5Yis still not yet known in detail. We, therefore, focus on the determination of phospholipid content of the whole cell and plasma membrane (PM) enriched SH-SY5Y cell isolates. First, we have determined the abundance of individual phospholipid headgroups by 31 P NMR, using approaches similar to those developed by Bosco et al. 16 . Then, we have determined the FA composition of individual lipid headgroups by LC-MS/MS with an iterative exclusion technique 7 . By this method, we have generated large amounts of data which is required to be processed on a batch-to-batch basis. To achieve this in an expedient and informative way, we have automated the approach for LC-MS/MS by developing a script in Matlab. This script, designated LipMat, is able to: (i) build libraries of predicted lipid fragmentations based on information collected from literature and based on user input, (ii) find possible lipid species by comparing intact m/z values and by scoring MS2 fragments, and (iii) do semi-quantitative FA abundance analysis from MS1 spectra chromatograms. ## Reagents and chemicals Lipids were purchased from Avanti Polar Lipids Inc. (Alabaster, Alamaba, US). These were all used without further purification. The following primary antibodies were used: β-Actin (AC-15) and Na + /K + ATPase α (H-3) from Santa Cruz Biotechnology (Texas, USA), Nucleophosmin (FC-61991) from ThermoFisher Scientific (Massachusetts, USA) and Calnexin from Abcam (Cambridge, UK). Lipase inhibitors (FIDI, U73122, and D609) were acquired from Tocris Bioscience Ltd (Bristol Somerset, UK). Growth medium, solvents, and fine chemicals were purchased from Sigma Aldrich (Germany), and used as described in the separate experimental sections, vide infra. ## SH-SY5Y cultivation SH-SY5Y cells (a generous gift from Prof. Kari Fladmark) were cultured in Dulbecco´s modified Eagle´s medium with high glucose supplemented with 10% fetal bovine serum and 1% Penicillin-Streptomycin, in a humidified atmosphere at 310 K and 5% CO2. ## Whole cell lipid extraction Whole cell lipid extraction was performed based on a modified version of a previously published method 17 . Overview of whole cell lipid sample preparation and plasma membrane isolation is provided in Supplementary Figure S1. Briefly, SH-SY5Y cells were harvested into PBS and pelleted by centrifugation (900 x g, 5 min, 277 K). The supernatant was removed, and the pellet resuspended using 1 mL PBS. To this, 330 µL of guanidine chloride and thiourea mix (3:0.75 molar ratio) was added, and the final mixture was vortexed and stored at 193 K until further use. The frozen material was freezedried and stored at 253 K until lipid extraction. For the extraction, the powdered material was resuspended in an organic extraction mixture (dichloromethane/methanol 3:1 with 0.5 mg/mL triethylammonium chloride, TEAC). Milli-Q water was added (1:1 by volume) and the mixture was transferred into a separating funnel where it was diluted with methanol until a monophasic solution was formed. The solution was agitated briefly and made biphasic again by the addition of more dichloromethane. The denser organic phase was collected using a separation funnel. The water fraction was washed three more times with dichloromethane. Collected organic fractions were pooled and concentrated in vacuo and stored in the form of a dry lipid layer under nitrogen, in the dark, at 293 K until analysis. ## Plasma membrane enriched lipid sample preparation The protocol used in this study was modified from two existing protocols 18,19 . Cells from 5 x 150 mm plates were harvested by trypsinization, pelleted and washed twice with ice-cold PBS and centrifuged (900 x g, 4 min, 277 K). Hypotonic buffer (2 mL, 10 mM Tris, pH 7.8) with FUD (0.1 mL of a concentrated stock of F191 (200 µM), U73122 (1 mM) and D609 (1.5 mM) in a solution of DMSO/water (1:9)) were layered over the cell pellet and removed immediately. The cell pellet was then resuspended using 6 mL of hypotonic buffer and rested on ice for 4 min. After the incubation, 15 mL of H2O was mixed into the suspension. The resulting cell suspension was subsequently incubated for 3 min on ice before the cells were passed through a hypodermic syringe 12 times and afterwards centrifuged (1000 x g, 1 min, 277 K). The pellet was washed with washing buffer (10 mM Tris pH 7.5, 2.5 mM MgCl2, 2.5 mM CaCl2, 10 mM NaCl, and FUD) and centrifuged (1000 x g, 1 min, 277 K). The supernatants from the two centrifugation steps were pooled and diluted with hypotonic buffer. The plasma membrane was isolated by loading the homogenate onto a sucrose gradient, consisting of 30% (w/v) and 45% (w/v) sucrose solutions, and centrifugation (3270 x g, 30 min, 277 K). The PM-enriched fraction was collected and diluted with hypotonic buffer and centrifuged (3270 x g, 20 min, 277 K). The pellet was resuspended in 500 µL of GTCU (6 M guanidinium chloride/1.5 M thiourea). The resuspension was freeze-dried within the same day, resulting material was then stored at 253 K until lipid extraction could be performed, using the same approach as for the whole-cell samples. ## Western blot analysis Samples collected from PM isolations were resuspended in 1xRIPA buffer (150 mM NaCl, 5 mM EDTA, 1 % NP-40, 0.5 % sodium deoxycholate, 0.1 % SDS, 50 mM Tris, pH 8.0), sonicated and centrifuged (17,000 x g, 5 min, 277 K). The total protein concentration was measured using a standard BCA assay 20 , and the protein composition of the samples was analyzed using 10% SDS-polyacrylamide gels and subsequently transferred to a nitrocellulose membrane. The membrane was blocked for one hour with 7% dry milk in PBS with Tween-20 (0.05%; v/v) and incubated overnight at 277 K with primary antibodies to Na + /K + ATPase (1:5000), Nucleophosmin (1:10000), Calnexin (1:1000) and β-Actin (1:1000). The primary antibodies were detected using HRP-conjugated secondary antibodies (1:10000, 1 h, 277 K). Protein bands were visualized using the SuperSignal West Femto maximum sensitivity substrate (Thermo Fisher, USA) and images collected with a Molecular Imager® ChemiDocTM XRS + imaging system and the ImageLabTM software version 3.0 (Bio-Rad, USA). ## Solution phase 31 P NMR Dried lipid films were dissolved in the Culeddu-Bosco "CUBO" solvent system 16 . Data acquisition was similar to our previously published work 17 , using a Bruker BioSpin NEO600 spectrometer equipped with cryogenic probe operating at 300 K for all data. 31 P spectra were acquired at 242.93 MHz using inverse gated proton decoupling, with 3072 scans per sample and a spectral width of 54.16 ppm. An overall recovery delay of 8 s was used, a setting that ensured full relaxation of the 31 P nuclei between scans. Data were processed using an exponential line broadening window function of 1.0 Hz prior to Fourier transformation, followed by manual phase correction and automatic baseline correction. The spectra were calibrated by setting the most abundant phospholipid signal in the samplephosphatidylcholine (PC)to zero ppm. All peaks were then deconvoluted. Peak assignment to individual phospholipids was done according to previously published work 17,21 . The analyses were done using TopSpin 4.0.1. A total of four independent whole cell samples and three plasma membrane samples 31 P spectra were collected. ## LC-MS/MS experimental setup Accurate mass LC-MS and MS/MS was performed on a Thermo Q-Exactive mass spectrometer and a Dionex Ultimate 3000 UPLC (Thermo Fisher, USA). Dry lipid mixtures from the whole cell and PM enriched extractions were dissolved in a mixture of water, dichloromethane, and acetonitrile (2:2:1) with 10 mM ammonium acetate. The analytes were then separated on a UPLC C18 column (1.7 μm particle size, Waters, USA) at 318 K at a rate of 0.4 ml/min; an injection volume of 20 μl was used. Mobile phase A consisted of 40% acetonitrile and 60% water, mobile phase B consisted of 10% acetonitrile and 90% isopropanol, and both phases were supplemented with 10 mM ammonium acetate. Lipid separation was then achieved using a multi-step gradient from 40% of solvent B to 100% of solvent B in 17 min (full gradient provided in Table S1). Ions were monitored both in positive and negative Full MS/data-dependent Top 5 mode. The Full MS scan range was 300-2000 m/z with a resolution of 140,000 at m/z = 200. Data-dependent peaks were fragmented using normalized collision energy of 24, and the resulting MS2 spectrum was collected with a resolution of 17,500 at m/z = 200, an isolation window of 0.4 m/z, and the dynamic exclusion parameter set to "auto." Each run was repeated three times for each mode using dynamic exclusion of previously analyzed ions 7 . Three biologically independent samples of SH-SY5Y whole cell samples were analyzed three times for each mode (positive or negative), i.e., 18 LC-MS/MS runs in total. The PM has been analyzed with just two iterative exclusions, a total of 12 LC-MS/MS runs. ## Lipid analysis by LipMat scripts For MS/MS analysis we have built highly flexible, customizable scripts using Matlab 2017b software. Please cite this paper when using LipMat. ## Lipid fragmentation libraries We have searched the literature for experimental and theoretical 29,30 lipid fragmentation pattern of the phospholipid groups identified by 31 ## Peak and Intensity Scoring The precursor m/z values are compared with the intact masses of adducts from the libraries. For each instance where a hit is found, the script proceeds with a peak (Speak) and intensity (SIntesity) scoring for each hit. Total score (Stotal) is the sum of the negative log of Speak and the negative log of Sintensity. The peak score (Speak) provides the largest sum of the total hit score, and it composes of a hypergeometric distribution (Equation 1), used previously by other authors to describe mass spectrometric peak distributions 29 . This function reflects the probability that any overlap of theoretical and experimental MS2 peak is random. Equation 1: Where x is the number of hits, Kthe number of theoretical MS2 fragments, nthe number of observed MS2 fragments and Mthe size of scanning range divided into bins, where the number of bins is the size of the scanning range divided by the ppm error. The ppm error is set by the user, typically to 5 ppm. The second component of the final score is the intensity score (Equation 2). This is a minor part of the final score and reflects the intensity component of identified peaks 29 . Experimental MS2 data and the number of identified peaks are used as input for this value. The same number of peaks are randomly selected from the spectra and if the total intensity of the resulting random spectra is lower than the total intensity of identified peaks, one point is added to the intensity score ratio. The process is repeated for all possible combinations of randomly generated spectra, and the ratio is then used to calculate Sintensity. However, to save computing power, if there are more than six identified peaks in the spectra, only 500 random combinations of peaks are selected. Equation 2: Where x is the number of hits, kthe number of theoretical MS2 fragments, nthe number of observed MS2 fragments and i -the intensity of the identified fragments. ## Construction of lipid-specific elution profiles After the identification of possible lipid compositions, the LipMat scripts will start to scan the MS1 spectrum. First, it will filter out non-distinguishable lipid species. These are lipid species with the same m/z ± a user-determined defined ppm error, which are eluted in same retention time ± user determined retention time offset. These lipid species are written to the LipMat output file and subject to manual check by the expert user. From these non-distinguishable species, only ones with the highest score are processed further. The area of each defined and distinguishable lipid species in the MS1 spectra is then calculated. The abundance of each unique FA is calculated as a sum of the area for each FA occurrence. The MS1 chromatogram for up to five of the most abundant lipid species associated with each headgroup category and bar plots for up to five of the most abundant FA are plotted and written to the LipMat output files. ## Results and discussion It has been shown that the phospholipid composition in the eukaryotic cell is highly variable. It differs between organisms, cell types or between healthy and diseases states of a cell . The current state of knowledge regarding the lipidome of eukaryotic cells starts to be insufficient when specific cell types are considered, and is often completely absent in many cases. Recent reports concerning the importance of phospholipids in development and progress of PD has led to our current goal: To perform a mutually supportive NMR and LC-MS/MS phospholipid analysis of a cell model system commonly used in PD research, the SH-SY5Y neuroblastoma cell line. The phospholipid content of the whole cell was isolated using a method we have previously established 17 . For PM isolation, we have employed a modified method of sucrose gradient separation 18,19 (Figure S1). Upon assessing the purity of the PM fraction using Western blot, we observed contamination of calnexin (a marker for the endoplasmic reticulum 40,41 ). However, there was a much lower presence of β-actin (cytoskeletal marker 42 ) and almost no nucleophosmin (nuclear marker 43 , Figure S2). We therefore concluded that our approach provided us with PM-enriched samples. We then used 31 P NMR to analyse the phospholipid content of whole cell SH-SY5Y lipid isolates (Figure 1). In accordance with other studies, the most abundant phospholipid was PC. However, the abundance was higher than that reported for human erythrocytes (~30%) 34 or yeast (~15%) 44 . With 54.7 ± 1.9 % of total abundance, PC is by far the most common phospholipid in SH-SY5Y. This is also in agreement with lipidomics analysis of human plasma where PC showed up as the most abundant phospholipids. However, this study did not further explore the FA profiles of lipid species quantified 8 . We have observed some difference in the fraction of PE and PS when we compare SH-SY5Y cell to brain matter. For whole brain extracts, the reported content of PE and PS is ~30% of brain phospholipids 45-49 , yet, for SH-SY5Y the amount of PE and PS is 17.8 ± 4.8% and 4.7 ± 2.6% respectively. Another interesting result is the relatively low abundance of SMs (2.7 ± 1.2 %) when compared to CL (8.0 ± 1.4 %). The only source of CL in the eukaryotic cell is the mitochondria, while almost all of the SMs is present only in PM. This suggests that SH-SY5Y is especially energetically active. Moreover, based on the amount of SMs, SH-SY5Y appears to have a lower PM surface than differentiated neurons in vivo, as the only source of SMs in cells is PM. Both of these observations reflect the carcinogenic proprieties of the non-differentiated SH-SY5Y cell line. We also strived to secure PM-specific, quantitative data using NMR. However, after PM isolation from SH-SY5Y cell, we did not obtain sufficient quantities of lipids for complete 31 P NMR analysis. We have observed just three of the most abundant phospholipids in the PM-enriched fraction: PC, PE and SMs. Although not complete with respect to minor lipid species, they suggest that almost all SMs in SH-SY5Y is localized in the PM. It is also apparent and not unexpected that 31 P NMR, although potentially straightforward and informative, suffers from lack of sensitivity when little material is available. In our case, working with samples containing ~30 μg of lipids on a modern 600MHz instrument fitted with a cryogenically cooled probe did not detect lipids species of low relative abundance. Increasing the number of scans did not seem to affect this, suggesting that the weakest resonances were not detectable at any receiver gain setting. The detection limit of mass spectrometry is significantly lower than that of NMR 50 , and we therefore proceeded with LC-MS/MS analysis of the FA chains of all phospholipids identified by 31 P NMR. Prior to the main experiments, the liquid chromatography conditions were optimized to obtain the elution of phospholipid as narrow peaks. We used an iterative exclusions protocol 7 to filter out abundant ions, as well as background signals. Our custom-built LipMat script written in Matlab was then used for further processing (Figure 2). The LipMat identifies the presence of individual lipids by comparing MS2 fragmentation spectra with the literature and in silico prediction (see Experimental section for more information and Figure S3). The lipid species deemed robustly detected by using a scoring function (sum of Eq. 1 and Eq. 2, Experimental Section), were then selected for MS1 search of the intact lipid m/z (Table S2). Because of the high overlap of m/z in different lipid species, only the area of retention time, where the lipid species was identified, was searched in this way. An additional retention offset of 1 min was added, so it would be possible to observe complete elution of a given peak. This leads to a reconstruction of chromatograms (Figure S4) for the elution of individual lipid species and to relative quantification of FA within each headgroup (Figure S5). Using a combination of 31 P NMR and the LC-MS/MS method described herein, we have successfully identified phospholipid abundances, as well as FA chain composition for whole cell isolates (Figure 3) and plasma membrane enriched samples (Figure 4). The most abundant lipid, PC, was easily ionized and we had high detection rate for positive mode MS runs. The diversity of PC FAs was low, as almost 75% of all PC lipid in the SH-SY5Y cell are composed of two of the four FA chains (18:1, 16:0, 16:1, 18:0 or 14:0). Similar observations were also made for the PM-enriched isolate where 16:0 FAs were particularly abundant. We also noted that the 16:1 PC FA chain (17.4 ± 3.4 % in the whole cell sample) deviates from other reports of brain composition, where 16:1 FA accounts for under 3% of PC content 45,51,52 . SMs was the second most abundant lipid detected in positive mode. Even at lower concentrations, we could observe a diverse array of SMs hits. However, we were unable to identify exact SMs FA acid composition, because most observed SM lipid species lacked FA-specific fragments. Two lipid species that could be identified, were SM d18:1/24:1 and SM d18:1/24:0, and these were both observed in whole cell and PM fraction samples. In the whole cell samples, all the PE FA chains were 18:1, 18:0 or 16:0. However, in the PM-enriched samples we observed a higher variability of different PE species (Figure 4) and disproportional FA lengths. One PE FA had a length of 18:0, while others were composed of long, unsaturated chains (22:6, 20:4 or 20:1). The PG lipid abundance was too low to perform FA distribution analysis. However, four FA chains were identified repeatedly within the PG lipid headgroup (16:0, 18:1, 20:3 and 20:4). Both PI and PS were much worse ionizers, associated with lower levels of detection and we therefore detected only a few lipid species from each headgroup repeatedly, namely: PI 18:0/20:4, PI 18:0/20:3 and PS 18:0/18:1. However, we did not identify any docosahexaenoic acid FA (22:6) in the whole cell sample regardless of the fact that this FA is reported in high abundance (>25%) of brain matter for both PS and PE 45,51,52 . Yet, PE 22:6 was present in PM sample with an abundance of 15.9 ± 2.2 %. This suggests that all 22:6 PE FA are located in the plasma membrane and that the presence of this FA was masked by other, more abundant PE FA in the whole cell samples. This underscores the importance of sample fractionation in order to provide a complete and reliable picture of the cell lipidome. ## Conclusion We have analyzed the phospholipid composition of SH-SY5Y cells by NMR and LC-MS/MS using customizable and highly flexible LipMat script for Matlab software. Compared to the phospholipid composition of the brain, there exists notable differences which should be taken into account when SH-SY5Y is used as a model for studying neurodegeneration or brain lipid metabolism. The most striking differences are a relatively low abundance of PE and PS, higher occurrence of 16:1 PC FA and missing 22:6 FA for PS. The lipid fragmentation library is generated using user input specifying the length of FA chains and number of double bonds. The analysis star with comparing intact ion masses with the library. In the case of a hit, MS2 spectrum is loaded and scored based on the presence of corresponding lipid fragments and their intensity. Afterward, retention time is analyzed, and MS1 spectrum is searched for elution peak of assigned lipid species. Generated outputs include lipid fragmentation figures, chromatogram plots and table output with identified lipid species and their respective scoring. were taken into account. For SMs we were unable to identify relative FA chain abundance.

chemsum

{"title": "Fast and quantitative phospholipidomic analysis of SH-SY5Y neuroblastoma cell cultures using LC-MS/MS and 31 P NMR", "journal": "ChemRxiv"}

simple_surface_modification_of_poly(dimethylsiloxane)_via_surface_segregating_smart_polymers_for_bio

## Abstract: Poly(dimethylsiloxane) (PDMS) is likely the most popular material for microfluidic devices in lab-on-achip and other biomedical applications. However, the hydrophobicity of PDMS leads to non-specific adsorption of proteins and other molecules such as therapeutic drugs, limiting its broader use. Here, we introduce a simple method for preparing PDMS materials to improve hydrophilicity and decrease nonspecific protein adsorption while retaining cellular biocompatibility, transparency, and good mechanical properties without the need for any post-cure surface treatment. This approach utilizes smart copolymers comprised of poly(ethylene glycol) (PEG) and PDMS segments (PDMS-PEG) that, when blended with PDMS during device manufacture, spontaneously segregate to surfaces in contact with aqueous solutions and reduce the hydrophobicity without any added manufacturing steps. PDMS-PEGmodified PDMS samples showed contact angles as low as 23.6° ± 1° and retained this hydrophilicity for at least twenty months. Their improved wettability was confirmed using capillary flow experiments. Modified devices exhibited considerably reduced non-specific adsorption of albumin, lysozyme, and immunoglobulin G. The modified PDMS was biocompatible, displaying no adverse effects when used in a simple liver-on-a-chip model using primary rat hepatocytes. This PDMS modification method can be further applied in analytical separations, biosensing, cell studies, and drug-related studies.The microfluidics industry encompasses a $2-4 billion market 1,2 , expected to grow by ~18%/year to $10-20 billion by the 2020s 1 . Academic interest in this field is growing at a similarly fast pace, with the number of publications on microfluidics doubling every 15 months 2 . This growth is driven mainly by biomicrofluidics such as point of care devices, drug manufacturing micro-reactors, toxicity screening with organs-on-chips, and microneedles/ pumps for drug delivery 3 . However, choosing the right materials is critical for avoiding artefacts and reduced sensitivity in biomedical and diagnostic applications, including those that can arise from the adsorption of compounds of interest onto surfaces.Poly(dimethylsiloxane) (PDMS) and other silicone elastomers offer a range of favorable properties for biomicrofluidics applications, including: (1) simple fabrication by replica molding, (2) good mechanical properties, (3) excellent optical transparency from 240 to 1100 nm, (4) biocompatibility and non-toxicity, and (5) high gas permeability 4 . Despite these merits, the hydrophobicity of PDMS (water contact angle ~108° ± 7°) 5 often limits its applications where solutions comprising of biological samples are concerned. The hydrophobicity of the PDMS surface results in undesired non-specific adsorption of proteins, which in turn affects analyte transport and reduces separation performance and detection sensitivity 6 . The hydrophobicity of PDMS microchannels also makes it difficult to introduce aqueous solutions or mixtures of aqueous and organic solutions 7 . Since most of the work in microfluidics relies on using polar liquids, this causes a significant obstacle in many applications. This has led many groups to develop approaches to render the PDMS surface hydrophilic and resistant to protein adsorption 8 . These strategies include the use of high-energy treatments in the form of O 2 plasma, UV/ozone treatments, and corona discharges to oxidize PDMS surfaces and to introduce alkoxy-or chloro-silanes for surface functionalization later on, coating of PDMS surfaces with polar functionalities using charged surfactants, polyelectrolyte multilayers (PEMs), chemical vapor deposition, silanization, phospholipid bilayers, and more recently, by attaching hydrophilic polymer brushes to the surface of PDMS via grafting-from and grafting-to approaches, hydrosilylation and click chemistry 8 . While these interventions have proved successful in improving surface hydrophilicity, their broader use was often limited by chemical stability, the need for special equipment and/or hazardous routes 9 , and/or the length and complexity of their process for fabrication that is restrictive for large-scale implementation. In addition, many of these methods lead to loss of transparency, change in mechanical properties, surface cracking and increased roughness 10,11 . Finally, most of these methods do not provide a hydrophilic surface long term. Due to the mobility of PDMS chains, the surface becomes hydrophobic again over time, negating the initial benefits of treatment 12,13 (Table 1). These issues curtail the benefits of these PDMS surface modification methods and emphasize the need for a new approach. An alternative approach for creating more hydrophilic and fouling-resistant surfaces involves the use of surface-segregating smart copolymers. In this approach, an amphiphilic copolymer additive is blended with the base polymer before the manufacture of the final component. The hydrophilic sections of the copolymer drive it to the polymer/water interface, leading to surface segregation. When successful, this results in increased surface hydrophilicity, but only minor changes in bulk properties. This approach has been previously used in other fields and base materials. For instance, it enabled the preparation of filtration membranes with excellent, complete fouling resistance made of polyacrylonitrile (PAN) and poly(vinylidene fluoride) (PVDF) 17 . It was also used to prevent non-specific adsorption and cell adhesion on poly(methyl methacrylate) (PMMA) surfaces 18,19 . Similarly, the use of amphiphilic or hydrophilic additives to PDMS during the manufacture of devices can lead to improved hydrophilicity. This approach is simple, often requiring no additional steps. If designed well, it can potentially lead to mechanical and optical properties similar to unmodified PDMS. Yet, to our knowledge, there are only a few studies that have focused on functionalizing the PDMS surface through a pre-mixing method, where functional additives are added to the liquid PDMS pre-polymer before curing. In some cases, the objective of such studies was not to improve hydrophilicity but to introduce specific functional groups on the surface. For example, Zare et al. 24 added a biotinylated phospholipid to PDMS prepolymer to enable protein immobilization. Another study introduced charged groups to PDMS microfluidic channels by adding undecylenic acid to the pre-polymer prior to curing 25 . This led to increased electroosmotic flow (EOF) in PDMS microchannels, improving the separation efficiency and reducing the peak broadening in PDMS microfluidic devices. In both studies, the use of the additive did not lead to any changes in surface hydrophobicity. Other researchers have tested additives to improve surface hydrophilicity (Table 1). For instance, Zhou et al. added vinyl-terminated polyethylene glycol (PEG) chains to PDMS before curing 22 , showing a slight decrease in the water contact angle (WCA) from 112° to about 78°, accompanied by improved resistance to non-specific adsorption of a protein. In another study, PDMS microchips were prepared via adding a poly(lactic acid)-poly(ethylene glycol) (PLA-PEG) amphiphilic diblock copolymer before curing 20 . These microchips exhibited reduced myoglobin adsorption, and slightly lower WCAs of 84° and 73° for 1.5 and 2% mass ratio of PLA-PEG to PDMS, respectively. The amphiphilic triblock copolymer Pluronic (PEG-b-poly(propylene oxide)-b-PEG) was also used as a similar additive 21 . Upon filling the PDMS microfluidic channel with water, Pluronic embedded in PDMS segregated towards the water/PDMS interface. The static contact angle of modified PDMS surface changed from 98.6° to 63° after soaking the sample in water for 24 hours, whereas that of the additive-free PDMS remained around 103°. Furthermore, thanks to the improved hydrophilicity, the modified surface exhibited lower non-specific adsorption of Immunoglobulin G (IgG) compared to native PDMS. However, the limited compatibility of the hydrophobic poly(propylene oxide) segments with PDMS can limit the success of this approach. Indeed, the researchers observed samples became cloudy with as little as 0.16% Pluronic. Furthermore, the Pluronic surfactant is water soluble, which led to some leaching during use. This may lead to the degradation of surface hydrophilicity in time, and affect cell viability. An alternative copolymer additive that has better compatibility with PDMS and can be integrated into the PDMS network during the preparation of the microchip would be beneficial. There is only one preliminary study www.nature.com/scientificreports www.nature.com/scientificreports/ that utilizes a PDMS-based additive, a PDMS-PEG block copolymer. This study shows that the addition of this PDMS-PEG block copolymer improved surface hydrophilicity of PDMS when used at concentrations between 1-1.9% (w/w), reducing the contact angle to 21.5-80.9°2 3 . These results indicate that PDMS-PEG copolymers may be a promising initial direction for improving the surface properties of PDMS, motivating the work described here. However, this study also leaves a lot of questions open regarding the effectiveness of PDMS-PEG additives in biomicrofluidic applications, and their true performance gains. Importantly, both this study and most others in the literature study surfaces prepared upon blending PDMS with the additive but do not account for the effect of common processes used in the manufacture of actual microfluidic devices such as plasma treatment on the eventual surface chemistry. This is a crucial short-coming for understanding the applicability of this work to practical systems. Furthermore, the work mentions compromised mechanical properties at high concentrations of this additive but does not quantify it. Importantly, no studies that utilize PDMS-based additives characterize non-specific protein adsorption, or how the surface hydrophilicity changes with time or upon exposure to common processes used in manufacturing microfluidic devices. Finally, none of the studies that focus on surface modification using additives, PDMS-based or not, tested their materials for biocompatibility. It should be noted that these additives are typically surfactants that are water-soluble and can cause cell rupture. Therefore, it is crucial to test any such new approaches for biocompatibility to ensure its usability in realistic systems in contact with cells, such as organs-on-chips. In this study, we focus on a practical and simple approach to improve the hydrophilicity of PDMS surfaces by adding a PDMS-PEG block copolymer (BCP) to the PDMS prepolymer before curing at concentrations between 0.25-2%, with the rest of the device manufacture process being conducted with no further changes, enabling this surface modification approach to be directly plugged into existing protocols. While a similar copolymer was previously used to improve the hydrophilicity of PDMS surfaces 23 , the experiments reported did not accurately address several key questions relevant to optimizing the overall biomicrofluidic device manufacturing process or the use of these devices in realistic applications. These results also implied compromised optical and mechanical properties at additive concentrations needed to improve hydrophilicity. Here, we seek to holistically study the use of a similar PDMS-PEG copolymer as an additive, with a focus on building a comprehensive understanding of how the overall device manufacture process, including alcohol soak and plasma treatment steps and exposure to cells, affects the surface chemistry and performance of devices prepared by this approach. We also seek to understand the capabilities of this approach in biomicrofluidic device applications, including the creation of devices with high optical clarity and mechanical properties. Importantly, we show that by tuning these manufacturing parameters and leveraging manufacturing steps already used for biomicrofluidic devices, we can achieve significantly enhanced hydrophilicity that is stable over at least 20 months, longer than all past reports, including those using more complex methods (Table 1). Compared to other additives that have been explored to date , the utilization of this PDMS-PEG block copolymer provides better compatibility between the additive and PDMS, keeping the device optically clear at concentrations up to 0.25%. Through dynamic water contact angle (WCA) measurements, we show the PDMS-PEG copolymer segregates to the surface when exposed to water/ aqueous solutions, which renders the surface more hydrophilic than all past studies using additives (Table 1). This approach also reduces non-specific adsorption of proteins (albumin, lysozyme and immunoglobulin G), as indicated by both fluorescent protein adsorption experiments on slabs and by quantitative experiments on fabricated microfluidic devices. The PDMS monolith and PDMS segments in block copolymers interact through van der Waals and hydrophobic interactions that improve the stability of the PEG layer on the PDMS surface 26 . Furthermore, the PDMS chains in the BCP can potentially be cross-linked with the chains of the monolith during the plasma treatment stage, further improving the stability of the hydrophilic surface. Indeed, we show that the hydrophilicity of PDMS modified with this copolymer is retained for at least twenty months, longer than all past reports, even after exposure to isopropanol (IPA) soaking and plasma treatment, crucial manufacturing steps that were not considered in previous studies. Mechanical properties are preserved at PDMS-PEG concentrations up to 1.0%, whereas optical clarity is retained at concentrations up to 0.25%. Unlike previous publications, this is the first report where the biocompatibility of PDMS modified with PDMS-PEG BCP was tested by culturing primary rat hepatocytes in glass-(PDMS-PEG BCP modified) PDMS microfluidic tissue culture devices. The PDMS-PEG modified devices performed just as well as unmodified PDMS devices and presented no adverse effects. These results demonstrate that the addition of this PDMS-PEG BCP to PDMS before the manufacture and curing of biomicrofluidic devices results in a durable increase in hydrophilicity and resistance to non-specific adsorption without sacrificing mechanical properties, optical clarity, or biocompatibility. Therefore, this method promises to be a very simple, rapid, and cost-effective approach to generate hydrophilic and protein repellent PDMS elastomer for microfluidic devices as well as other uses such as tubing and sealants. ## Results and Discussion Surface modification of PDMS with PDMS-PEG BCP additives. We selected a PDMS-PEG BCP as the smart copolymer additive for hydrophilizing the PDMS surface. This copolymer, a commercially available surfactant (Gelest, product code DBE-712), includes a hydrophobic PDMS segment compatible with the base elastomer (e.g. PDMS) and a hydrophilic, fouling resistant PEG block. Its molar mass is 600, and contains 60-70% PEG. The PDMS segment solubilizes the additive within the elastomer matrix during preparation and then anchors the additive in the cured PDMS. It can also be linked with the base PDMS during the plasma treatment used for bonding the device together, improving the longevity of the surface modification. The short chain length and BCP architecture of the additive leads to its segregation to the sample surface 18,27 . When the sample surface is exposed to water (e.g. when the microfluidic channel is filled with aqueous media), the copolymer self-organizes at the PDMS/water interface to expose the PEG segments to the aqueous solution and create a stable hydrophilic surface that prevents the adsorption of proteins and other bio-macromolecules (Fig. 1) without using any additional steps or changing the manufacturing process. To date, most approaches, to reduce hydrophobicity of PDMS, were developed using post-treatment methods 8,23, that add several new, cumbersome steps to the micro-device manufacturing process, often requiring special equipment and/or hazardous routes 9 . This renders them unfavorable for large-scale fabrication and for adoption by a wide user base. Further, they cannot be adapted to the manufacture of other silicone-based elastomers (e.g. tubing, seals). This limits their impact. The approach we present here is differentiated by its simplicity during use, compared with other approaches that rely on coatings or post-processing (Table 1). While a handful of studies have utilized surface-segregating amphiphilic copolymers to improve the surface hydrophilicity of PDMS, none have demonstrated high degrees of hydrophilicity without loss in mechanical properties and/or optical clarity (Table 1). Furthermore, these studies have almost exclusively focused on characterizing surfaces that have not been subjected to the full slew of processes involved in microfluidic device manufacture, including an alcohol soak for disinfection and plasma treatment for bonding of the device. These processes can leach these additives and/or significantly alter surface chemistry. In addition, none of the past studies characterize the viability of cells upon exposure to these PDMS blends. These amphiphilic additives may leach into the feed going through a biomicrofluidic device, killing cells and thus rendering these approaches moot in practical settings. Thus, there is a significant knowledge gap in not only developing novel additives for PDMS for surface modification but also in better understanding their behavior throughout the life cycle of a biomicrofluidic device. ## Hydrophilicity and wettability of PDMS with PDMS-PEG BCP additives. To test our hypothesis that the PDMS-PEG BCP additive would lead to increased hydrophilicity that remains stable over long timescales, we measured sessile drop water contact angles (WCA) on PDMS-PEG BCP modified PDMS surfaces and compared them to the unmodified PDMS over a 20-month duration. We used dynamic contact angle measurements, which are useful for evaluating the wettability and hydrophilicity of modified PDMS surfaces 32,33 . Figure 2a shows the variation of the WCA of PDMS samples prepared with varying amounts of PDMS-PEG BCP additive in time. The initial contact angles of all samples (except the one containing 2% PDMS-PEG BCP additive) were quite high, between 94-106°. This indicates that in air, the sample surface is mostly covered with hydrophobic PDMS segments. However, while the WCA of PDMS with no PDMS-PEG BCP remained steady above 101° during the 45-minute experiments, the WCA of all PDMS with PDMS-PEG BCP additives decreased in time. Furthermore, this decrease was generally proportional to the concentration of PDMS-PEG BCP additives. After 45 minutes of exposure to water, the PDMS-PEG BCP additive containing sample surfaces became significantly more hydrophilic than additive-free PDMS. As little as 0.125% PDMS-PEG BCP additive led to a final contact angle of 69.6° (Supporting Information Fig. S1), comparable with the lowest contact angles reported for other additive-modified PDMS systems . The highest BCP containing samples (1.5% and 2% PDMS-PEG BCP) were fully wetted (WCA ≈ 0°) in our dynamic measurements. Nevertheless, it is important to note that increasing BCP concentration for reducing hydrophobicity is not the only requirement for successful and stable surface modification. We www.nature.com/scientificreports www.nature.com/scientificreports/ encountered bonding problems on glass slides during oxygen plasma treatment at higher copolymer concentrations (1.5 and 2 (w/w %)), so we eliminated these concentrations for further experiments. These results confirm that upon exposure to water, the PDMS-PEG BCP additive self-assembles at the interface to create a hydrophilic PEG layer, and indicate that this rearrangement occurs faster and more effectively with increasing PDMS-PEG BCP content. Furthermore, this approach can lead to final WCA values much below previous reports for additive-modified PDMS materials, which range between 84° and 63°2 0, 21 . In practical applications, PDMS is not used directly after molding. PDMS devices are typically sterilized by immersion into an alcohol such as IPA, which may leach out additives. Then, they are treated with O 2 plasma and bonded to glass, a silicon wafer, or another piece of PDMS. It is critical for the improved surface hydrophilicity to be stable during these processing steps. Furthermore, microfluidic devices are not necessarily used directly after manufacture. Therefore, the modified surface needs to be stable over long time periods. To test these parameters, we first established the soaking time of PDMS with and without PDMS-PEG BCP additives in IPA. We measured the WCA of PDMS without PDMS-PEG and of PDMS with 0.5% PDMS-PEG BCP additive after soaking in IPA for 6, 12 and 24 hours (Supporting Information Fig. S2). The hydrophilicity of samples after 6 hours of IPA soaking was higher compared with that of samples with 12 hours and 24 hours of soaking. We believe that during soaking in IPA, the lower molar mass fractions of the PDMS-PEG BCP diffused out 21 . This resulted in a decrease in copolymer concentration in the PDMS and a significant increase in the contact angle. We observed that 12 hours of IPA soaking was sufficient to remove all the lower mass fractions of the BCP because no significant change in hydrophilicity was observed between 12 hours and 24 hours IPA soaking. Still, some PDMS-PEG BCP remained in the PDMS as the final contact angles were still much lower than that of PDMS with no PDMS-PEG BCP. The remaining PDMS-PEG BCP molecules were likely of higher molar mass, which improved the long-term stability of the layer. Furthermore, this decreased the risk of the additive, particularly low molar mass fractions likely to act as cytotoxic surfactants, leaching out of the PDMS during operation, which could negatively impact cell viability. Nonetheless, we selected 24 h IPA soaking for experimental practicality and to ensure consistent results in further experiments (samples labeled AS). The improved surface hydrophilicity of IPA-soaked samples was stable for at least 20 months (Fig. S3). Some of the samples were then treated with O 2 plasma (samples labeled AS + PT) (Fig. 2b). A day after O 2 plasma treatment (AS + PT 1 d), the hydrophobicities of PDMS samples both with and without PDMS-PEG BCP additives were significantly reduced (Fig. 2b). The WCA of PDMS with no PDMS-PEG BCP became 63.3°. www.nature.com/scientificreports www.nature.com/scientificreports/ It has previously been reported that PDMS surfaces exposed to plasma exhibit increased oxygen content and possibly silicon (Si) atoms are bonded to three or four oxygen atoms and that this reduces hyrophobicity 34 . The main challenge posed by the plasma oxidation is the eventual hydrophobic recovery. This is a result of the reorientation of pre-existing oligomers from the bulk to the surface 35 . Indeed, 3 days after plasma treatment (AS + PT 3 d), the WCA of the PDMS with no PDMS-PEG BCP additive returned to its initial value of 102°. PDMS with PDMS-PEG BCP additives also exhibited an increase in hydrophilicity upon O 2 plasma treatment. The surfaces became fully wettable a day after plasma treatment with WCA values around 0°. As seen in Fig. 2b, although a minor increase in wettability was observed, all PDMS with PDMS-PEG BCP additives maintained their hydrophilicity, with WCA values between 54.2° ± 2.7° and 25.7° ± 2°. These values are significantly lower than previous reports . We believe that the existence of PEG on the modified PDMS surface enhances Si-O bonding and as a result, more SiO x -rich layer and more hydrophilic surfaces can be obtained as compared to PDMS with no PDMS-PEG BCP. Importantly, this enhanced surface hydrophilicity was stable for at least twenty months. The increased surface hydrophilicity may have enhanced the surface segregation of the PDMS-PEG BCP by creating a local gradient, drawing the copolymer to the surface even before exposure to water. The increased degree and stability of surface hydrophilicity may also be linked with the complex and competing etching, deposition and reaction processes that occur during plasma treatment. During the O 2 plasma treatment, PDMS repeat units are partially etched on the surface, losing their methyl groups and forming silica. The plasma treatment may also cause cross-linking, but this effect is relatively limited in PDMS 36 . In contrast, oxygen-containing polymers such as PEO tend to undergo atomic re-arrangement reactions such as cross-linking as opposed to etching 37 . This implies that the plasma treatment may preferentially etch the hydrophobic methyl groups from PDMS chains on the surface, exposing PEG segments that were right below. The plasma treatment can also chemically cross-link the PDMS-PEG BCP additive to the PDMS network. Furthermore, it may lead to cross-linking between PEO chains on the surface. This may anchor the PDMS-PEG BCP specifically on the top surface of the sample, improving the longevity of surface modification. Additionally, PDMS samples with different PDMS-PEG BCP preserved their hydrophilic characteristics even after 20 months of storage (with/without plasma treatment) indicating that samples prepared with PDMS-PEG BCP additives are stable for a long period (Fig. 2c). ## Characterization of the physical properties of PDMS with PDMS-PEG BCP additives. Transparency. Microfluidic devices are commonly used together with bright field and fluorescence microscopy for imaging cells 38 to monitor their health and motility. Therefore, materials used for manufacturing such devices must be transparent. Blue light [460-500 nm] excitation is commonly used to image green fluorescent protein (GFP) and Calcein AM, and green light [528-553 nm] excitation is useful for imaging red fluorophores 38 . Accordingly, we assessed the optical clarity of PDMS with and without PDMS-PEG BCP additives by measuring light transmittance through 8 mm thick slabs between 400-600 nm wavelengths in the UV-visible range before and after an IPA soak (Supporting Information Fig. S4) after the fabrication. Transmittance values for the center wavelengths of blue light (480 nm) and green light (540 nm) are given in Table 2. Before soaking in IPA (Fig. S4(a)), transparency values for PDMS with up to 0.5% PDMS-PEG BCP were comparable to additive free PDMS, with all transmittance values above 96%. The transparency of the PDMS sample with 1% PDMS-PEG was slightly lower, with transmittance values in the 80-88% range. This may arise from the formation of micelles or similar aggregates of the PDMS-PEG BCP surfactant within the bulk PDMS at these higher concentrations, as observed in other studies 20,21 . After soaking in IPA (Fig. S4(b)), samples modified with 0.125% and 0.25% PDMS-PEG BCP additives exhibited approximately the same optical clarity as unmodified PDMS. However, the optical clarity of modified PDMS with 0.5% and 1% PDMS-PEG BCP decreased, with transmittance values around 75% and 50%, respectively. PDMS samples containing 0.25% PDMS-PEG BCP successfully combined high optical clarity with a hydrophilic surface. Surface Characterization. We used X-ray photoelectron spectroscopy (XPS) to gain a better understanding of the changes in surface chemistry during the manufacture of biomicrofluidic devices from PDMS with and without the PDMS-PEG BCP additive. In this study, we focused on PDMS with 0.25% PDMS-PEG BCP, selected according to the criteria described above, and PDMS with no PDMS-PEG BCP. We analyzed their surface chemistry at each stage of the microfluidic device manufacture process. The elemental surface compositions of both the PDMS with no PDMS-PEG and 0.25% PDMS-PEG, determined by the survey scan, remained essentially www.nature.com/scientificreports www.nature.com/scientificreports/ unchanged after soaking in IPA (Supporting Information Fig. S5). After plasma treatment, survey scans of PDMS with and without PDMS-PEG BCP additives both indicate an increase in carbon and oxygen content and a corresponding decrease in silicon content (Supporting Information Fig. S6). High-resolution scans of C1s spectra were used to gain deeper insight into the chemical changes that occurred during these processes (Fig. 3(a,b)). PDMS with no PDMS-PEG spectra featured only one peak, at 284.2 eV, corresponding to C-Si bonds. This was unchanged upon soaking in IPA. Upon plasma treatment, peaks appeared at 286.3 eV and 289.1 eV, assigned to C-O and C=O bonds, respectively 39 . These peaks, however, completely disappeared after a week. This is due to the recovery of hydrophobicity after oxidation by reorientation of the surface silanol groups into the bulk polymer, which provides for the movement of free PDMS chains from the bulk phase to the surface and condensation of silanol groups at the surface 8 . In contrast, PDMS with 0.25% PDMS-PEG BCP showed both a strong peak near 284.6 eV arising from C-Si bonds and a shoulder near 286.3 eV, corresponding to C-O bonds, indicating the presence of PEG segments from the PDMS-PEG BCP additive near the surface. Upon plasma treatment, the intensity of the C-O peak increased, and a new peak corresponding to C=O appeared. These groups may arise both from reactions of PDMS and from reactions and cross-linking of PEG. Unlike pure PDMS, the intensity of the C-O and C=O peaks in the PDMS with 0.25% PDMS-PEG remained unchanged a week after plasma treatment. This phenomenon confirms the existence of PEG molecules on the modified surface for long-term stability after plasma treatment, which is in good agreement with the hydrophilicity data (Fig. 2a,b). Mechanical properties. PDMS is a good candidate for use in microfluidic devices due to its high compliance and flexibility. Its Young's modulus depends on the exact formulation, and is around ∼1.32-2.12 MPa for the commonly used pre-polymer to curing agent ratio of 10:1 . Ideally, surface modification by any approach should not compromise these mechanical properties. To check the mechanical properties of PDMS-PEG-modified PDMS samples, tensile strength and compressive modulus were evaluated by dynamic mechanical analysis (DMA) immediately after fabrication and also 20 months after fabrication. Young's modulus and compressive modulus of the modified samples were calculated for the linear elastic region (<40% strain). No significant change was observed with the mechanical properties of the PDMS-PEG BCP modified PDMS when compared with literature studies even after 20 months of storage. ## Biocompatibility. Many microfluidic applications that utilize PDMS and its alternatives involve culture or circulation of cells from different tissues. Therefore, when designing a new material for biomicrofluidics, it is crucial to take its biocompatibility into account. For instance, the surface modifying additives may leach from the device into the microfluidic channel and affect cell viability and/or function. This may lead to poor device performance even if surface hydrophilicity is enhanced. To date, there are some studies that evaluated the biocompatibility or cell adhesion of modified PDMS microfluidic devices or slabs using mammalian A549 cells 43 , L929 mouse fibroblasts 44 , tendon stem cells 45 , mesenchymal stem cells (MSCs) 46 , brain cerebral cortex cells 47 , HeLa cells 48 , and stroma cells 49 . To our knowledge, none of the previous PDMS modification strategies were evaluated for compatibility with hepatocytes, the parenchymal cells of the liver, which are highly susceptible to adverse reactions. The liver plays a central role in drug metabolism and detoxification so the development of liver-on-a-chip models for successful prediction of toxic response is at the center of the recent initiatives towards in vitro human clinical trial approaches 50,51 . Here, we used rat primary hepatocytes to test the biocompatibility of our modified PDMS substrate in a simple microfluidic liver-on-a-chip model. To ensure that the use of the PDMS-PEG BCP in microfluidic device manufacture does not adversely impact cell function, we manufactured microfluidic devices using a glass bottom and PDMS top with or without PDMS-PEG BCP additives, and cultured primary rat hepatocytes in these devices. In order to quantitatively evaluate cell viability, the cells were stained with a live (green)/dead (red) stain 3 days after the culture (Fig. 4). Cells had high viability (>99.0%) throughout the 3 day culture period following the initial cell seeding into the microdevice. The use of the PDMS-PEG BCP additive led to no visible or significant differences in cell viability or morphology. PDMS-PEG modified microfluidic devices performed just as well as PDMS with no PDMS-PEG additives and presented no adverse effects. Since in vitro systems are often preferred as models to predict drug toxicity and pharmacokinetics for clinical cases, this design can be easily scaled to create an array of in vitro studies for rapid drug development or studying the toxicity of drugs due to the simplicity of the device. ## Protein adsorption on PDMS with PDMS-PEG BCP additives. The main goal of developing this PDMS surface modification approach was to create a fouling resistant surface and prevent the non-specific adsorption of proteins onto the microfluidic device. This is motivated by two phenomena. First, most of the undesired bioreactions and bio-responses in artificial materials are promoted due to adsorbed proteins 52,53 . Second, many applications of biomicrofluidics involve controlling the exposure of cells to a known concentration of a specific, desired protein such as a biologic drug. Non-specific adsorption leads to the loss of this drug through adsorption, exposing the cells to a lower concentration than presumed. This can lead to a severe underestimation of the toxicity and activity of such drugs. While hydrophilicity is broadly correlated with decreased protein adsorption, the relationship is not necessarily straightforward 54,55 . Therefore, we quantitatively measured the adsorption of two fluorescently-labeled proteins, albumin and lysozyme, on PDMS slabs with and without PDMS-PEG BCP additives (Fig. 5(a,b)), both directly upon manufacture (Fig. 5a) and following processes that simulate biomicrofluidic device manufacture (Fig. 5b, IPA soak and 1 week after O 2 plasma treatment). PDMS with no PDMS-PEG BCP adsorbed significantly more protein than all PDMS with PDMS-PEG-BCP additives, confirming that this approach led to decreased non-specific adsorption (Fig. 5a). PDMS with 0.125% PDMS-PEG BCP exhibited some protein adsorption. No adsorption was visible for any of the other samples. The same trend continued following soaking in IPA and O 2 plasma treatment (Fig. 5b). PDMS slabs with PDMS-PEG BCP additives indicated substantially reduced adsorption as compared to additive free PDMS. To further quantify protein adsorption in a more realistic setting for biomicrofluidic device applications, we manufactured microfluidic devices from PDMS with or without PDMS-PEG additives. We then introduced a protein solution containing 0.05 mg/mL BSA, lysozyme or IgG into the microchannel (30-90 min), and measured the loss of protein due to adsorption on the device by micro-BCA analysis (Fig. 5c). Devices with PDMS-PEG additives adsorbed significantly lower quantities of each protein as compared to PDMS with no PDMS-PEG (Fig. 5c). As the BCP concentration increased in the mixture, the amount of adsorbed protein decreased. PDMS-PEG BCP additives significantly reduced protein adsorption at concentrations as low as 0.125% (w/w). The use of only 1% PDMS-PEG additive led to 98.9%, 89.4%, and 99.6% lower adsorption of albumin, lysozyme, and IgG, respectively when compared to PDMS without PDMS-PEG BCP. An additive concentration of 0.25% PDMS-PEG led to a ~90% the reduction in protein adsorption while also retaining excellent optical clarity. ## Capillary-driven microfluidic devices with PDMS-PEG BCP additives. Having demonstrated the successful hydrophilization of PDMS with PDMS-PEG BCP additives, we investigated the flow characteristic of PDMS with and without PDMS-PEG BCP additives (0.25% and 0.5%) in the capillary microchannels which were bonded on the glass substrates. Two linear channels (height: 0.1 mm, length: 40 mm) with different widths (0.25 mm and 0.5 mm) were tested for capillary-driven flow experiments. PDMS with no PDMS-PEG BCP was utilized as a control. All samples were tested 3 days after plasma treatment. Liquid was introduced into the inlet of the capillary channel and fluid flow through the channel was recorded by a camera to calculate the experimental flow rates. Table 3 and Fig. S7 show the variation of flow velocities of liquid using PDMS samples with varying amounts of PDMS-PEG BCP. All modified devices were shown to fill through steady capillary action while PDMS without PDMS-PEG BCP www.nature.com/scientificreports www.nature.com/scientificreports/ failed to fill with liquid. We did not observe a significant difference in capillary flow rates through the 0.25% and 0.5% PDMS-PEG BCP modified samples. This was consistent with the WCA results, which showed very similar hydrophilicity (Fig. 2). The advantage of the fabrication technique presented here is that hydrophilic PDMS microfluidic channels can be obtained with a simple, one-step method through inexpensive bench-top methods. We finally evaluated our results to choose the most preferred PDMS-PEG BCP concentration for a given application. As the biocompatibility and mechanical properties of all PDMS samples with PDMS-PEG BCP additives are almost identical with PDMS, we compared the different samples for their transparency, WCA after plasma treatment (t = 45 min) and protein (IgG as a sample protein) adsorption data. We selected 0.25% PDMS-PEG BCP concentration (WCA = 25.7° ± 2°, transmittance = 99%, reduction in IgG adsorption relative to PDMS with no additive = 92.2%) to be the best performing composition for applications where optical clarity is of importance, since the transparency of the modified samples decreased down to 73% and below after IPA soaking with BCP concentrations at or above 0.5%. ## Conclusion This manuscript introduces a simple approach to address non-specific protein adsorption, a key problem encountered in the use of PDMS in biomicrofluidic applications, without making any changes to the existing workflow for manufacturing such devices. This involves simply adding a PDMS-PEG BCP additive to PDMS during device manufacture. This BCP segregates to the surface during device manufacture and rearranges to create a hydrophilic surface upon exposure to aqueous media. As little as 0.25% additive leads to contact angles as low as 31.4° ± 1.5, whereas 2% additive leads to a fully wettable (WCA ≈ 0) surface. Surface hydrophilicity is retained through common processes used in microfluidic device manufacture (e.g. immersion in IPA and plasma treatment), and after prolonged storage at the bench top for at least 20 months. The extent and durability of surface hydrophilicity obtained by this method surpass others reported in the literature . Only 0.25% PDMS-PEG additive leads to ~90% reduction in the adsorption of three proteins, whereas 1% additive led to 89-99.6% reduction in protein adsorption, comparable to or better than the highest reductions in protein adsorption in the literature 56 . While we did not test protein adsorption resistance after extensive storage, the long-term stability of surface hydrophilization implies that these devices will likely exhibit reduced non-specific adsorption for long time periods. Furthermore, devices prepared with this approach preserve their transparency, flexibility, and biocompatibility with primary rat hepatocytes. According to all results, 0.25% (w/w) copolymer concentration was selected as an optimum value for applications requiring high transparency, whereas 1% additive led to samples with the lowest fouling while preserving mechanical properties. The PDMS modification method introduced here does not require any additional steps or equipment for device fabrication. This allows easy adoption and scale-up and is more compatible with mass production of microfluidic devices compared to silicon, glass or thermoplastic alternatives. It is our opinion that this method has a potential for applications including drug-related studies, analytical separations, biosensing, cell targeting, and isolation. Apart from the applications in microfluidics, we expect our invention to remove barriers that currently prevent the use of PDMS in critical commercial applications such as those in applications in pharmaceutical and biomedical industries. ## Microfluidic device fabrication for cell culture studies and protein adsorption experiments. Silicon wafer templates served as negative molds to fabricate microfluidic devices using PDMS, (Sylgard 184, Dow Corning, Tewksbury, MA) with and without PDMS-PEG BCP additives and utilizing standard soft lithography protocols. The microfluidic platform consisted of media fluid inlet/outlet and cell inlet/outlet in the same place, and a cell culture chamber. The dimensions of the chamber were 11 mm 2 × 0.1 mm (Surface area x height). Inlet and outlet ports of the device were punched into the PDMS microfluidic device using a 1.5 mm biopsy punch piercing tool (Ted Pella Inc.). The face of the PDMS with microchannel and a glass microscope slides (75 × 25 mm, Thermo scientific) were bonded with O 2 plasma (80 W, 35 sec) using a vacuum plasma cleaner. Capillary-driven microfluidic device fabrication. Capillary-driven microfluidic devices were fabricated with and without PDMS-PEG BCP using replica molding on silicon wafer templates as discussed above. Two linear microfluidic channel designs consisted of media fluid inlet/outlet were fabricated with varied geometries (0.25 mm, 0.5 mm widths, 0.1 mm height, and 40 mm length). Inlet and outlet ports of the microfluidic devices were punched using a 3.5 mm biopsy punch piercing tool (Ted Pella Inc.). The devices were then bonded to glass microscope slides (75 × 25 mm, Thermo Scientific) using an O 2 plasma cleaner (80 W, 35 sec). We placed a drop of DDI water with food coloring into the inlet port of the capillary channel with no applied pressure. The progress of the aqueous solution through the capillary was recorded using a camera. The recording was analyzed to calculate the experimental capillary flow rates. www.nature.com/scientificreports www.nature.com/scientificreports/ Production of PDMS with PDMS-PEG BCP additives. A block copolymer with a poly(dimethylsiloxane) (PDMS) and hydrophilic poly(ethylene glycol) (PEG) blocks, PDMS-PEG, was purchased from Gelest (product code DBE-712, dimethylsiloxane-(60-70% ethylene oxide) block copolymer, MW 600, 20 cSt, specific gravity: 1.01, refractive index: 1.442, Gelest, USA) and utilized as an additive in the modification of microfluidic devices. Silicone pre-polymer and curing agent were mixed in a mass ratio of 10:1 (w/w). Desired amount of PDMS-PEG BCP was then added to the polymer base-curing agent mix to obtain a final additive concentration of 0.125%, 0.25%, 0.5% 1.0%, 1.5%, 2.0% (w/w) in the mixtures. The mixtures were blended using a glass a rod and poured onto a silicon wafer or into a petri dish for the fabrication of microfluidic devices and slabs, respectively. Trapped air bubbles were removed by keeping the mixture at +4 °C for 15 min. After removing air bubbles, the blended mixture was cured at 70 °C for 24 h. All devices and slabs (~2 mm thick) were rinsed with isopropyl alcohol (IPA) for 24 h and dried at room temperature (RT). Steam sterilization was applied to microfluidic devices before performing experiments. Primary rat hepatocyte isolation and cell seeding. Primary rat hepatocytes were isolated from adult female Lewis rats (Charles River Laboratories, MA) as described previously 57 . All methods were performed in accordance with the guidelines and regulations of National Research Council. For isolation, protocol #2011N000111 approved by the Institutional Animal Care and Use Committee (IACUC) at the Massachusetts General Hospital (MGH) was implemented by the Cell Resource Core (CRC). In general, as determined by trypan blue exclusion, 100-150 million hepatocytes with 90-95% cell viability were obtained and a suspension consisting of primary rat hepatocytes at a final concentration of 5 million cells (M) mL −1 was prepared to plate into microfluidic devices. Before introducing rat hepatocytes, glass bottom of the devices was coated with 50 μg/ mL fibronectin (Sigma-Aldrich) for 30-45 min at 37 °C in 5% CO 2 . Then the cells were plated into the cell culture chamber and the device was connected to a syringe pump with a flow rate of 10 μl/hr and incubated at 37 °C in 5% CO 2 . After 24 hours of seeding, the flow of the fresh media was replaced in the cell culture chamber of perfusion devices and continued thereafter. Dulbecco's modified eagle's medium (DMEM, Life Technologies, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS, Sigma, St. Louis, MO, USA), 0.5 U/mL insulin, 7 ng/mL glucagon, 20 ng/mL epidermal growth factor, 7.5 μg/mL hydrocortisone, 200 U/mL penicillin, 200 μg/mL streptomycin, and 50 μg/mL gentamycin was utilized for culturing primary rat hepatocytes. For all fluidic connections and media perfusion, Tygon tubing (0.01"ID × 0.03" OD, Cole Parmer) was used. Hepatocyte morphology and cell viability. Hepatocyte morphology and viability were assessed by phase contrast microscopy (Evos FL Imaging System, ThermoFisher Scientific). Live/Dead Cell Viability/Cytotoxicity Kit (Thermo Fisher Scientific) were utilized to determine cell viability. For this purpose, Live/Dead assay reagents (calcein AM (10 μL), ethidium homodimer-1 (100 μL)) and PBS (2.5 mL) were combined and vortexed to ensure thorough mixing. Reagents were introduced into the culture chamber and after 30 min incubation (37 °C) and PBS rinsing, images were captured on the EVOS fluorescence microscope to evaluate the cell viability. Protein adsorption study. PDMS-PEG BCP at ratios between 0.125-1.0 (w/w %) was blended with PDMS and poured into a petri dish and cured at 70 °C for 24 h, as described in Section 2.2. After polymerization, round swatches of PDMS samples (5 mm Dia × 4 mm) were cut using a 5 mm dermal punch (Ted Pella Inc.). These samples were immersed in phosphate buffered saline (PBS, pH 7.4) for 2 h to reach pre-equilibration. 0.5 mg/mL solutions of each fluorescently labeled protein, bovine serum albumin (BSA) (Alexa Fluor 594-labeled BSA, Thermo Fisher Scientific) or lysozyme (FITC-labeled, Nanocs), were prepared in PBS separately. To study protein adsorption, 50 μL of fluorescently labeled protein solution was placed on the modified PDMS swatch and incubated in the dark at 37 °C for 1.5 h. For comparison, the same procedure was followed for PDMS with no PDMS-PEG BCP. After 1.5 h, each sample was rinsed with 200 μL PBS. Fluorescence microscope images were captured by Evos FL Imaging System (ThermoFisher Scientific) using 10X objective. Quantitative protein adsorption experiments were also performed using PDMS microfluidic devices with/without PDMS-PEG additives. For this purpose, microfluidic devices were conditioned with PBS at a flow rate of 20 μL/min for 4 hours using a syringe pump and then emptied. 0.05 mg/mL solutions of BSA from the chicken egg (Sigma Aldrich), lysozyme from chicken egg white (Sigma Aldrich) and Immunoglobulin G from human serum (IgG) (Sigma Aldrich) were introduced into the device (30-90 min). The amount of adsorbed BSA, lysozyme and IgG were measured comparing the influx and efflux concentrations utilizing the Pierce BCA Protein Assay Kit (Thermo Scientific) according to the manufacturer's protocol. Characterization. Optical Properties. Optical clarity was quantified using UV-Vis spectrophotometer (Thermo Scientific, Genesys 10S equipped with a high-intensity xenon lamp and dual-beam optical geometry) within the wavelength range of 400-600 nm both for PDMS and BCP modified PDMS samples (0.125-1.0% (w/w)). Samples were tested before and after IPA soaking. All samples were prepared with similar thicknesses (~8 mm) with the purpose of avoiding any disparity in the data. Mechanical properties. Mechanical properties (Young's modulus, compressive modulus) were tested using TA Instruments RSAIII Dynamic Mechanical Analyzer (DMA), (Rheometrics Solids Analyzer). PDMS samples with/ without PDMS-PEG additives for tensile and compressive testing were fabricated according to ASTM standards. For tensile testing, crosshead velocity was 250 mm/min. At strain levels below 40%, the linear behavior allows utilizing Hooke's law (E = σ/ε, where σ is the applied stress and ε is the resultant strain) to calculate Young's modulus 40 . For compression testing, crosshead velocity was set to a maximum of 20 mm/min. Prior to all subsequent compression tests, a drop of machine oil was applied to the parallel surfaces of the PDMS cylinder to prevent excessive friction and the resultant barreling. www.nature.com/scientificreports www.nature.com/scientificreports/ Surface characterization. Sessile drop water droplet contact angles (WCA) were measured at the polymer-air interface using a contact angle goniometer (Rame-Hart Instrument Co., Netcong, NJ) to assess the wettability of PDMS modified with PDMS-PEG BCP additives. Briefly, 6 μL volume of distilled water (18.2 MΩ cm −1 water) was dropped onto the BCP modified PDMS slab (2 cm × 2 cm) and the contact angle was measured at regular time intervals to observe the timeline of surface arrangement. WCA of PDMS without BCP additive substrates (2 cm × 2 cm) was also measured as a control. To characterize the surface chemistry of PDMS with and without PDMS-PEG BCP additives, square samples (1 cm × 1 cm) were prepared. Samples were analyzed using X-ray photoelectron spectroscopy (XPS) using the K-Alpha + XPS system (Thermo Scientific) at Harvard University's Center for Nanoscale Systems. The probe for the measurement was aluminum k-α X-ray line with energy at 1.4866 keV and X-ray spot size at 400 μm with 90 degrees take-off angle (sampling depth is around 10 nm from the surface). A flood gun, which supplies low energy electrons and ions was used throughout the entire experiment for sample surface charge compensation. Both survey spectra and high-resolution scan data were collected at each sample. For survey spectra, the scan was completed by taking an average of 5 scans in 1 eV steps with passing energy at 200 eV from −10 eV to 1350 eV binding energy. For high-resolution scans, the data were collected by taking an average of 10 scans in 0.1 eV steps with passing energy at 50 eV for Si 2p, O 1s, and C 1s photoelectron lines. Statistical analysis. Each biocompatibility experiment was conducted in triplicate using cells from at least three different rat isolations. Three different samples were utilized to quantify the WCA, protein adsorption, and mechanical analysis measurements. Capillary-driven flow experiments were performed with three different samples (n = 3). XPS of each sample was obtained by taking an average of 5 and 10 scans for survey spectrum and high resolution scan data respectively. Wherever indicated, quantitative data were plotted as the mean ± standard error of the mean (n = 3).

chemsum

{"title": "Simple Surface Modification of Poly(dimethylsiloxane) via Surface Segregating Smart Polymers for Biomicrofluidics", "journal": "Scientific Reports - Nature"}

rapid_controlled_release_by_photo-irradiation_using_morphological_changes_in_micelles_formed_by_amph

## Abstract: Photo-induced rapid control of molecular assemblies, such as micelles and vesicles, enables effective and on-demand release of drugs or active components, with applications such as drug delivery systems (DDS) and cosmetics. Thus far, no attempts to optimize the responsiveness of photoresponsive molecular assemblies have been published. We previously reported photoresponsive surfactants bearing a lophine dimer moiety that exhibit fast photochromism in confined spaces, such as inside a molecular assembly. However, rapid control of the micelle structures and solubilization capacity have not yet been demonstrated. In the present work, photo-induced morphological changes in micelles were monitored using in-situ small-angle neutron scattering (SANS) and UV/Vis absorption spectroscopy. An amphiphilic lophine dimer (3TEG-LPD) formed elliptical micelles. These were rapidly elongated by ultraviolet light irradiation, which could be reversed by dark treatment, both within 60 s. For a solution of 3TEG-LPD micelles solubilizing calcein as a model drug molecule, fluorescence and SANS measurements indicated rapid release of the incorporated calcein into the bulk solvent under UV irradiation. Building on these results, we investigated rapid controlled release via hierarchical chemical processes: photoisomerization, morphological changes in the micelles, and drug release. This rapid controlled release system allows for effective and on-demand DDS.Amphiphilic compounds or surfactants form a variety of molecular assemblies, such as micelles, vesicles in water. The aqueous solutions of molecular assemblies can solubilize water-insoluble or poorly water-soluble compounds by incorporation into the molecular assemblies. This solubilization technique is used to dissolve active substances in aqueous media and is an indispensable step for the production of commercial products in the field of beverages, foods, cosmetics, medicines, and also household products.Controlled assembly of the amphiphilic compounds causing deformation and/or morphological changes can be applied for the regulated release of drugs and other active compounds (Fig. 1a). To trigger morphological changes in the molecular assemblies, external stimuli are applied, such as light, redox reaction, magnetic field, or temperature or pH change 1,2 . Among these, light is a promising external stimulus because it is clean and easily applied with high spatial resolution and specific wavelength. Therefore, many photoresponsive amphiphilic compounds have been developed, and the formation of molecular assemblies, such as micelles and vesicles, by photo-irradiation has already been realized 1,3,4 .For decades, morphological changes in molecular assemblies by photo irradiation and their applications have been accomplished. For example, our group demonstrated controlled release of volatile oils as model perfumes with an azobenzene-type cationic surfactant that exhibits photoswitchable formation of micelles upon UV irradiation 5,6 . However, to induce significant changes, these conventional systems require minutes to hours of photo-irradiation. Considering practical applications, the response time needs to be fast, requiring a change in solution properties at an arbitrary moment (ON state) which ceases when the stimulus stops (OFF state) (Fig. 1a). To the best of our knowledge, there has been no attempt to speed up the controlled formation of photoresponsive molecular assemblies. To achieve rapid control of amphiphile self-assembly, we focused on the lophine dimer (LPD) as a photochromic moiety. LPD molecules quickly dissociate into two lophyl radicals and are recovered by a thermal recombination reaction (Fig. 1b) 7 . Although this recombination is extremely slow, it has been reported that the reaction rate is significantly enhanced by inhibiting free diffusion of the radical species. This has been achieved by connecting two lophine moieties through a linker , confining lophine dimers in micelles formed by surfactants , or confining them to microscopic domains formed by ionic liquids 14,15 . We previously achieved rapid photoisomerization of the lophine dimer in micelles formed by amphiphilic lophine dimers themselves (3TEG-LPD, shown in Fig. 1c) 12 and demonstrated rapid and reversible variations in the surface tension of the aqueous solution 13 . However, so far rapid morphological control of the micelles and controlled release system has not yet been accomplished. In the present study, we examine rapid morphological changes in micelles formed by amphiphilic lophine dimers and demonstrate controlled release of a model drug solubilized in the micelles. However, conventional particle-sizing equipment using light scattering and laser diffraction are problematic for the detailed morphological analysis of molecular assemblies and their changes under light irradiation. Small-angle X-ray or neutron scattering (SAXS or SANS) provides precise and direct information about the structures of molecular assemblies. Another interesting feature of small-angle scattering is its capacity for insitu and time-resolved measurement, allowing the monitoring of the dynamics of phase changes triggered by external stimuli . Combining a light source and a spectrometer into a SAXS or SANS measurement system enables the monitoring of the morphological changes in globular to wormlike micelles formed by photoresponsive surfactants 19,20 . This analytical system is suitable for in-situ analysis of micelles formed by amphiphilic lophine dimers. In this work, using a simultaneous SANS and Ultraviolet/Visible (UV/Vis) absorption measuring system, we performed in-situ observations of photo-induced morphological changes in micelles formed by two different amphiphilic lophine dimers (3TEG-LPD and 6TEG-LPD) in the order of several tens of seconds. Furthermore, the ability to control the release of a model drug by photo-irradiation was examined. From these experiments, we investigated rapid controlled release owing to hierarchical chemical process, consisting of photoisomerization, morphological changes in the micelles, and drug release. ## Results and discussion To reveal morphological changes in the micelles and the photoisomerization upon UV irradiation, we performed simultaneous in-situ SANS and UV/Vis absorption measurements by installing a mercury lamp and an UV/ Vis absorption spectrometer on the sample table of a SANS instrument (Figure S1). Figure 2a shows the SANS profiles of 10 mM 3TEG-LPD in D 2 O. To perform SANS measurements at a sufficiently larger concentration than the cmc of 3TEG-LPD (0.80 μM 13 ), the concentration of 3TEG-LPD was set at 10 mM. Several theoretical models, such sas uniform or polydisperse spheres and cylinders, were considered. The experimental data could be well fitted using a uniform prolate ellipsoid model 21 with a long radius (r a ) of ~ 47 and a short radius (r b ) of ~ 28 (Fig. 2a, Table 1). UV irradiation caused an increase in the scattering intensity toward the lower q region, and this reached a constant state within 2 min. The post-irradiated SANS profile was also well fitted using a uniform prolate ellipsoid model. The short radius remained at ~ 28 and a long radius increased to ~ 70 , indicating that UV irradiation induced elongation of the longer axis of the micelles. This is because the hydrophobic portion www.nature.com/scientificreports/ of the resulting lophyl radical has a relatively larger volume in comparison with that of the original dimer form, whose LPD unit is tightly linked with a covalent bond, and photo-irradiation caused formation of micelles having a lower curvature 22 . When the UV irradiation ceased, the SANS profile readily recovered to its initial state (Fig. 2a). Figure 2c shows the integrated scattering intensity in the q-region of 0.01-0.05 −1 , for the scattering profiles at 60 s intervals during cycles of 2 min UV irradiation followed by 4 min standing in the dark (Figure S2). This observation revealed that the elongation of the elliptical micelles completed in 60 s of irradiation and reversed in 60 s of standing in the dark. These reversible changes in the SANS profiles were repeatable over 10 cycles, after which there was incomplete recovery due to partial decomposition. During the SANS measurements, UV/Vis absorption spectra were simultaneously collected. As shown in Fig. 2d, a characteristic absorption band of the lophyl radical at 580 nm was observed during UV irradiation. The temporal change in the absorption at 580 nm during the irradiation cycle revealed that morphological changes in the micelles and production of the lophyl radicals occurred without time lag within the analytical time resolution (Fig. 2c,d). These results indicate that the reorganization of the surfactant molecules in the micelles following photoisomerization is complete within the time scale of the SANS measurements. Interestingly, the photoisomerization and morphological changes in the molecular assembly of the amphiphilic lophine dimers readily proceed even in highly concentrated systems, unlike azobenzene derivatives, which are the most popular photochromic compounds. Azobenzene derivatives show a lower rate and yield of photoisomerization in a confined system 23 or in the solid state 24 because sufficient free volume is necessary for conformational changes in the molecular structure during the trans-cis isomerization. Therefore, the lophine dimer analogue is suitable for photoresponsive functional systems based on molecular assemblies or supramolecules. To examine the effects of the position of hydrophilic groups and the balance of hydrophilicity and lipophilicity on the speed of the recombination reaction and morphological changes in the micelle, 6TEG-LPD was synthesized according to Scheme S1. According to the static surface tension measurements of aqueous solutions of 6TEG-LPD, the critical micellar concentration (cmc) was 0.35 μM (Figure S3), which is lower than that of 3TEG-LPD (0.80 μM 13 ). We assume that the introduction of alkyl chains promoted adsorption of 6TEG-LPD at the air/water interface and formation of the micelles. For the micellar solution of 6TEG-LPD, the dimer-monomer photoisomerization proceeded readily according to the UV/Vis absorption measurements (Figure S4). The apparent rate of the latter recombination of the lophyl radicals (k′), with the assumption of a second-order rate reaction, was 1.5 s −1 , which was approximately a 1500-fold enhancement compared to that in tetrahydrofuran (THF), where 6TEG-LPD does not form micelles and the lophyl radicals freely diffuse. From this observation, 6TEG-LPD showed better enhancement in the radical recombination than 3TEG-LPD with an approximate 800-fold increase 13 , although there is almost no difference in the absolute values of the apparent recombination rates (k′) in the micellar solutions. SANS data of 5.0 mM 6TEG-LPD in D 2 O were well fitted with a model of polydisperse spheres with an average radius (r) of ~ 27 (Fig. 2b). We assume that 6TEG-LPD bearing more bulky hydrophilic groups is likely to form micelles with larger curvature. However, the SANS profile was unchanged by UV irradiation for over 80 min (Fig. 2b). This suggests that UV irradiation induces no morphological change in 6TEG-LPD micelles, despite the photoisomerization proceeding under UV irradiation. We interpret that flexible alkyl chains compensate for UVinduced structural changes in the surfactant, explaining the nonresponsive behavior of the micellar morphology. The abovementioned results show that UV irradiation induced rapid and reversible changes in the morphology of the 3TEG-LPD micelles. Using this promising photoresponsive surfactant, substances solubilized in the micelles can be readily released by UV irradiation. To demonstrate controlled release, we prepared an aqueous solution of 3TEG-LPD containing calcein as a fluorescent model drug. Figure 3a shows the fluorescence spectra of a 5.0 mM 3TEG-LPD solution with calcein at the solubilization limit (1.0 mM) as a function of UV irradiation time. Under UV irradiation, the band at 545 nm assigned to calcein in the micelles decreased, with the peak shifting to a band at 556 nm assigned to aggregated calcein molecules in water. This indicated calcein molecules incorporated in the micelles released under UV irradiation. Figure 3b shows the temporal changes of the relative intensity of the fluorescence peak. The intensity reached a constant state within 60 s of UV irradiation, which is slightly shorter than the change in the SANS profile. We assume that UV irradiation for 60 s induced effective Table 1. SANS profiles with fitting results, after UV irradiation and standing in the dark, of 10 mM 3TEG-LPD (a) and 6TEG-LPD (b) solutions. Prolate ellipsoids for 3TEG-LPD and spheres with a Schultz size distribution for 6TEG-LPD were modeled. ϕ is the volume fraction, and r a or r b is the long or short radius of ellipsoids, respectively. r and p are the mean radius and polydispersity of the radius, respectively. ρ m is the scattering length density of the micelle. Here, the scattering length density of the micelle or solvent (ρ s ) was fixed at 6.30 × 10 −6 −2 for D 2 O. χ 2 is the chi-square value for fitting. www.nature.com/scientificreports/ morphological changes in the micelles, affecting the solubilization state of calcein. In contrast, the fluorescence spectra of the calcein/3TEG-LPD aqueous solution remained in the absence of UV irradiation (Figs. 3b and S5a), representing that UV light induced release of calcein from inside of the 3TEG-LPD. The fluorescence of calcein in the 6TEG-LPD micelles was unchanged upon UV irradiation (Figs. 3b and S5b), corresponding to the nonresponsive behavior of 6TEG-LPD in the SANS profile (Fig. 2b). Furthermore, the fluorescence spectra of aqueous calcein solution almost unchanged during UV irradiation (Figure S5c). These results indicate that the photo-bleaching of calcein was not responsible for the fluorescence changes. To elucidate the changes in solubilization under UV irradiation, we also performed in-situ SANS on the solutions. As shown in Fig. 4, a SANS profile of 5.0 mM 3TEG-LPD in D 2 O with 1.0 mM calcein was well fitted using a model of a uniform prolate ellipsoid with a long radius (r a ) of ~ 66 and a short radius (r b ) of ~ 32 (Table 2), indicating that the 3TEG-LPD micelles swelled as a result of calcein incorporation owing to hydrophobic interaction and/or π-π interaction between aromatic groups of 3TEG-LPD and calcein. UV irradiation caused gradual changes in the SANS data. UV irradiation for 3 min produced a SANS profile similar to that of the empty 3TEG-LPD micelles under UV irradiation (Fig. 2a). Further irradiation induced a significant increase in the scattering profile in the lower q-range of < 0.02 −1 and the slope was q -3 , indicating that coarse aggregates of surplus calcein formed, as visualized in Figure S7. In the SANS profile of 6TEG-LPD in D 2 O, calcein solubilization was www.nature.com/scientificreports/ unchanged by UV irradiation (Figure S6). These results reveal that the morphological change in the 3TEG-LPD micelles under UV irradiation caused rapid and effective release of calcein. This rapidly controlled release system allows for effective and on-demand delivery systems of active components, such as drugs and perfumes. ## Conclusions To investigate photo-induced rapid controlled release systems, we developed photoresponsive surfactants bearing a lophine dimer moiety that proceeds with fast photochromism in confined spaces such as inside molecular assemblies. Rapid control of micelle structures and solubilization capacity have not been accomplished yet. An in-situ SANS system simultaneously incorporating a light source and a UV/Vis absorption spectrometer showed that the morphological changes in micelles formed by the amphiphilic lophine dimer (3TEG-LPD) and the photochromism between dimer and radical forms proceeded within several tens of seconds. On the other hand, another surfactant bearing alkyl chains (6TEG-LPD) showed no morphological change in the micelles under UV irradiation. We assume that the flexible alkyl chains compensate for the change in the hydrophobic section of 6TEG-LPD radicals. We studied the controlled release of calcein as a model drug using micellar solutions of amphiphilic lophine dimers. The fluorescence spectra revealed that UV irradiation caused the release of calcein from inside the micelles within 60 s, which is faster than the morphological change in the empty micelles. This result indicates that the change in the solubilization of calcein under UV light is faster than the completion of morphological changes in the micelles. In-situ SANS revealed similar morphological changes in the calcein-incorporating micelles and in the empty micelles. In the previous studies, we demonstrated rapid photoisomerization with the lophine dimer in micelles and control of surface tension of the aqueous solution. In the present study, rapid morphological controls of the micelles and controlled release system were accomplished by using the amphiphilic lophine dimer. From these results, rapid controlled release via hierarchical chemical processes: photoisomerization, morphological changes in the micelles, and drug release was investigated. This rapidly controlled release system allows for effective and on-demand delivery systems of drugs, perfumes, or other active components. Currently, to avoid damages of these guest molecules and organisms by the UV irradiation, we are developing molecular assemblies responding red or near infrared light. Furthermore, we are working on application of fast morphological control of molecular assemblies to wormlike micelles that have characteristic viscoelasticity in solution, owing to a polymer-like network, which can be applied in control of texture, heat exchange, etc. Therefore, this study could contribute to medicine, cosmetics, food, and device development. ## Methods Materials. Solvents and reagents were purchased from Tokyo Chemical Industry Co. Ltd. (Tokyo, Japan) and FUJIFILM Wako Pure Chemical Co. (Osaka, Japan) and used without further purification. All reaction mixtures and fractions eluted by column chromatography were monitored using thin layer chromatography (TLC) plates (Merck, Kieselgel 60 F254). The TLC plates were observed under UV light at 254 and 365 nm. Flash column chromatography over silica gel (Wakosil C-200, 64-201 μm) was used for separation. Measurements. 1 H-and 13 C-NMR spectra were measured at 298 K in a DMSO-d 6 solution of the samples, using a JEOL model JNM-AL300 (300 MHz) or JNM-ECP500 (500 MHz) spectrometer with Si(CH 3 ) 4 as an internal standard. Chemical shifts (δ) and coupling constants (J) are reported in parts per million (ppm) and Hertz (Hz), respectively. Electrospray ionization mass spectrometry (ESI-MS) spectra were recorded using a JMS-T100CS instrument (JASCO, Tokyo, Japan). High-resolution mass spectrometry (HRMS) spectra (ESInegative) were recorded using a JMS-MS700 system (JEOL, Tokyo, Japan). Elemental analysis was performed using a PE 2400II (PerkinElmer, Massachusetts, US) system. UV/Vis absorption spectra were measured using an Agilent 222 UV/Vis spectrophotometer with a quartz cuvette (1.0 cm path length). Synthesis and the structural characterization. Synthesis of 3TEG-LPD was performed according to a previously reported procedure 13 . A synthetic route for 6TEG-LPD is shown in the supplementary information. www.nature.com/scientificreports/ Photoisomerization of the amphiphilic lophine dimers. UV irradiation was performed using a 200 W Hg-Xe Lamp (SUPERCUR UVF-203S, SAN-EI ELECTRONIC). The irradiation wavelength (260-390 nm) was achieved using a color filter (U340, HOYA). Each solution was irradiated with light (20.0 mW/ cm 2 ) in a 1.0 cm path quartz cuvette. The photochromic behavior of 6TEG-LPD was characterized by UV/Vis absorption spectroscopy (Agilent 222). The apparent rates for the latter recombination of the lophyl radicals (k′) were calculated from the slope of the reciprocal of absorbance at 620 nm originating from the lophyl radical vs. time plots. In-situ SANS. Small-angle neutron scattering (SANS) measurements were performed using TAIKAN on the BL15 beamline at the Material and Life Science Experimental Facility (MLF) in the Japan Proton Accelerator Research Complex (J-PARC) 25 . The neutron wavelength was set to 0.8-7.8 . The sample-to-detector distance was 5.65 m. The q range was from 0.01 to 1.0 . The q value is defined by q = 4πsin(θ/2)/λ, where θ and λ are the scattering angle and wavelength, respectively. The samples were analyzed in a quartz cuvette of 2.0 mm thickness. SANS measurements were performed at 25 °C. The intensity of the scattering profile was converted to absolute intensity by subtraction with a profile of D 2 O and standardization with a profile of glassy carbon. Data reduction and analysis were performed with Igor Pro (Wave-metrics Inc., version 8.03) using the NIST reduction and analysis macros (version 7.99) 26 . The set-up of the in-situ SANS apparatus is shown in Figure S1 27 . REX-250 (Asahi Spectra Co., Ltd.) was used as a UV lamp (365 nm, 26.2 mW/cm 2 ). The UV/Vis absorption system consisted of a C10082CAH spectroscope and an L10290 light source (Hamamatsu Photonics K.K.). The UV light was reflected with a mirror and directed into a sample cuvette. Neutrons could penetrate the mirror. The light for UV/Vis absorption spectroscopy was passed through the upper part of the sample cuvette. A photo detector was installed to observe the exact time at which the light was emitted. During light irradiation, temperature changes in the samples were negligible. In general, the scattering function for an assembly of particles is given by where N p , �ρ , and V p are the number of particles per unit volume in solution, the scattering contrast, and the volume of a single particle, respectively. P(q) and S(q) are the form factor and structure factor, respectively. We employed the scattering functions of uniform ellipsoid and polydispersed sphere models. For a prolate ellipsoid with minor axis R 1 and major axis R 2 , the form factor P(q) is given by where J 1 (x) is the first order Bessel function of x. R s is given by where u is the axis ratio (R 2 /R 1 ). For a polydispersed sphere model, the form factor P(q) is given by where r is the mean radius of the micelle. The volume distribution function D(r) is represented by a Schulz distribution using the mean particle radius (r m ): where Γ(Z) is the gamma function and Z is given by Here, p represents the polydispersity index. Controlled solubilization upon UV irradiation with a fluorometer. 1.0 mM of calcein was added to 5.0 mM aqueous 3TEG-LPD or 6TEG-LPD solution and stirred in the dark for more than two days to reach equilibrium. The 3TEG-LPD solution solubilizing calcein was added to a quartz cuvette of 1.0 cm thickness. Upon UV irradiation (260-390 nm, 20.0 mW/cm 2 ), changes in fluorescence originating from calcein were monitored using an RF-5300PC (Shimadzu Corporation). (1) I q = N p • (�ρ) 2 • V p 2 • P(q) • S(q) (2) P q = 9

chemsum

{"title": "Rapid controlled release by photo-irradiation using morphological changes in micelles formed by amphiphilic lophine dimers", "journal": "Scientific Reports - Nature"}

the_gas_diffusion_electrode_setup_as_a_testing_platform_for_evaluating_fuel_cell_catalysts:_a_compar

## Abstract: Gas diffusion electrode (GDE) setups have been recently introduced as a new experimental approach to test the performance of fuel cell catalysts. As compared to the state-of-the-art in fundamental research, i.e., rotating disk electrode (RDE) measurements, GDE measurements offer several advantages. Most importantly mass transport limitations, inherent to RDE measurements are avoided. In a GDE setup the reactant, e.g., oxygen gas, is not dissolved into a liquid electrolyte but distributed through a gas diffusion layer (GDL), as it is actually the case in fuel cells. Consequently, much higher current densities can be achieved, and the catalysts can be studied in a wider and more relevant potential range. Furthermore, direct contact to a liquid electrolyte can be avoided and elevated temperatures can be employed in a straight-forward manner. However, the use of GDE setups also comes with some challenges. The determined performance is not strictly related to the catalyst itself (intrinsic activity), but also to the quality of the catalyst film preparation. Therefore, it might be even more important than in RDE testing to develop standardized procedures to prepare catalysts inks and films that can be reproduced effortlessly in research laboratories for fundamental and applied experimentation. To develop such standardized testing protocols, we present a comparative RDE -GDE study, where we investigate several commercial standard Pt/C fuel cell catalysts with respect to the oxygen reduction reaction (ORR). The study highlights the strengths of the GDE approach as an intermediate "testing step" between RDE and membrane electrode assembly (MEA) tests when developing new fuel catalysts. ## Introduction To minimize the global impact of climate change on human civilization, human welfare, and biodiversity, it remains crucial to reduce mankind global greenhouse gas emissions. To fulfil social and economic transformations defined by the legally binding Paris Climate Agreement, such as lowering the net CO2 emissions to zero by 2050 thereby limiting the pre-industrial global temperature rise to 1.5 °C, alternative energy sources need to be developed. . So far however, there have been obstacles to their development. A promising system are fuel cells that efficiently convert chemical energy to electric energy by combining hydrogen and oxygen to form water. Despite their potential advantages, many potential applications of fuel cells are still considered too expensive and not yet commercially viable. A major factor that determines the costs of the fuel cell technology is the use of Pt in the fuel cell catalysts. Thus, the catalyst layers need to be improved in a way that they provide maximal power by minimal Pt content. Additional challenges are the scarcity of the active catalyst materials and the limited conversion efficiency (as compared to battery storage). Developing new and improved ORR catalysts with lower platinum content that achieve higher power densities is therefore crucial. One major challenge thereby is the implementation of new catalysts established in fundamental research to applications in fuel cells. In fundamental research most fuel cell catalysts are investigated with a rotating disc electrode (RDE) setup. However, due to the limited mass transport, inherent to RDE setups, the potential ranges at which the kinetics of an ORR catalyst can be investigated is narrow. This limits the transferability of results gained with an RDE setup towards an application in fuel cells. There is a lack of evidence that high performing fuel cell catalysts measured with the RDE setup can unfold their full potential in membrane electrode assemblies (MEAs) that constitute a fuel cell . In order to facilitate the full exploitation of results and knowledge obtained mainly in RDE measurements conducted in fundamental research, new measurement setups with increased mass transport properties have been introduced , - . These setups allow to apply more realistic conditions in the catalyst testing and at the same time should be widely applicable in standard research laboratories. The gas diffusion electrode (GDE) approach fulfils these criteria. - However, one major challenge is to develop and standardize procedures for catalyst testing in GDE setups. In the presented study, we therefore compare the ORR performance of six different commercial Pt/C catalysts in a GDE setup. Standardized RDE measurements serve as benchmark. It is demonstrated that a GDE approach allows a straight-forward optimization of a given catalyst film under conditions relevant for applications. On the other hand, GDE testing using standardized ink recipes might not uncover the full potential of a respective catalyst. ## Chemicals, Gases, and commercial catalyst samples Ultrapure water (resistivity > 18.2MΩ•cm, total organic carbon [TOC] < 5 ppb) from a Milli-Q system (Millipore) was used for catalyst ink formulation, acid/base dilutions and the GDE cell cleaning. For the ink formulation and electrolyte preparation following chemicals were used: isopropanol (IPA, 99.7 + %, Alfa Aesar), 70% perchloric acid (HClO4, Suprapur, Merck), potassium hydroxide hydrate (KOH • H2O, Suprapur, Merck), commercial Pt/C catalysts (19.4 wt. % TEC10E20A, 46.0 wt. % TEC10E50E and 50.6 wt. % TEC10E50E-HT, Tanaka kikinzoku kogyo, as well as HiSPEC 3000, HiSPEC 9100, and HiSPEC 13100, Alpha Aesar) and Nafion dispersion (D1021, 10 wt%, EW1100, Fuel Cell Store). A Nafion membrane (Nafion 117, 183 μm thick, Fuel Cell Store), a gas diffusion layer (GDL) with a microporous layer (MPL) (Freudenberg H23C8) and a GDL without an MPL (Freudenberg H23) were employed in the GDE cell measurements. Before use, the Nafion membrane was prepared and activated as follows: after punching several discs with 2 cm diameter from a Nafion sheet, the discs were treated for 30 min at 80 °C in 5 wt.% H2O2, followed by rinsing with Milli-Q water. Then, the membrane discs were treated for 30 min at 80 °C in Milli-Q water followed by rinsing with Milli-Q water. Finally, the membrane discs were treated for 30 min at 80 °C in 8 wt.% H2SO4, again followed by rinsing with Milli-Q water. All membranes were kept in a glass vial filled with Milli-Q water. For the electrochemical measurements the following gases from Air Liquide were used: Ar (99.999%), O2 (99.999%) and CO (99.97%). ## Small angle X-ray scattering (SAXS) Small angle X-ray scattering (SAXS) measurements were conducted at the University of Copenhagen, Niels Bohr Institute, Denmark, using a SAXSLab instrument as previously reported. The instrument is equipped with a 100XL + micro-focus sealed X-ray tube from Rigaku, producing a photon beam with a wavelength of 1.54 . A 2D 300 K Dectris Pilatus detector was used to record the scattering patterns. The sample powders were placed in-between mica windows in home-made cells. The two-dimensional scattering data were azimuthally averaged and normalized by the incident radiation intensity, the sample exposure time, and the transmission, and then corrected for the background (carbon materal without nanoparticles) and detector inhomogeneities using the SAXGUI reduction software. The resulting dataset is the radially averaged intensity I(q) expressed as a function of the scattering vector q = 4π•sin(θ)/ λ, where λ is the wavelength and 2θ is the scattering angle. The data were fitted assuming a power law and polydisperse spheres. The background corrected scattering data were fitted using a power law to take into account the behaviour at low q value and a model of polydisperse spheres described by a volume-weighted log-normal distribution. Some data were best fitted by adding a second model of polydisperse spheres also described by a volume-weighted log-normal distribution. The scattering data are fitted to the following general expression: where A•q -n corresponds to the power law where A and n are free parameters; C1 and C2 are scaling constants, PS1 and PS2 the sphere form factors, V1 and V2 the particle volumes and Dv1 and Dv2 the log-normal size distribution. The sphere form factor is given by DV was assumed to be a log-normal distribution: where σ is the variance and R0 the geometric mean of the log-normal distribution (evaluated here in ). The fitting was conducted with the home-written MATLAB code available on request. The free parameters in the model are A, n, R1, R2, σ1, σ2, C1, C2. The scattering data and related fits are reported in Figure S 1 and the values obtained for the fitting parameters are reported in Table S 1. ## Transmission electron microscopy (TEM) and scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM-EDS) A Jeol 2100 transmission electron microscope (TEM) operated at 200 kV was used for the TEM analysis. The samples were prepared by suspending the commercial catalyst powders in ethanol and then dropping the sample suspension onto carbon coated copper TEM grids (Copper or Nickel grids, Quantifoil). Micrographs were recorded in at three different magnifications at least, and in at least three randomly selected areas. At least 200 nanoparticle diameters were evaluated using the software ImageJ to evaluate the size distribution. The SEM-EDS cross-section measurements were performed as described before ## Electrochemical characterization 2.4.1 Catalyst ink and film formation for the RDE measurements The inks for the RDE measurements were prepared from the respective dried catalyst powder and dispersed in a mixture of Milli-Q water and isopropanol (Vwater: VIPA= 3: 1). To the ink 1.6 μL/ml 1 M KOH (aq) was added and then homogenized in a sonicator bath for 10 min. The resulting homogeneous catalyst ink had a total Pt concentration of 0.218 gPt L -1 . Thin catalysts films were prepared by pipetting 9 µL (0.218 gPt L -1 ) of each catalyst ink onto a newly polished glassy carbon (GC) disc. The disc was then dried in an argon atmosphere. The resulting films had a Pt loading of 10 µg cm -2 and were dried at ambient atmosphere for further electrochemical measurements. ## Catalyst ink and film preparation for the GDE measurements Catalysts inks were prepared from different dried catalyst powders and dispersed in a mixture of Milli-Q water and isopropanol (mixture volume ratio of 3:1). To disperse the powder, the mixture was sonicated for 5min at room temperature. Subsequently, Nafion solution was added so that the ink contained a mass C:Nafion ratio of 1. The ink was sonicated again for 5min. The final inks had a Pt concentration of 0.5 mg/ml for all catalysts. The catalyst films were produced by a vacuum filtration of the catalyst ink onto GDL. To conduct the vacuum filtration, the ink was first diluted by Milli-Q water to a Pt concentration of 0.05 mg/ml. The ink was then added to a vacuum apparatus and filtrated through an MPL-coated GDL (Freudenberg H23C8). The resulting catalyst films (Ø =4cm) were stored in petri dishes. From this film coated GDL, a disk (Ø = 3mm) was extruded and used as GDE. All investigated GDEs prepared from the commercially available Pt/C catalysts had a Pt loading of 208 ugPt • cm -2 on the GDL. ## Rotating disk electrode (RDE) measurements All RDE electrochemical measurements were performed at room temperature with a computer controlled potentiostat (ECi 200, Nordic Electrochemistry ApS) and a glass cell equipped with 3 electrodes as previously reported. The working electrode (WE) was a glassy carbon (GC) disk (5 mm in diameter) embedded into a Teflon tip. A Pt wire served as counter electrode (CE) and a reversible hydrogen electrode (RHE) served as a reference electrode. An aqueous 0.1 M perchloric acid electrolyte was used which was saturated with argon prior to the start of the electrochemical measurements. The solution resistance was measured with a superposed AC signal (5mV, 5kHz) and was compensated down to 2 Ω. The analytical procedure to electrochemically analyse the Pt/C catalyst layers was repeated for all six investigated Pt/C catalysts and included the following steps: Surface cleaning, Ar background, ORR activity, and CO stripping to determine Pt active surface area. The Pt catalyst surface was cleaned under an argon atmosphere by cycling the potential between 0.05 VRHE and 1.20 VRHE with a scan rate of 0.50 V s -1 . After roughly 50 cycles a stable cyclic voltammogram (CV) was observed. Afterwards, an Ar background was measured in a potential range between 0.05 VRHE and 1.10 VRHE with a scan rate of 0.05 V s -1 in Ar saturated electrolyte. Prior to the ORR performance measurements, the electrolyte was purged with O2 for 10 min. During the ORR activity measurement, the potential window and the scan rate were the same as that applied for Ar background measurements, while the RDE had a rotation speed of 1600 rpm. To determine the electrochemical active surface area (ECSA) of the investigated catalysts, the oxidation charge obtained from a CO monolayer stripping experiments was analysed. In brief, the electrode was held at 0.05 VRHE in CO saturated electrolyte for 2 min. Subsequently, the electrolyte was saturated with Ar (~10 min) to purge the electrolyte from CO. The potential was swept from 0.05 to 1.10 VRHE with a scan rate of 50 mV s -1 to oxidize the adsorbed CO monolayer to CO2. The ECSA was then calculated from the ration of resulting oxidative charge (QCO), after background subtraction, and the oxidation charge of a monolayer, 400 µC cmPt -2 , and finally normalized to the mass of the Pt (mPt). The ORR data was analysed from the background corrected polarization curves. The background polarization curves were recorded in Ar-purged electrolyte. The ORR activity was then evaluated at 0.90 VRHE from positive going scans. The mass activity (MA) was obtained by normalizing the activity by the Pt mass. The specific activity (SA) was obtained by normalizing the measured current density (mA cmGeo -2 ) to the ECSA. ## Measurements in the Gas Diffusion Electrode Setup The GDE-setup was assembled as the follows : A 3 mm disc was punched out of the catalyst film covered GDL. The catalyst containing disc was placed into an MPL-coated GDL disc (Ø = 2cm, Freudenberg H23C8) which had a 3mm hole in the middle. A Nafion membrane was placed on top (Nafion 117, thickness 183um). With a tablet press (pressure range: 0-15T), the whole stack was pressed together at a pressure of two tonnes and a duration of 10 min. Afterwards, a GDL (Freudenberg H23) was placed into the gas flow field of the lower cell body, followed by the stack containing the GDE and the Nafion membrane. Finally, the upper cell body was placed on top of the Nafion membrane. The two body parts were held in place by a clamp. The compartments of the upper cell body were filled with 15 ml of 4 M perchloric acid. Finally, an RHE and the CE (Pt wire) where put into the electrolyte. All electrochemical measurements were performed at 30°C with a computer controlled potentiostat (ECi 240, Nordic Electrochemistry ApS) and a GDE-setup as reported. The analytical procedure to electrochemically analyse the Pt/C catalyst layers was the same for all six investigated Pt/C catalysts and included the following steps: First, the GDE was purged from the backside (through the GDL) with argon gas. Doing so, the catalyst was cleaned by potential cycles between 0.05 and 1.10 VRHE at a scan rate of 0.2 V s −1 until a stable cyclic voltammogram (CV) could be observed (∼50 cycles). Afterwards, a CO-stripping measurement was performed followed by electrochemical impedance spectroscopy (EIS) and ORR-activity measurements. To conclude the investigations, a second CO-stripping measurement was performed. Throughout the entire experiment, a bubbler was used to humidify the gas and the membrane. CO-stripping measurements were conducted to determine the ECSA. In essence, the catalyst layer got covered by CO gas which adsorbed onto the Pt surface. Afterwards, the catalyst was purged with Ar to remove the excess of CO. As a next step, a CV was recorded (scan rate 50 mV/s), which records the oxidative current originating from the oxidation of CO to CO2. Finally, multiple CVs under Ar atmosphere were conducted until the Ar background was re-gained. The value of the ECSA was then obtained as previously described in section "2.4.3 Rotating disk electrode (RDE) measurements". Prior to the ORR activity measurements, oxygen was flowed through the pipes for 10 minutes. For the last 5 minutes, a potential of 0.80 VRHE was applied. This ensured that all gas lines were fully filled with oxygen and that the catalyst layer was equally wet over the entire surface. The ORR-activity measurements were conducted in potential control mode with a potential range between 1.00 VRHE and 0.10 VRHE. The potential was pre-set to 1.00 VRHE and then lowered in steps of 25 mV until 0.10VRHE are reached. At every step the potential was held constant for 1 minute to reach steady state conditions. For analysis, the measured current was averaged over the last 10 s. The solution resistance was measured by superposing a signal with a fixed frequency of 5kHz and an amplitude of 5mV. Finally, all ORR activity measurements were post-corrected for the potential errors introduced by the solution resistance. We start with the physical characterization of the investigated Pt/C fuel cell catalysts. All examined catalysts are commercially available and can serve as benchmarks in studies investigating new, home-made fuel cell catalysts. Their Pt to C ratio (Pt loading), as indicated by the supplier, ranges from roughly 20 wt. % up to 70 wt. %. In Figure 1, we present representative TEM micrographs to demonstrate the physical characteristics of each Pt/C catalyst. In addition, size histograms and average particle sizes derived from a TEM analysis as well as probability density functions derived from fitting the SAXS data are shown. As Figure 1 shows, within the accuracy (error) of the measurements both methods lead to the same average particle size. ## Physical Characterization of the commercial Pt/C catalysts by TEM and SAXS However, with a closer look at the size retrieved, the average Pt particle size determined from TEM is slightly smaller (except for TKK 46 wt.% Pt/C) in comparison to the values derived from the SAXS analysis. This difference can be explained by the fact that the particle size distributions are based on different analyses that are sensitive to different sizes in different ways: For the TEM analysis, one determines the relative number of particles with the same size based on defined bin sizes and only relatively few individuals NP are accounted for. In contrast, SAXS analysis is performed in a larger volume of sample and so more NP are considered for the size evaluation. Additionally, the size retrieved from TEM is often number-or surface-weighted, whereas it is volume weighted for SAXS: i.e., SAXS is more sensitive to the contribution of larger NP sizes. This explains why SAXS analysis led to an estimated diameter slightly larger than for TEM analysis in this study. Nevertheless, due to the good agreement of the results obtained by both analysis techniques in the present study, we do not distinguish in the following between the two methods when referring to the average particle size and size distribution. The analysis shows that the average particle sizes range from roughly 2 to 5 nm (Figure 1). In addition, the carbon support of each investigated catalyst is relatively homogeneously decorated by Pt particles; in particular the TKK 19.4 wt. % Pt/C sample. The limited particle agglomeration on the carbon support of this catalysts is also reflected by the very narrow size distribution with a standard deviation of only 0.4 nm in the TEM analysis. As expected, it can be clearly seen that at increased Pt loadings the carbon support is more densely covered with Pt particles and agglomeration increases. Characteristically in the TEM micrographs of the HISPEC 60 wt.% and 70 wt.% Pt/C samples, some darker spots are seen that most likely are related to the slightly agglomerated Pt particles and the size distribution exhibits a clearly discernable tail towards larger sizes (also the size distribution of the TKK 50.6 wt. % Pt/C sample displays such a feature). ## Physical Characterization of the as prepared GDEs from commercial Pt/C catalysts In Figure 2 ## Electrochemical Catalyst Characterization: RDE vs. GDE As mentioned in the introduction, one of the main motivations for introducing new measurement setups with increased mass transport has been the challenging implementation of promising catalysts identified in RDE measurements to fuel cells. Often the performance measured in a RDE setup cannot be translated to corresponding improvements in MEA measurements and there is a large gap between fundamental research and applications. Therefore, it is of interest to systematically compare the performance of the Pt/C catalysts as determined by RDE measurements with their performance in GDE measurements. In previous work of our and other groups, the TKK 46 wt. % Pt/C sample was used as a benchmark or reference catalyst - . Therefore, in the following we discuss the GDE setup with this catalyst. In Figure 3a representative CVs of the TKK 46 wt.% Pt/C samples recorded in the two setups are compared. In both cases, the typical "electrochemical features" of a Pt/C catalyst are depicted. In the low potential region of the CVs (0.05 -0.35 VRHE), both hydrogen adsorption (negative scanning direction) and desorption (positive scanning direction) are visible, typically referred to as Hupd peaks. However, the Hupd peaks in the CV of the GDE measurements slightly differ from the ones in the RDE, which are typical for measurements in aqueous perchloric acid electrolyte. In particular, in the CV recorded in the GDE setup the "second" peak at around 0.25 VRHE is less pronounced, and the hydrogen evolution reaction (HER) starts earlier, around 0.07 VRHE. In contrast to these differences, the adjacent potential region between 0.35 VRHE and 0.60 VRHE, the double layer region defined by capacitive currents from charging and discharging the interphase, displays identical double layer capacities in both setups. Finally, in the potential region of Pt oxidation and reduction (0.60 -1.10 VRHE), the Pt oxidation and reduction peaks in the CV recorded with the GDE setup are slightly shifted towards higher potentials. These observed differences are most likely a consequence of the different local ion environments. It is well-known that the hydrogen features are sensitive to the local ion-population. In the GDE setup the catalyst is surrounded by a solid Nafion electrolyte, whereas it is surrounded by a liquid aqueous electrolyte in the RDE setup. The earlier onset of the HER in the GDE setup might be related to the reduced local partial pressure of hydrogen. The effect of different reaction environment manifests itself even more in the CO stripping measurements that are typically used to determine the electrochemically active Pt surface area , see lower graph in Figure 3. The CO oxidation peaks recorded in both setups are clearly shifted against each other. Interestingly, in the GDE setup the CO stripping peak appears at lower potentials than in the RDE setup (ca. 0.8 vs. 0.9 VRHE). Thus, the shift is more pronounced and in opposite direction as compared to the potential difference in oxide formation observed in the CVs recorded in Ar atmosphere. It should be pointed out that this shift is not related to an incomplete CO monolayer formation as can be seen from the absence of Hupd features in the forward going CO stripping scan. Furthermore, the Hupd peaks in the CV recorded after the CO stripping indicate that the surface area in both environments is very similar, see also below. Therefore it can be argued that the shift in the CO stripping peak is related to a reduced anion blocking in the GDE membrane-catalyst environment . Interestingly, the peak position observed in the CO stripping curve recorded in the GDE setup is similar to the one observed in an MEA measurement by Harzer et al. , although it needs to be stressed out that a direct comparison is difficult due to the different catalyst and different experimental parameters such as scan rate and temperature. In Figure 4, the behavior of the Pt/C benchmark catalyst in O2 saturated atmosphere is presented from measurements by both setups, i.e., the RDE and the GDE setup. Focusing first on the measurement limitations of the RDE setup, it is demonstrated that the maximum ORR current density which can be reached (at 1600 rpm) is around 6 mA cm -2 Geo. The broad current plateau indicates that in a wide potential region the ORR is limited by mass transport through the hydrodynamic layer at the electrode interface. By contrast, in the GDE setup a current density up to 1400 mA cm -2 Geo can be reached in the same potential region because oxygen gas can directly diffuse through the GDL to the catalyst layer. A GDE setup is thus particularly apt at investigating catalysts with higher current densities and lower potentials, which reflect more realistic conditions that are closer to the operational window for a real fuel cell. It should be mentioned though, that the maximum current density reached in the GDE setup can vary up to 50 % between different samples, highlighting the influence of the catalyst layer on the obtained results. To further compare the results, the geometric current densities were normalized to the ECSA derived from the CO stripping measurements. The SA is shown in a Tafel plot, i.e. a logarithmic yaxis, in Figure 4's lower part. From this plot, the broader kinetic (linear) region can be observed, that goes from 1 VRHE down to 0.75 VRHE form GDE data. In contrast, for the RDE data, the linear region ends at around 0.90 VRHE, due to the onset of diffusion limitations. Astonishingly, the SA obtained in low-current regions (above 0.80 VRHE) from RDE setup is significantly higher than the one measured in a GDE setup. This is caused by the fact that for GDE measurements we started at 1.00 VRHE and went stepwise more negative to 0.10 VRHE, where each step took 1 min. Thus, during the first few steps the catalyst had a steady state coverage of oxygenated species which partially block active sites. In contrast, dynamic potential cycling is used for RDE measurements. Here, the steady state coverage has not yet been obtained, and thus leads to an apparent higher activity. This is well-known from the different activities recorded in an RDE for the positive and negative going sweeps, respectively. Both measurements protocols are typically used for respective setup, and thus, can lead to the appearance of different SA. To support this statement, we performed a potential holding experiment in an RDE setup. (Figure S 3). However, a GDE setup is not particularly designed to investigate the catalysts properties at 0.90 VRHE, instead the focus is set to lower voltages with higher current densities, which reflects more realistic conditions, i.e., in the range between 0.70 VRHE and 0.80 VRHE. This range equals the operational window for a real fuel cell and is thus especially important. The kinetic region with the traditionally applied RDE setup is, however, not lower than 0.85 VRHE (at around 0.85 VRHE or below, the diffusion limited region is already reached due to reactant transport limitation), in comparison, the kinetic region from a GDE setup can be extended to 0.75VRHE. Hence, a loss in activity at low current region in GDE is not decisive as a loss in high current regions. ## Characterization and Comparison of different commercial Pt/C catalysts In the following, the performance of the different commercial Pt/C catalysts in both setups is compared based on key values to highlight general trends. ## ECSA In Figure 5 the ECSA values of the different Pt/C catalysts measured in both setups is plotted vs. their "theoretical" ECSA, which is calculated from the TEM size histograms assuming that the Pt NPs are perfect, free-standing spheres, i.e., no Pt surface area is blocked by the carbon support. Part of the Pt/C samples can also be compared to previous measurements. The diagonal line in Figure 5 indicates where measured and "theoretical" ECSA values are equal. It is demonstrated that there is in general a good agreement between the measured ECSAs and the expected ECSA based on the particle size distribution. However, the ECSA values determined in the GDE setup are slightly lower than the ones obtained in the RDE measurements. This general trend is visualized in Figure 5 by fitting linear trendlines to the data points. The difference most likely is related to the presence of Nafion in the GDE catalyst layer. Nafion is known to partially block the active surface area of the active catalyst phase and thus reduces the ECSA . By comparison, in the RDE measurements no Nafion binder was used. ## ORR performance The goal of an RDE characterization is to determine the intrinsic kinetic ORR activity of a catalyst. Such task is challenging as in the past even the results of relatively "simple" Pt/C catalysts had been varying by one order of magnitude . As a result, benchmarks such as polycrystalline Pt have been introduced and several works on measurement procedures and best practices have been published , , , , , . The basic assumption is that procedures and conditions can be defined where all catalysts exhibit their maximum performance. Implementing such approach to GDE measurements we adopted a procedure of Yarlagadda et al. to prepare Pt/C films on top of a GDL using the same Pt loading and a standardized ink composition for all Pt/C catalysts. Furthermore, the same automized testing protocol has been applied, see experimental section. The activity results then can be compared at either a fixed potential or at fixed current density. In the following, the performance of the different Pt/C catalysts is compared at a fixed current density of 5 A mPt -2 . As Figure 4 shows, at low overpotentials the SA measured in a GDE setup is significantly lower than the one measured in an RDE setup. This loss in SA was observed for all six investigated Pt/C catalysts and is summarized in Figure 6. Figure 6 shows that with the used standardized procedure a potential shift in the range of 0.067 VRHE and 0.108 VRHE is observed between the two approaches which constitutes a substantial difference. These differences in performance can be attributed to several reasons. On the one hand, it needs to be noticed that RDE and GDE measurements are conducted using different procedures. RDE measurements are typically conducted potentiodynamically and only the forward-going, more active potential scan is analyzed at a fixed potential of 0.9 VRHE as shown above in Figure 4. The obtained kinetic current density (after correcting for mass transport limitations) is considered as the intrinsic ORR activity of the catalyst surface. However, the scan rate dependence of such measurements clearly indicates that the ORR performance under such conditions might be overestimated. By comparison, the high current densities obtained in an GDE setup make such procedure difficult. Any uncompensated resistance (iR drop) leads not only to a shift in potential but also to a current dependent change in the scan rate. Potentiostatic or galvanostatic measurements by comparison can be corrected for the iR drop in a straightforward manner but face the challenge of a more or less pronounced time dependence in the recorded current or potential. Hence in the current work, we choose to average the currents recorded in a set time interval, see experimental section. On the other hand, apart from these systematic differences which should lead to a constant shift in activity between all catalysts, the different measurement results of the 50.6 wt. % catalyst in Figure 6 indicate that an automized and standardized procedure might not always be suitable to ensure that each catalyst exhibits its optimal, intrinsic performance. For example, an improved cleaning procedure in oxygen might improve the performance (in the specific case shown here, in the RDE measurements) while for other catalysts it might lead to slight degradation, e.g., in case of small particles. Furthermore, the SEM-EDS cross sections demonstrate substantially different thicknesses of the Pt/C catalyst films depending on the Pt loading on the carbon support. Furthermore, the activity difference in RDE and GDE can be in part be attributed to the same, unoptimized ink composition for all Pt/C catalysts. For example, a fixed carbon to Nafion (C:N) mass ratio, might not be the best recipe for all the different catalysts. The different Pt loadings on the carbon support, the different Pt particle size distributions as well as different carbon supports might require specific ink compositions for every single catalyst to optimize the performance in the GDE setup; knowledge that is commonly known for MEA measurements and is part of the optimization of fuel cell catalyst layers. ## Catalyst layer optimization To investigate this hypothesis, we analyzed the specific ORR activity (SA) of a moderately performing catalyst, i.e., the HISPEC 70 wt. % at different C:N mass ratios. As demonstrated in Figure 7, the conventional ink recipe (C:N=1:1) does not lead to the best performance of the HISPEC 70 wt. % catalyst. The obtained maximum power density strongly depends on the C:N ratio in the ink (Figure 7). By changing the C:N ratio, the maximum power density can be almost doubled from about 0.4 W cm -2 to about 0.7 W cm -2 . A standardized ink recipe therefore leads to an "underperformance" of certain catalysts. For a meaningful comparison of different catalysts in a GDE, it is therefore important to consider optimizing the ink composition for every single catalyst. To demonstrate this conclusion even further, and to analyze which characteristics are crucial for a good performance of a specific Pt/C catalyst, we also analyzed the influence of the Pt to Nafion (Pt:N) ratio of this specific catalyst by introducing additional carbon support in the catalyst ink. In the plot in Figure 8 it is demonstrated that at 0.9 VRHE the ORR performance increases with increasing C:N ratio. With regards to the Pt:N ratio, it seems that the ORR performance increases with increasing ratio as well. However, this behavior changes as soon as higher current densities are reached. At 0.8 VRHE the highest ORR current density was reached with a Pt:N ratio of 1 instead of a ratio more than 2 at 0.9 VRHE. Furthermore, it is shown that at this Pt:N ratio, the C:N ratio does not have a substantial influence on the current density anymore. This trend gets even more pronounced at 0.7 VRHE. The highest current density for the HISPEC 70 wt. % catalyst were obtained with a Pt:N ratio of 1 (by adding carbon support to the ink) and a C:N ratio of 1. establishing standardized procedures and the use of benchmarks will be essential to obtain meaningful results. On the other hand, the results clearly demonstrate the potential of the GDE approach to bridge RDE and MEA measurements thus helping to commercialize new ORR catalysts. Most importantly, the GDE approach allows focusing on relevant current densities that are inaccessible in RDE measurements. Moreover, the optimization of characteristics such as the ink recipe or the applied catalyst loading on the GDL for each individual catalyst is feasible in a much simpler manner than in elaborate MEA testing. Therefore, the GDE approach has the clear potential to reach similar popularity as the RDE approach. Surf. Sci. Rep., vol. 45, no. 4-6, pp. 117-229, 2002, doi: 10.1016/s0167-5729(01)00022-x. T. Yoshizumi, H. Kubo, and M. Okumura, "Development of High-Performance FC Stack for the New MIRAI," 2021, doi: https://doi.org/10.4271/2021-01-0740. M. Inaba, Y. Kamitaka, and K. Kodama, "Eliminating the need for craftsmanship: Facile and precise determination of oxygen reduction reaction activity by spraying catalyst ink on rotating disk electrode," J. Electroanal. Chem., vol. 886, no. December 2020, p. 115115, 2021, doi: 10.1016/j.jelechem.2021.115115. ## Details SAXS analysis and results The average volume of nanoparticle from population 1 and from population 2, <V>1 and <V>2 respectively, lead to define volume fraction of population 1, ႴV1, and volume fraction of population 2, ႴV2, as: where N1 and N2 are the number of nanoparticles in the population 1 or 2 respectively. From the SAXS data acquisition we have the relationship between the retrieved coefficient C1 and C2 given by Ci = k. Ⴔvi. <V>i where i=1 or 2 and k is a constant. In order to weight the probability density function by the area or surface fractions we consider <A>1 and <A>2 as the average area of the nanoparticles from population 1 and 2, respectively:

chemsum

{"title": "The Gas diffusion electrode setup as a testing platform for evaluating fuel cell catalysts: a comparative RDE-GDE study", "journal": "ChemRxiv"}

autodesigner,_a_de_novo_design_algorithm_for_rapidly_exploring_large_chemical_space_for_lead_optimiz

## Abstract: The lead optimization stage of a drug discovery program generally involves the design, synthesis and assaying of hundreds to thousands of compounds. The design phase is usually carried out via traditional medicinal chemistry approaches and/or structure based drug design (SBDD) when suitable structural information is available. Two of the major limitations of this of a SBDD project where limited data is available. To assess the effectiveness of AutoDesigner, we applied it to the design of novel inhibitors of D-amino acid oxidase (DAO), a target for the treatment of schizophrenia. AutoDesigner was able to generate and efficiently explore over 1 billion molecules to successfully address a variety of project goals. The compounds generated by AutoDesigner that were synthesized and assayed (1) simultaneously met not only physicochemical criteria, clearance and central nervous system (CNS) penetration (Kp,uu) cutoffs, but also potency thresholds; (2) fully utilize structural data to discover and explore novel interactions and a previously unexplored subpocket in the DAO active site. The reported data demonstrate that AutoDesigner can play a key role in accelerating the discovery of novel, potent chemical matter within the constraints of a given drug discovery lead optimization campaign. approach are (1) difficulty in rapidly designing potent molecules that adhere to myriad project criteria, or the multiparameter optimization (MPO) problem, and (2) the relatively small number of molecules explored compared to the vast size of chemical space. To address these limitations we have developed AutoDesigner, a de novo design algorithm. AutoDesigner employs a cloud-native, multi-stage search algorithm to carry out successive rounds of chemical space exploration and filtering. Millions to billions of virtual molecules are explored and optimized while adhering to a customizable set of project criteria such as physicochemical properties and potency. Additionally, the algorithm only requires a single ligand with measurable affinity and a putative binding model as a starting point, making it amenable to the early stages ## Introduction The aim of most current small-molecule drug discovery efforts is to develop several diverse, patentable series of compounds that potently bind to a target protein while also having favorable properties such as solubility, permeability, binding specificity and metabolic stability as specified in a Target Product Profile (TPP). 1 It is well documented that discovery teams find this multiparameter optimization problem challenging, often delivering high molecular weight (MW), lipophilic compounds that consequently possess poor pharmacokinetic (PK) profiles, a phenomenon described as "molecular obesity". 2 As a result, there has been considerable interest recently in using Artificial Intelligence (AI) to generate compounds de novo under the premise that algorithms will not be subject to such human behavioral biases and thus deliver higher quality compounds. 3 While a number of these algorithms have reported success on retrospective 4,5 and prospective tests, 6 there are comparatively fewer reports of their performance in the setting of an active drug discovery program. We hypothesized that a de novo compound design algorithm that emulates the design approaches of medicinal chemists, searches chemical space on a vast scale, and remains constrained to a chemical property space desirable for good PK would be able to produce diverse, drug-like, and potent compounds. To test this hypothesis, we have created a de novo design algorithm, AutoDesigner, that is based on a number of ligand design strategies that medicinal chemists commonly use during the Hit-to-Lead (H2L) and Lead Optimization (LO) discovery stages, does not require a large amount of prior experimental data, and is inexpensive to run relative to the cost of synthesizing compounds. The algorithm uses a combination of matched molecular pair (MMP) transforms, reaction-based enumeration, and R-group decoration to explore chemical space on an ultra-large scale, routinely on the order of 100 million to >1 billion molecules, usually with runtimes on the order of 24-48 hours. These iterative generative cycles of exploration are coupled with an extensive filtering cascade that evaluates the molecules against a large number of endpoints such as physicochemical properties, complexity, and the ability to dock in the receptor. The end result is a large pool of diverse, drug-like compounds that can fit into the target protein's binding pocket, thoroughly explore available vectors in a synthetically tractable fashion, and whose potency can readily be evaluated using structure-based computational methods like Free-Energy Perturbation (FEP+). 7 As a challenging real-world, prospective test case for the AutoDesigner algorithm, we chose the development of inhibitors of D-amino acid oxidase (DAO). Studies have shown that inhibiting DAO to increase levels of D-serine, an endogenous agonist of the glycine modulatory site of the N-methyl D-aspartate receptor (NMDAR), could be a valid therapeutic approach for treating schizophrenia. 8,9,10,11 A variety of DAO inhibitors binding the catalytic site have been identified 12,13 and several public and in-house high-quality co-crystal structures have been made available to use as the starting point for an AutoDesigner effort (Figure 1). Full details of our DAO discovery program can be found elsewhere. 14 In this work, we demonstrate how suggest more broadly how it might find utility in a variety of discovery programs. ## Methods The AutoDesigner algorithm consists of a multi-stage search algorithm that utilizes cloud computing to carry out successive rounds of chemical space exploration and filtering. This de novo design algorithm only requires a single ligand with measurable binding affinity and a putative binding model, making it amenable to the early stages of a SBDD project where limited data is available. Over the course of several generative stages and subsequent filtering stages, millions to billions of molecules are explored and evaluated against a customizable set of project criteria. A general overview of the AutoDesigner algorithm is depicted in Figure 2. predictions. If an AutoDesigner run generates tens of thousands of compounds, then the processing of all compounds can be sped up effectively using an active learning approach based on FEP+ predictions. 15,16 In Active-Learning FEP (AL-FEP), single-edge FEP+ predictions (utilizing one central reference compound) from a random subset are used to train a machine learning (ML) model, and the resulting ML model is used to score the remaining compounds in the library. Based on the ML scores, the next set of compounds is selected for single-edge FEP+ profiling, and a new ML model is trained on the cumulative predictions. This iterative cycle is repeated until all promising compounds have been retrieved from the library. 17 ## Matched Molecular Pair Transformations Matched molecular pairs (MMP) were generated by fragmentation of the molecules from the PubChem 18 and ChEMBL 19 databases using a maximum heavy atom difference of 8 atoms and a maximum of 2 cuts. 20 The resulting outputs were converted to their respective reaction SMARTS to be applied as transformations. In total, 291 million and 26 million unique transformations were generated from the PubChem and ChEMBL databases, respectively. A more detailed analysis of the extracted transforms can be found in the Supplementary Material (S1). Before applying the set of transforms, a set of input compounds is generated from a single starting ligand to introduce additional chemical matter to maximize the number of applicable transformations. The full set of 300 million transforms is applied to these input molecules and only ligands that maintain the predefined immutable core are kept. The resulting output is trimmed recursively, as described below, to generate additional ligands. The ligands are filtered using an intermediary filtering cascade, as described below, and used as inputs in the following generative stage. ## Reaction-based Enumeration All the passing compounds from the MMP transformation stage are used as inputs for the reaction-based enumeration stage. This generative stage performs a retrosynthetic analysis of the input molecules, up to a configurable depth, followed by combinatorial synthesis of the resulting reaction pathways using our PathFinder tool. 15 All routes for each input ligand are enumerated using commercially available reactant libraries. The reactant containing the immutable region, or common core, of the ligand is kept constant while the other reactant(s) are varied. The diverse nature of the input ligands, originating from the matched molecular pair transformation stage, generates a large variety of possible routes, and ensures a diverse exploration of synthetically-accessible chemical space. After deduplication, the ligands are recursively trimmed, as described below, and the resulting output is filtered in the intermediary filtering cascade. ## Recursive trimming Before the outputs resulting from both the MMP transformation stage as well as the reaction-based enumeration stage are progressed into the filtering cascade, a recursive trimming algorithm is applied to generate additional ligands. This algorithm breaks predefined single bonds in a recursive fashion, while at the same time keeping the bonds of the immutable core intact, leading to a large number of additional ligands generated from all possible relevant permutations (Figure 3). Rules are applied to prevent trimming of bonds that would lead to undesirable outputs (e.g., aromatic bonds). ## R-group Decoration In the final two generative stages, all ligands that survived the intermediary filtering cascades following the MMP transformation stage and the reaction-based enumeration stage are combined. The resulting ligands are then decorated using a curated library of R-groups. Each input is decorated at every available site excluding the fixed core. After the first decoration round, the resulting products are funneled through the intermediary filtering cascade, and the remaining compounds are subjected to another round of R-group decoration. Only ligands that can accommodate the extra atoms from the R-group are decorated. Two rounds of decorations ensure a broad exploration of the available chemical space. After the second round of decoration, the resulting products are combined with all of the ligands that passed the intermediary filtering cascades of the first three generative steps (MMP transformation, reaction-based enumeration, and R-group decoration). After deduplication of this set, the ligands are funneled into the final filtering cascade. ## Filtering Stages Of equal importance to generating ligands and exploring chemical space is the ability to effectively filter these large numbers of compounds so that only the compounds that match the desired profile are retained. The AutoDesigner algorithm contains a progressive filtering cascade that employs a range of different filtering techniques. The order of the filtering steps is based on their computational efficiency and there are two different types of filtering cascades. The intermediary filtering stage uses fewer filters and is set up to let compounds through that could potentially be rescued in later stages, e.g., compounds are not filtered based on the number of hydrogen bond donors as these can be added or removed in the decoration stage. The goal of the intermediary filtering steps is to remove undesired compounds efficiently without removing promising compounds. The final filtering stage uses all available filters and the strictest settings so that only compounds that fit all the desired criteria are retained. Due to the ultra-large-scale exploration of chemical space, strict filtering is both advantageous and necessary in order to generate the best possible compounds. In the first step of the filtering cascade, compounds are filtered based on >20 physicochemical properties (e.g., MW, TPSA, logP etc.) and their overall complexity (e.g., number of chiral centers, spiro atoms etc.; see Supplementary Material S2-S4 for additional details). Additionally, compounds are also analyzed for their overall complexity compared to the starting ligand by looking at structural features, e.g., the number of chiral centers, spiro atoms, and all-carbon quaternary centers. The complexity filter reduces the number of compounds that would be too challenging to synthesize, independent from the complexity of the core and the overall physicochemical property space. The compounds that pass all of these criteria are progressed to the next filtering stage in which the ligands are filtered using a proprietary set of ∼1,500 SMARTS patterns to remove known and potential chemical liabilities, PAINS offenders, and other undesired chemotypes. Every AutoDesigner run explores different targets and areas of chemical space, and therefore the outputs of the runs are periodically reviewed by our in-house team of experienced medicinal chemists to flag compounds that are either chemically unstable, contain medicinal chemistry liabilities, or for which the synthesis is unprecedented. This resulted in continuous improvement of the algorithm and the SMARTS patterns used for substructure filtering over an eight month period. Although originally as many as 20% of compounds were manually flagged, in later runs less than 1% of the AutoDesigner-generated compounds were flagged by visual inspection (Figure 4). In the next step of the filtering process the ligands are evaluated for their FEP+ amenability, which is a key property required to ensure accurate FEP+ predictions. 21 Each ligand that makes it to this part of the filtering cascade is compared with relevant FEP+ references that are provided as part of the setup, and the heavy atom perturbation between ligand and reference is calculated. Based on FEP+ performance for the area of interest, a maximum heavy atom perturbation is set and the ligands are filtered accordingly. The ligands are then prepared for docking using the LigPrep module from Maestro (Schrӧdinger, Inc.: New York, NY, USA), and only ligands with a total and absolute charge equal to the desired charge state used in the docking and FEP+ models are retained. The state penalties (i.e., relative energies) of the different ionization and tautomeric states for the individual ligands are calculated and ligands with a high state penalty are removed to reduce false positives and irrelevant ligand states. The resulting library of ligands is then subjected to a number of optional filters that are fully configurable. If the ligands contain a titratable center in the newly designed R-group, then a pKa filter can be used to filter for ligands in the desired pKa range. This allows for precise tuning of the basicity and associated ADMET properties. Additionally, compounds can be filtered on their calculated logD values. Next, the filtered and prepared 3D structures are funneled into the docking stage. In this stage the ligands are docked in the receptor using a maximum-common-substructure-based docking model combined with optional hydrogen bond constraints. 22 Ligands that do not fit in the binding site and/or do not match the hydrogen bond constraints are discarded and the ligands that docked successfully proceed to the next stage. Following the docking stage there is an optional filtering step that analyzes the docked poses based on a pharmacophore model using a Phase hypothesis. 23,24 All the final ligands are then combined, deduplicated, and written to a ligand library file containing the docked poses. The output from the AutoDesigner algorithm can be triaged further by applying additional filters, e.g., machine learning models trained on relevant endpoints, and the final set of docked poses are used directly as input poses for active-learning FEP. 15,16 Relative Binding Free Energy Calculation protocol Protein-ligand bound X-ray structures were obtained via collaboration with Takeda Pharmaceuticals and processed with the Protein Preparation Wizard in Maestro 25 using default settings. All ligand structures were processed using LigPrep to enumerate all stereoisomers, and protonation states were assigned using Epik. 26 Possible tautomer and charge states were evaluated using Jaguar pKa 27 and corrected for in the Free Energy calculations. 28 OPLS3e 29 force field torsion parameters for each ligand were generated using the Force Field Builder. Prospective free energy calculations were run using Schrödinger FEP+ 7 with a simulation time of 20 ns, 24 λ-windows, and grand canonical Monte Carlo (GCMC) enhanced water sampling. 30 ## Generation of AutoQSAR machine learning models for Kp,uu and Clearance Machine learning models trained on experimental in vivo mouse PO Kp,uu (brain/plasma) and mouse IV plasma clearance (Cl,p) project endpoints were generated using AutoQSAR, an automated QSAR model builder application. 31 In brief, all experimental data measured at that particular point in time during the course of the project (45 Kp,uu and 54 Cl,p data points at initial model build), were separated into a training set and a holdout set using an 80/20 time-split. The top scoring models were then used prospectively as part of the AutoDesigner algorithm. Each model was retrained over time as more experimental data was collected. ## Implementation on a cloud computing platform The AutoDesigner algorithm has been implemented as a multi-stage algorithm which is, by design, very bursty in terms of compute resource utilization: it can typically require a compute capacity in the 2K-150K CPU cores range, with peaks that can last for relatively short amounts of time, at different stages of the algorithm. Therefore, we designed and implemented a bespoke infrastructure, deployed on a public cloud platform, that was optimized to run the AutoDesigner algorithm in a highly efficient manner. Such infrastructure has the ability to automatically expand and shrink its compute capacity according to the requirements of each and every stage. Having a fully elastic cluster allows for optimization of both cost and performance. On average an end-to-end AutoDesigner run takes 24 -48 hours. To prevent unnecessary workloads, the AutoDesigner algorithm always executes inline compound deduplication at the end of each and every generative stage. To remove all possible duplicates, the AutoDesigner algorithm executes an across-the-board deduplication stage after each series of generative stages. The deduplication stage merges all generated compounds into a single dataset, and removes duplicates from the entire dataset. We take advantage of a cloud-based, fully serverless data warehouse to perform this task. By ingesting the dataset into a table, and implementing the deduplication process as SQL-like code, we deduplicate from tens to hundreds of billions of compounds from a few minutes to a few hours in rare cases. Finally, we built automation code that can automatically provision the exact amount of required compute resources. The high performance computing (HPC) part relies on various compute node sizes to run computations. The memory to CPU ratio is 2 GB of RAM per CPU core. Based on the job requirements, one or more compute nodes will be provisioned on-demand, each one with the proper specifications, ranging from a lower bound of 8 CPUs / 16 GB per node, up to 64 CPUs / 128 GB. The cloud infrastructure operates with a 30 CPUs / 60 GB compute node size. The data warehouse part, instead, is completely serverless and autoscaled by design, and therefore does not require any type of pre-configuration other than creating the schema. ## Results and Discussion To efficiently explore relevant chemical space, it is critical to set the physicochemical property space so that it aligns with the overall project goals. In this case, the ranges of the physicochemical properties were set to match the desired CNS property space. A fine balance for these property ranges is required: too narrow, and the diversity and number of compounds generated is limited; too wide, and the exploration is costly and inefficient. The optimal physicochemical property space, along with a robust filtering cascade, allows for efficient and thorough exploration of millions of compounds while at the same time minimizing computational time spent on exploring irrelevant chemical matter. Here, we describe how AutoDesigner solved five program challenges by performing ultra-large-scale design within the property space defined by the project team (Table S2b). ## Generation of DAO inhibitors with AutoDesigner Statistics for three AutoDesigner campaigns carried out during this program are shown in Table 2, on a per-stage basis. For the AutoDesigner run based on compound 1, a total of 761M ligands were explored, and ultimately 515K compounds survived to the end of the process. For the run based on compound 5, a total of 353M ligands were explored and 15K passed through all filters. Finally, for the AutoDesigner campaign launched on the basis of compound 7, although 199M ligands were explored, only 156 were retained. In addition, for each of the three runs the CNS MPO 32 was computed for all surviving AutoDesigner compounds (Figure 5). For all three runs, the distributions were fairly tight and centered around median CNS MPO values of roughly 5. This value is greater than the cutoff of 4 which is generally considered a "good" score. 32 While some compounds did have CNS MPO values below this value, they represented a minority overall. Taken together, these data suggest that AutoDesigner produced compounds of high quality as measured by the CNS MPO metric, which was calculated subsequent to the AutoDesigner runs and not during the design process. ## Application of AutoDesigner to Identify Compounds with Moderate Lipophilicity One of the key parameters of an efficacious DAO inhibitor is adequate CNS exposure, which in part requires molecules with a moderate lipophilicity. 33 For this reason we decided to use AutoDesigner to explore the relevant physicochemical property space and generate compounds that introduce some additional polarity to offset the higher hydrophobicity of the tail region, while still maintaining other favorable properties of the starting ligand. Compound 1, consisting of a pyrazine dione head piece, a 2-carbon linker, and a trifluoromethyl phenyl tail piece, was used as the starting point. The AutoDesigner algorithm was used to explore the SAR of the aromatic tail region, and the 2-carbon linker and pyrazine dione were kept constant. The goal was to explore whether additional polarity could be introduced in the tail region without impacting other favorable properties and without a dramatic reduction in enzymatic potency. The algorithm consists of several expansion stages that generate hundreds of millions of compounds and thoroughly explore the desired property space. Each expansion stage is followed by a filtering cascade that filters ligands based on the desired physicochemical property space, removes undesired and/or unstable chemotypes, and ensures FEP+ amenability. Additionally, only the ligands that docked successfully were promoted to the next stage (see Figure 2 for the general algorithm). The filtering stage is crucial to the algorithm, as it ensures that only compounds that adhere to the predefined project criteria, do not introduce chemical liabilities, and fit within the binding site, propagate to later stages. Following the PathFinder reaction-based enumeration stage, and the subsequent filtering cascade, the ligands are passed on to the decoration stage, which decorates the designed ligands with R-groups (up to 15 heavy atoms). This decoration stage increases compound diversity and further explores chemical space. It also attempts to capture the "magic-methyl" effect. 34 In the example highlighted in Figure 6, the matched molecular pair (MMP) transformation stage converts the phenyl-CF 3 to an indole (Figure 6B), which passes all filters and fits in the binding site. In the next generative stage, the PathFinder reaction-based enumeration stage, the indole is converted into an indazole (Figure 6C). The decoration stage adds a CF 3 group to the indazole-containing ligand (Figure 6C) generated in the PathFinder stage, and compound 2, together with hundreds of millions of other compounds, enter the final, most comprehensive, filtering cascade. In the end, compound 2 (Figure 6D) and 515,000 other diverse compounds passed this final filtering cascade, confirming their FEP+ amenability, desired physicochemical property space, chemical stability, and ability to fit into the protein binding pocket. The most promising compounds that passed the final filtering cascade were evaluated on their CNS MPO score, FEP+ predicted potency, and predicted LLE. Based on these parameters compound 2 was one of the analogs that was synthesized, assayed, and confirmed as a potent inhibitor of DAO with a pIC 50 = 6.19, and an improved CNS MPO score (Table 3). In line with the specified design criteria, the final compound has a higher TPSA than the starting compound, while retaining potency. ## Rapid Design of Compounds to Accommodate Differences in Protein Conformation SBDD relies heavily on high resolution crystal structures of the target protein in order to design potent ligands. Initially, an unpublished proprietary X-ray structure was used to profile potential ligands with FEP+. Later on in the project we obtained a crystal structure of the protein bound to compound 1 and two significant differences were observed: (1) His217 moves slightly into the pocket, and (2) Ile223 swings out thereby creating a larger subpocket able to accommodate larger ligands. AutoDesigner was employed to quickly explore the binding site of the new crystal structure. Simply changing the docking model in the algorithm allows AutoDesigner to rapidly probe the larger binding site by exploring a large variety of possible R-groups that would fit in the additional binding pocket present in the new crystal structure. For instance, the algorithm explored an oxazolidinone substituent at the ortho position (Scheme 1, compound 3) that is able to fit into the new crystal structure but would clash with the Ile223 in the original crystal structure (Figure 7). The FEP+ predicted potency for both enantiomers of the oxazolidinone-substituted compound showed a distinct difference in potency for both enantiomers; pred. pIC 50 = 8.96 for (S)-3, and pred. pIC 50 = 5.10 for (R)-3. This potency of the racemic compound was confirmed experimentally after synthesis, and evaluation (3); pIC 50 = 6.31 Different crystal structures of the same protein will be able to accomodate ligands of different sizes and R-group vectors. With the help of AutoDesigner we are able to quickly probe the binding site of every crystal structure obtained for the target protein through the generation of large libraries of project-relevant chemical matter. ## Introduction of Polarity via Novel Water Mediated Interactions Besides fully exploring the protein's binding pocket, and exploring billions of compounds that have the potential to fit the desired physicochemical property space, such large-scale exploration of chemical space also has the added benefit of discovering unexpected and non-intuitive interactions with the receptor. The region of the DAO binding site that interacts with the tail section of the ligand is primarily hydrophobic, but to improve physicochemical properties we were interested to see if more polar groups could be accommodated in this region without sacrificing potency. In this particular example, one of the final compounds generated by AutoDesigner contains a benzothiadiazole R-group (Figure 8). The compound was predicted to have improved potency (FEP+ pIC 50 = 8.05), and improved LLE (6.35) while at the same time maintaining an excellent CNS MPO score (4.99). Though there are no additional hydrogen bond interactions with the protein, the predicted potency increases as both nitrogen atoms of the benzothiadiazole engage in a water network and the sulfur atom makes additional hydrophobic interactions. The compound was synthesized and assayed, and, at that stage of the project, compound 4 was one of the most potent compounds in the series (pIC 50 = 7.41). The tight integration of FEP+ into the AutoDesigner algorithm enables us to exhaustively explore chemical space and predict the most potent compounds, leading to the discovery of novel chemotypes with unexpected interactions and potency. ## Multiparameter Optimization of Multiple Experimental and Computational Endpoints Crucial for the lead optimization stage of the project was to design compounds that are not only potent, fit within the desired physicochemical property space and CNS MPO, but at the same time exhibit the desired ADME properties, particularly mouse IV plasma clearance (Cl,p < 40 mL/min/Kg) and brain exposure (Kp,uu > 0.1). In this AutoDesigner run, compound 5 was used as the starting ligand, consisting of a pyrazine dione head piece, a thioether linker, and a trifluoromethyl phenyl tail. Following the final filtering cascade, as described in Figure 2, over 15,000 unique, chemically stable, and drug-like ligands were generated that passed all filters and fit the binding pocket (See Table 2). In order to enrich for compounds with the requisite properties, machine learning models were generated to predict mouse IV plasma clearance (Cl,p), and the unbound partition coefficient (Kp,uu) using Schrödinger AutoQSAR and available project data. These machine learning models were applied to the AutoDesigner output and act as additional filtering stages at the end of the algorithm depicted in Figure 2. In total, 12,521 compounds had a predicted Cl,p < 40 and a predicted Kp,uu > 0.1. Additionally, compounds were evaluated on their closest Tanimoto similarity to competitor compounds (<0.5), CNS MPO score (>4), and FEP+ predicted potency. One of the AutoDesigner compounds that was synthesized based on these strict criteria was verified experimentally and met all the desired endpoints was compound 6 (Table 4). a FEP+ predicted pIC 50 in parentheses. In order to produce 12,521 ligands that satisfy the desired mouse IV plasma clearance (Cl,p), and unbound partition coefficient (Kp,uu) criteria, 353 million compounds were evaluated. Taking into account additional project criteria such as low Tanimoto similarity to competitor compounds, FEP+ predicted potency, LLE, and CNS MPO score, the final number of compounds that meet all of the desired criteria is much lower. These results highlight the challenges of the drug discovery process, and the need for ultra-large-scale exploration of chemical space in order to find the proverbial "needle in the haystack". ## Exploration of Novel DAO Subpocket to Improve Potency Based on the relatively limited SAR exploration of the DAO subpocket described in the literature, it was postulated that larger groups cannot be accommodated by the relatively narrow subpocket. Ligands with additional substitution on the phenyl ring suffered from a significant decrease in potency. 35 We decided to use the AutoDesigner algorithm to quickly and thoroughly explore whether this hypothesis was correct, and what type of groups could potentially fit within the subpocket. As a starting point for the AutoDesigner algorithm we selected compound 7, consisting of the pyrazine dione head group linked to a fused bicyclic ring system, and decided to explore the 7-position of the ring system. Introduction of the fused 2,3-dihydro-1,4-benzoxathiine ring led to a gain in potency compared to the uncyclized ligand (5) due to stabilization of the linker and tail piece conformation. The AutoDesigner algorithm designed and evaluated 199 million unique ligands, and ultimately 156 ligands passed the final filtering cascade, suggesting that this is a relatively difficult design problem. Even though the subpocket is relatively narrow, there are some compounds that were able to dock into this new subpocket. Not only was the AutoDesigner algorithm able to design compounds that fit all desired criteria, and fit in the new subpocket, but also FEP+ predictions indicated that potency can be improved by exploration of the new subpocket (Figure 9). Synthesis of one of the most promising AutoDesigner compounds in this series, compound 8, confirmed the predicted potency and challenged the hypothesis that substitution of the ring system reduces potency. This supports the argument that the AutoDesigner algorithm, and its ability to explore billions of compounds, is beneficial to understand the SAR of the binding pocket, and is a robust and efficient ligand design tool. Additionally, this produced the most potent compound in the project at the time (pIC 50 = 8.11). ## Conclusions We have recently seen improvements in compute power simultaneously matched by increases in predictive accuracy of computational methods. This has created the opportunity for accurate computational exploration and profiling of chemical space at a scale far larger than possible via traditional experimental approaches. To fully capitalize on these advances, we developed AutoDesigner, which employs a broad search algorithm and cloud computing to carry out successive rounds of chemical space exploration and filtering. We demonstrated the utility of AutoDesigner by applying it towards the design of DAO inhibitors. Over the course of the drug-discovery campaign, over one billion compounds were explored by AutoDesigner using only three ligands as starting points. This highlights the ability to effectively use AutoDesigner even in the absence of large sets of experimental data. Depending on the series being explored and the specific MPO criteria, it was observed that the size and nature of the chemical space could vary widely. This data helped the team strategically prioritize which compounds to pursue for synthesis. We have also found that AutoDesigner is most effective when run in close consultation with medicinal chemists. The first reason for this is that although the algorithm is designed to be synthetically aware at each chemical exploration stage (i.e., using previously published transformations, retrosynthetic enumeration, and small R-group decorations), the resulting molecules usually occupy a wide range of synthetic tractability ranging from simple to difficult. This is not a limitation of AutoDesigner per se, as the computational prediction of synthetic feasibility at scale is still an area of active research. 36 Nonetheless, in practice this means that collaboration with medicinal chemists to prioritize AutoDesigner compounds for synthesis is essential. Second, medicinal chemists play a key role in defining the permissible property space ranges for the AutoDesigner compounds. As noted above, one of the strengths of AutoDesigner is that the physicochemical ranges are fully customizable, allowing for design beyond the simple "rule of 5" cutoffs. 37 However, delineating these project specific ranges, which often change over the course of the program, benefits from the experience and analytic capabilities of experienced medicinal chemists to ensure that compounds with reasonable properties are designed. Thus, by working collaboratively, computational and medicinal chemists can utilize AutoDesigner to rapidly advance SBDD programs for challenging targets. Table S1. Classification criteria. These properties are discussed below. Note that NCC, NR, and HAC correspond to the number of chiral centers, number of rings, and number of heavy atoms respectively. Structures are assigned to a class by the last successful match on all criteria, proceeding from right to left.

chemsum

{"title": "AutoDesigner, a De Novo Design Algorithm for Rapidly Exploring Large Chemical Space for Lead Optimization: Application to the Design and Synthesis of D-Amino Acid Oxidase Inhibitors", "journal": "ChemRxiv"}

tailored_metastable_ce–zr_oxides_with_highly_distorted_lattice_oxygen_for_accelerating_redox_cycles

## Abstract: Ceria-based catalysts are widely used in oxidation or oxidation-reduction reactions in the field of environmental science. Their catalytic functions are determined by their ability to exchange oxygen species with oxidants. The enhancement of oxygen release is desired since it is often the ratedetermining step in redox cycles. Herein, we developed a lattice oxygen distortion method to enhance oxygen activation by quenching the Ce-Zr oxide nanoparticles formed from an extremely high temperature. This process can ensure the formation of solid solutions as well as avoiding atomic rearrangement during calcination, retaining the lattice oxygen at a metastable and disordered state without vacancies. Reduction, vacuum or metal deposition will easily induce oxygen release accompanied by vacancy creation. The metastable oxides can provide about 19 times more oxygen vacancies than traditional ones in a CO atmosphere. CO oxidation rates increased with increasing Zr content from 25 to 75% and achieved a new level, which is attributed to the acceleration of oxygen circulation via promoting oxygen release and supplying plenty of oxygen vacancies for redox cycles. This strategy is expected to be applied in the design and fabrication of improved oxygen storage materials. ## Introduction The development of three-way catalysts (TWCs) is desired due to the urgent requirement for eliminating the hydrocarbon, CO and NO x pollutants in automotive exhaust. Ceria-based composites as typical TWCs are directly involved in this oxidation-reduction catalytic process as "active" components via exchanging oxygen species with oxidants. 1 They are widely used due to their essential role in supplying active oxygen species for oxygen activation and their ability to act as an oxygen buffer by storing/releasing O 2 due to the Ce 4+ /Ce 3+ redox couple. 2,3 Take CO oxidation for example, this activation process follows the MvK mechanism, where active lattice oxygen on the surface of an oxide is removed by reaction with CO, and the oxygen vacancies (OVs) are subsequently replenished by reaction with O 2 from the gas phase. 4,5 Isotopically labeled oxygen studies have also proved that the oxygen atom incorporated in the reactant is directly extracted from the solid surface rather than derived from the gas phase. 6 This redox cycle can be enhanced by accelerating the oxygen release since it is often the ratedetermining step. Recently, Safonova et al. 7 proposed that the activity in CO oxidation is independent of the oxygen storage capacity (OSC) and Ce 3+ amount in the steady state. The release of oxygen atoms from the support which react with CO to form OVs and CO 2 has been identifed as the rate-determining step. The dynamic oxygen mobility or oxygen diffusivity is related more closely to the oxygen release rate in practical catalytic processes. Moreover, Haruta et al. 8 observed a clear volcano-like correlation between the CO oxidation activity of oxides and their metal-oxygen binding energy (E M-O ). Oxides with low E M-O , such as Fe 2 O 3 , may slow down the incorporation of oxygen from the gas phase into OVs, while for those with high E M-O , such as CeO 2 and ZrO 2 , it is hard to release oxygen. Therefore, decreasing E Ce-O may be an effective way to enhance oxygen activation and accelerate redox cycles. This fact provides signifcant motivation for intense research in developing techniques which could weaken the Ce-O interaction to create active oxygen and promote oxygen release. For instance, placing the smaller Zr 4+ (ionic radius of 0.84 ) at the position of the larger Ce 4+ (0.97 ) disturbs Ce-O interactions, leading to a decrease in the local Ce-O coordination number from 8 to 7 and to the formation of structural defects. 9 Moreover, the formation of heterotypic ceria could change the surface Ce-O interaction through different exposed surface planes. 10,11 The Ce-O interaction is also related to the metal-oxide interface, lattice strain and metastable t 00 phase. 5,12,13 However, weakened Ce-O interactions would lead to oxygen release and oxygen vacancy formation during calcination treatment in traditional catalyst preparation methods, resulting in a loss of active oxygen. Therefore, how to create active oxygen species as much as possible and at the same time keep them stable until reaction becomes an intractable problem. Flame spray pyrolysis (FSP) is a highly promising and versatile technique for the rapid synthesis of nanostructured materials without any subsequent calcination treatment. The combustion of aerosols at high temperature (ca. 2000 C (ref. 14)) followed by rapid quenching (ca. 60 C ms 1 (ref. 15)) produces highly dispersed nanosized oxide powders with metastable structures, which may not be easily accessible by conventional processes. 19,20 Herein, we design a series of metastable Ce-Zr oxides by employing the FSP method. A strong correlation between the oxygen distortion and the redox properties of the studied Au/Ce 1x Zr x O 2 catalysts for a CO oxidation probe reaction has been established. Rapid quenching could inhibit atomic rearrangement and make the oxygen atoms stable at a highly disordered state. These oxygen atoms are active and readily released in the reaction atmosphere, solving the intractable problem mentioned above. The target of this work is to reveal the role of oxygen distortion in oxygen activation and the key factors determining oxygen release. It is expected to explore the potential of metastable oxides for enhancing oxygen activation and provide a new strategy for accelerating redox reaction rates. ## Materials preparation The FSP samples were prepared by using the setup at the Karlsruhe Institute of Technology (KIT), which was described elsewhere. 21 Appropriate amounts of cerium(III) 2-ethylhexanoate (49% in 2-ethylhexanoic acid, 12% Ce) and zirconium(IV) 2-ethylhexanoate were dissolved in xylene to a fnal cerium and zirconium concentration of 0.5 M. The precursor was fed to the flame at 5 ml min 1 using a syringe pump (PHD Ultra™, Harvard) and dispersed by oxygen (feed rate 3.5 l min 1 and pressure 4.5 bar). The flame was ignited by premixed methane (0.6 l min 1 ) and oxygen (1.9 l min 1 ) issued from an annular gap. The product particles were collected on a watercooled glass fber flter (Whatman GF/D, 25.7 cm in diameter) with the help of a vacuum pump. A series of mixed oxides with different Ce/Zr molar ratios were synthesized by the above procedures and denoted as FSP-Ce 1x Zr x O 2 (x ¼ 0, 0.25, 0.5, and 0.75). Thermal treatment was conducted for the FSP-Ce 0.25 Zr 0.75 O 2 sample at 800 C for 30 h in air, which was then denoted as FSP-Ce 0.25 Zr 0.75 O 2 -C. For comparison, a series of the same ratio Ce 1x Zr x O 2 mixed oxides (denoted as CP-Ce 1x Zr x O 2 , x ¼ 0, 0.25, 0.5, and 0.75) were prepared by a traditional co-precipitation method (CP). Adequate amounts of ammonium cerium(IV) nitrate and zirconium(IV) nitrate were dissolved in deionized water to a fnal Ce and Zr concentration of 0.3 M. Then the (NH 4 ) 2 CO 3 solution (0.6 M) was slowly added into the above solution at 50 C with continuous stirring until the pH value reached 8-9. The solution was centrifuged and washed until pH ¼ 7.0. The obtained precipitate was dried at 110 C for 10 h and calcined in air at 500 C for 4 h. The Zr content was determined by SEM-EDS. Au was deposited onto supports via a deposition-precipitation (DP) method. 22 At room temperature, an adequate amount of HAuCl 4 $4H 2 O to obtain 1 wt% of gold was dissolved in 40 ml deionized water (6.0 10 4 M). The pH of the solution was adjusted to 9 by adding 0.1 M NaOH. Then 0.5 g support was added and the mixture was kept for 1 h under continuous stirring, maintaining the pH at around 9 via continuous NaOH addition. After that, the suspension was heated to 65 C and stirred for 1 h in a water bath. The precipitates were fltered and washed with deionized water to eliminate Cl . The obtained solid was dried at 60 C for 10 h and calcined in He (30 ml min 1 ) at 300 C for 2 h. ## Characterization Powder X-ray diffraction (XRD) patterns were recorded on an X'Pert Pro (PANAlytical) diffractometer with Cu Ka radiation at 40 kV and 40 mA over a 2q range from 5 to 80 . The crystal grain size was calculated from the FWHM of the strongest (111) reflection according to the Scherrer equation. Both lattice parameters and lattice strain were calculated using the Scherrer equation based on the (111) and (101) reflections in space groups Fm 3m and P4 2 /nmc, respectively. Electron paramagnetic resonance (EPR) spectra were recorded with a Bruker A200 EPR spectrometer at 100 K. Equal amounts of the catalysts were placed into a homemade quartz tube with stopcocks. Transmission electron microscopy (TEM) images were obtained with a JEM-2100 system (JEOL) with an acceleration voltage of 200 kV. The samples were ultrasonically suspended in ethanol and dropped onto a carbon flm supported over a Cu grid. Raman spectra were recorded on a commercial micro-Raman spectrometer (Renishaw, UK) using an Ar laser with a wavelength of l ¼ 514 nm working at 30 mW. In situ Raman spectra were obtained in a 2% CO/He or pure He (99.99%) atmosphere at 6 mW. The samples were heated from 25 to 300 C with a temperature ramp rate of 10 C min 1 . The data were collected after stabilization for 5 min at each temperature. Raman peak shifts and areas were determined by ftting to Lorentzian line shapes (R 2 > 0.95). X-ray photoelectron spectroscopy (XPS) was carried out to determine the surface compositions of the asprepared catalysts with an ESCALAB 250Xi instrument using Al Ka radiation (hn ¼ 1486.6 eV). The C1s peak was set at 284.6 eV and taken as the reference for binding energy calibration. Temperature programmed reduction with hydrogen (H 2 -TPR) was carried out in a homemade setup with a conventional U-shaped quartz reactor connected to a thermal conductivity detector (TCD). 50 mg samples were loaded and pretreated in Ar (30 ml min 1 ) at 300 C for 0.5 h. The TPR was performed in 5% H 2 /Ar (30 ml min 1 ) from room temperature to 900 C at a rate of 10 C min 1 . The loading of Au was determined by inductively coupled plasma optical emission spectroscopy (ICP-OES). The samples (10-20 mg) were chemically digested by dissolution in a mixture of 12 ml HCl and 4 ml HNO 3 in an autoclave at 160 C for 4 h. ## Reaction tests The CO oxidation reaction was executed in a fxed bed with 8 mm inner diameter under ambient pressure. A gas mixture containing 1.67 vol% CO, 3.33 vol% O 2 and 95 vol% He as balance was used with a flow of 37.5 ml min 1 for Au-based catalysts (25 mg) and 10 ml min 1 for the Ce-Zr oxide solid solution (50 mg). The CO conversion was measured in the temperature range of 25-100 C for Au-based catalysts and 100-450 C for the Ce-Zr oxide solid solution. The effluents from the reactor were analyzed using a gas chromatograph (Shimadzu GC-8A) equipped with a carbon molecular sieve column (TDX-1, Dalian Zhonghuida Scientifc Instrument Co. Ltd) and a thermal conductivity detector (TCD). At each reaction temperature, the CO conversion was calculated based on the composition of the product gas after 1 h stabilization. ## Results and discussion Properties of the metastable Ce-Zr oxide solid solution Here, a series of Ce 1x Zr x O 2 with a wide range of compositions (x ¼ 0-0.75) were prepared by both FSP and co-precipitation (CP) methods. XRD patterns shown in Fig. 1(a and b) reveal the single-phase nature of Ce 1x Zr x O 2 , varying from cubic to tetragonal structure with the Zr content. Fig. 1(c) displays a nearly linear relationship between the lattice parameters and the Zr content. The details can be seen in Table S1 in the ESI. † The observed change in the cell parameter with Zr contents is in agreement with Vegard's law and indicates the penetration of smaller Zr 4+ cations into the ceria cubic lattice resulting in the formation of solid solutions in the whole doping range x ¼ 0.25-0.75 by both FSP and CP methods. It is known that lattice strain, which represents a partially distorted lattice in a real crystal compared to the perfect structural model, originates from the lattice contraction and O 2 release followed by O 2 vacancy production during the substitution process of Ce 4+ ions by Zr 4+ ions. 12 Thus, as observed in Fig. 1(d), the increase of the lattice strain with the Zr dopant content for CP-made samples indicates the enhancement of oxygen vacancy formation in high Zr content samples. In contrast, a slightly decreasing tendency is found in FSP-made samples. This reveals that O 2 release and OV formation are suppressed by the rapid quenching process, making the structure still close to the perfect model. And this suppression is more pronounced with higher incorporation of ZrO 2 . This can also be seen from HRTEM images in Fig. S1 † that the FSP-made sample exhibits an intermediate of the agglomeration of cubic-shaped particles in a high temperature calcination process, and the CP-made one seems like the fnal product. 23,24 This reveals that, in the FSP process, the oxides remain in a transition state as a result of instantaneous high temperature decomposition without long-time thermal treatment. In general, the O 2 molecules are facile to be adsorbed at OVs and charged to form active oxygen species, such as superoxo (O 2 ) and/or peroxo species (O 2 2 ). 25 It has been reported that electron paramagnetic resonance (EPR) lines observed in the g range between 2.047 and 2.009 can be attributed to paramagnetic species O 2 . ions, which are capable of being transformed to O 2 molecules. 28 The EPR line with g ¼ 2.047 is assigned to the superoxide O 2 located in OVs. All of these signals appear in CP-made samples as shown in Fig. 2. In contrast, only a weak signal corresponding to OVs (g ¼ 2.047) is observed for the FSP-made CeO 2 samples. This signal gets weaker with Zr addition and fnally disappears, indicating that the FSP process dramatically suppresses the formation of oxygen vacancies especially for high Zr content. This result is consistent with the observed lattice strain tendency. ## Investigation of lattice oxygen behaviour As discussed above, the FSP-made Ce-Zr solid solution is in a metastable state without oxygen vacancies due to the quenching preparation process. The performance of oxygen release via breaking the metastable state is investigated through CO reduction, calcination, vacuum pretreatment and Au deposition. Generally, a single sharp peak at about 465 cm 1 (F2g symmetry mode) in the Raman spectrum is usually used to characterize the surroundings of Ce 4+ since it is very sensitive to oxygen sublattice disorder and nonstoichiometry. 29,30 It is observed from Fig. 3(A) that the F2g peaks broaden, diminish and become asymmetric with rising temperature for all tested samples. Simultaneously, a new broad feature appears at about 570 cm 1 , indicating the reduction of Ce 4+ to Ce 3+ and the formation of oxygen vacancies (OVs). 31 As we know, the precise assignment of OVs in Raman spectra is still controversial. It was reported that the peak at around 600-620 cm 1 was ascribed to OVs in Zr-doped CeO 2 in the literature. 32,33 Others assigned the OVs to 570 cm 1 in Raman spectra. In the present in situ Raman investigation of FSP samples, the peaks appear at around 570 cm 1 and increase steadily with the increase of reduction temperature, while the peaks at around 600 cm 1 remain unchanged. This is a strong indication of OVs at around 570 cm 1 . According to the literature, the Raman bands at ca. 600-620 cm 1 can be attributed to a nondegenerate Raman inactive Longitudinal Optical (LO) mode of ceria due to the substitution of zirconium into the ceria lattice. 38 The peak blue shift from 600 to 620 cm 1 can be observed with increasing Zr content, which may be attributed to the structural transformation from cubic to tetragonal. Weak bands at around 305 cm 1 can be attributed to the displacement of the oxygen atoms from their ideal fluorite lattice positions. 39 The enhancement of these bands with Zr content indicates the increasing disorder degree of oxygen atoms in the Ce 1x Zr x O 2 lattice. A new broad signal appears at around 260 cm 1 for Ce 0.25 Zr 0.75 O 2 samples and is assigned to the tetragonalization of the cubic structure. 32 According to the Raman spectra, tetragonalization is more strongly promoted by the inclusion of Zr in FSP-Ce 0.25 Zr 0.75 O 2 than in CP-Ce 0.25 Zr 0.75 O 2 . It can be seen from Fig. 3(B) that the amount of OVs created is much larger for FSP-made samples than for the corresponding CP-made ones, especially for the composition at x ¼ 0.75. And the increasing rates with rising temperature are faster for the former than for the latter. For example, the ratio slightly increases and reaches 0.77 at 300 C for CP-Ce 0.25 Zr 0.75 O 2 . By contrast, it sharply goes up to 4.24 for FSP-Ce 0.25 Zr 0.75 O 2 even at a relatively low temperature (200 C). This fnding demonstrates that Ce reduction and oxygen vacancy formation are more accessible for FSP-made Ce-Zr oxide solid solutions. The Ce-O interaction is remarkably weakened by the FSP method, producing considerable active oxygen species. And the proportion of active oxygen species created in lattice oxygen around Ce 4+ is more pronounced with increasing Zr content. It is also observed that the F2g signal of the FSP-Ce 0.25 Zr 0.75 O 2 sample becomes very small at 200 C and is barely observed on further increasing the temperature, indicating that most of the Ce 4+ ions on the surface can be reduced by CO under this condition. Note that no relations can be observed between oxygen release behaviors and the surface area which is listed in Table S1. † The oxygen release behaviors of CP-Ce 0.75 Zr 0.25 O 2 and FSP-Ce 0.75 Zr 0.25 O 2 samples are different although their surface areas are the same. Moreover, the amount of oxygen release increases with Zr content for FSP samples, while they possess the same surface area. Therefore, we suppose that the oxygen release behavior depends on the activity of lattice oxygen rather than the redox in the bulk phase. The catalytic performance in Fig. S2 † exhibits that the formation of active oxygen species cannot be enhanced to infnity via increasing the Zr content since the Ce content is decreasing at the same time. In FSPmade samples, the amount of active oxygen species around Ce 4+ is fnally restricted by the Ce content, leading to a limited formation of OVs. Peak shifts of the F2g mode for different gas flows and temperatures for FSP-and CP-made Ce 0.25 Zr 0.75 O 2 samples are shown as a bar diagram in Fig. 3(C). It shows that the F2g mode positions in He decrease with increasing temperature for both the tested samples. This red shift, which depends on temperature, is mostly attributed to thermal expansion as well as to phonon coupling and decay. 41 Compared to the CP-made samples, the F2g mode positions of FSP-Ce 0.25 Zr 0.75 O 2 at different temperatures are much lower due to mode softening, illustrating the enhancement of oxygen disorder. High static positional disorder can enhance the diffusivity and mobility of oxygen ions, 42 facilitating oxygen release. In addition, the bar reveals the change of the chemical environment from He to 2% CO/He atmosphere. The offsets of the F2g for the FSP-made samples are much larger, exhibiting 8.2, 10.7 and 11.2 cm 1 at 20, 100 and 200 C, respectively, compared to those for CP-made ones (0.5, 0.9 and 1.0 cm 1 corresponding to the same temperatures as above). It is reported that this additional downshifting with the CO flow is caused by lattice expansion or chemical strain. 41 It can be represented by Du ¼ gu(DV/V 0 ), where g is the Grüneisen parameter (1.24), 43 and DV is the volume change from the reference case (He atmosphere) volume V 0 . This fractional volume change of Ce 0.25 Zr 0.75 O 2d is a consequence of oxygen defcit d formation, i.e. expansion due to the replacement of small Ce 4+ (0.970 ) ions by large Ce 3+ (1.143 ) ions, which is partially offset by effective compression owing to the loss of large O 2 (1.380 ) and the creation of small OVs (1.164 ). The oxygen defcit can be calculated by using the equation: d ¼ 2.66(Du/u). 41 The d values of the two samples, also shown in Fig. 3(C), exhibit an increasing trend with increasing temperature. The oxygen defcits of the FSP-Ce 0.25 Zr 0.75 O 2d sample are 0.0467, 0.0607 and 0.0622 at 20, 100 and 200 C, respectively, which are more than one order of magnitude larger than those of CP-Ce 0.25 Zr 0.75 O 2d (0.0025, 0.0048 and 0.0057 at the same temperatures, respectively). It is worth noting that the amount of oxygen defcit of the FSP-made Ce-Zr oxide solid solution produced in the reductive atmosphere is extremely high, exceeding that of CeO 2 rods with Au deposition at 300 C (d ¼ 0.039), 41 considering that the rod shape, the existence of Au and high temperature are all positive factors for enhancing oxygen defcits. This demonstrates that the FSP preparation method is effective in softening the Ce-Zr oxide solid solution, enhancing the Ce 4+ reducibility and facilitating the creation of large amounts of oxygen defcits. Thus, the oxygen release in the redox cycle is strengthened. The formation of Ce-Zr oxide solid solutions with high oxygen activation ability in one step via the FSP method may be attributed to instantaneous high temperature and rapid quenching. The precursors directly decompose into oxide crystals in the flame at high temperature, and then the oxide particles quickly get away from the flame area. High temperature facilitates the formation of the crystal structure. Rapid quenching inhibits atomic rearrangement and the formation of strong metal-oxygen interactions. It also makes them stay in an intermediate state, leading to the O atoms lacking restriction and being released more readily. Therefore, long-time thermal treatment is executed in our work on the FSP-Ce 0.25 Zr 0.75 O 2 sample to break this state. From Fig. 4(A), it can be seen that a large amount of OVs is formed in FSP-Ce 0.25 Zr 0.75 O 2 at low temperature before the thermal treatment, indicating that the active oxygen atoms are easy to release and react with CO. However, no obvious OVs can be observed below 200 C in the calcined sample, which is similar to the situation for CP-Ce 0.25 Zr 0.75 O 2 . A sharp decrease of the peak area ratio (A Ov /A F2g ) from 4.24 to 0.85 at 200 C is found after the thermal treatment from Fig. 4(B). It can also be seen from Fig. 4(C) that the F2g peak positions in the 2% CO/He atmosphere increase from 456.3 to 467.2 cm 1 at 200 C, and their deviation from those in the He atmosphere diminishes from 10.9 to 0.95 cm 1 . Meanwhile, the oxygen defcits decline from 0.0622 to 0.0054, representing that mode hardening happened during the thermal treatment. This makes the Ce-O bonding energy of FSP-Ce 0.25 Zr 0.75 O 2 increase and get close to that of CP-Ce 0.25 Zr 0.75 O 2 , leading to a decrease of oxygen disorder degree. This also proves that the FSP-made Ce-Zr solid solutions stay in a metastable state owing to the rapid quenching. The crystals are produced and get away from the high-temperature zone in ultrashort time before the interactions among the atoms are strengthened. Subsequent long-time thermal treatment will transform the chemical bonds from metastable to stable states, leading to relatively hard oxygen release and low oxygen activation ability, which is similar to that observed in CP-made samples. The influence of thermal treatment on the F2g peak is observed from Fig. 4(D and E) by narrowing the line, increasing its symmetry and moving it to a higher Raman shift, which may be related to the rearrangement of atoms during the calcination process. Note that no obvious variations can be detected in morphology, particle size or phase from Fig. S3 and S4. † The effect of particle size on the Raman peak position can be excluded. 44 Compared to fresh CP-made ones, about 5 to 13 cm 1 (Table S2 †) lower F2g peak position is found with Zr content increasing from 25 to 75% for fresh FSP-made samples. However, these gaps disappear, i.e., the peak positions are getting close to those of the fresh CP-made ones after thermal treatment at 800 C. The F2g downward shift and the presence of a new peak at 258 cm 1 of CP-Ce 0.25 Zr 0.75 O 2 indicate the phase transformation. 12 The degree of oxygen disorder can be strengthened by the quenching process in FSP. And this effect is more pronounced in the samples with high Zr content and can be diminished through high temperature thermal treatment. XPS studies are performed in order to analyze oxygen release and OV formation during vacuum treatment via measuring the amount of surface Ce 3+ /Ce ratio. In Fig. 5, XPS Ce 3d signals of catalysts show eight peaks corresponding to the spin-orbit doublet (SOD) and satellite lines characteristic of Ce in the oxidized state. 45 It is clearly shown in Table S3 † that Ce 3+ /Ce surface atomic ratios of FSP-made catalysts are higher than those of the corresponding CP-made ones. In particular, the ratios of FSP-Ce 0.25 Zr 0.75 O 2 and CP-Ce 0.25 Zr 0.75 O 2 are 0.24 and 0.17, respectively. According to EPR results, no sign of oxygen vacancy existence can be found in FSP-Ce 0.25 Zr 0.75 O 2 . It can be supposed that about 24% of Ce 4+ turn into Ce 3+ due to oxygen release during vacuum treatment, far exceeding that observed in the CP-made one. This indicates that the surface lattice oxygen is more readily released for the former than the latter. Fig. 6(A and B) show the effect of Au deposition on the Raman spectra of Ce-Zr oxide solid solutions. In contrast to the calcination effect, Au deposition will broaden the F2g peaks and lead to a red shift ranging from 8.2 to 12.6 cm 1 and from 12.6 to 29.6 cm 1 for CP-and FSP-made samples, respectively, as shown in Table S4. † Au deposition has a more remarkable effect on the enhancement of oxygen distortion for FSP-made catalysts by weakening the Ce-O interaction which makes the oxygen atoms more accessible. In particular, a clear Raman shift from 461.7 to 432.1 is observed for the Au/FSP-Ce 0.25 Zr 0.75 O 2 catalyst. At the same time, a Raman peak at 570 cm 1 , which is assigned to OVs, appears accompanied by a decrease of the F2g peak area. This phenomenon is also observed from EPR as shown in Fig. 6 Although Au deposition can dramatically enhance the surface Ce 4+ reduction, a large portion of Ce 4+ is still reduced at high temperature in CP-made samples. Therefore, we propose that the oxygen release and the creation of OVs preferentially happen at the interface between Au and the support. The OVs created at the interface are the active sites in the CO oxidation reaction due to the promotion of oxygen mobility. 47,48 From previous analysis, the schematic diagram of oxygen behaviour is proposed in Fig. 7. The distorted oxygen atoms of metastable Ce-Zr oxide solid solutions are released readily in S3, † respectively. the reduction atmosphere or under vacuum treatment, in situ producing plenty of OVs and supplying a considerable amount of active oxygen species. They can accelerate the redox cycles and enhance the oxygen activation. Besides, the creation of OVs preferentially happens at the interface between Au and the support during Au deposition, which promotes the oxygen atom mobility and their reaction with the activated CO molecule. Therefore, it is supposed that the redox properties are determined by the rate of oxygen release rather than the existing oxygen vacancies. However, long-time thermal treatment at high temperature (above 800 C) can break the metastable state. The distorted oxygen atoms will rearrange with OV formation during calcination. This could bring the oxides into a stable state, where it is then hard to supply more active oxygen in the process of reaction. ## Catalytic performance in CO oxidation It is reported that the optimal catalyst composition has been analysed by a temperature scanning method (TSM) based on the experimental data of CO oxidation for a series of Au/Ce 1x Zr x O 2 (x ¼ 0, 0.25, 0.5, 0.75, and 1) catalysts and identifed to be x ¼ 0.25. 53 Regular behaviour has been reported that Ce-Zr oxide solid solutions with extremely high Zr content (higher than 60%) exhibit low redox activity due to the formation of a non-defective t phase. 12 The CO oxidation reaction rates over Ce-Zr oxide solid solutions reported in the literature and in our work are summarized in Table 1. It can be found that the CO reaction rate obviously increases with the Zr content for the FSPmade samples, which is in contrast to that for CP-made ones and has never been reported before. The activity of the Au/FSP-Ce 0.25 Zr 0.75 O 2 sample is much higher than that of the most active ones employed in the literature, including allotropic ceria and Ce-Zr oxide solid solutions. Generally, Zr addition leads to structure relaxation in the lattice, decreasing the oxygen vacancy formation energy 54 and promoting the oxygen mobility. However, the Ce-O bond length becomes shorter with Zr doping into the CeO 2 framework, which could suppress oxygen release. 55 Simultaneously, the amount of oxygen bonding with Ce decreases with increasing Zr content. This can explain why the Zr content always has an optimum value. The usage of the quenching method for producing Ce-Zr oxide solid solutions could weaken the Ce-O interaction and activate oxygen atoms around Ce, which compensates the negative effect of Ce reduction. This makes its oxygen activation capability stay at a high level even for low Ce content. The activation energy of CP-and FSP-made Au/Ce 1x Zr x O 2 has been tested in the CO oxidation reaction while keeping the CO conversion below 20%. The activation energy of all samples is almost unchanged, fluctuating in the range of 39.1 to 41.1 kJ mol 1 . This indicates that the reaction mechanism is the same for both catalysts, including the same rate-determining step and similar energy potential diagrams. The improvement in the catalytic activity of CO oxidation lies in that the FSP catalyst can supply more active sites than the CP one. ## Conclusions Rapid quenching from extremely high temperature can freeze Ce-Zr oxide solid solutions in a metastable state with highly disordered oxygen, which results in an enhancement of the capability for oxygen release. A reduction atmosphere can induce 19 times more oxygen defcit formation from FSP-Ce 0.25 Zr 0.75 O 2 than the corresponding CP-made one at room temperature. The lattice oxygen atoms readily detach from the constraint of metals in the vacuum treatment, producing large amounts of OVs. Besides, during Au deposition, the creation of OVs preferentially happens at the interface between Au and the support. CO oxidation reaction rates obviously increase with the Zr content for Au/FSP-Ce 1x Zr x O 2 , which is in contrast to conventional CP-made samples and has never been reported before. The quenching method used in our work has proved to be effective in enhancing oxygen release and accelerating redox cycles. Foreseeably, this new concept can be widely employed in many oxygen activation reactions in heterogenous catalysis. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Tailored metastable Ce\u2013Zr oxides with highly distorted lattice oxygen for accelerating redox cycles", "journal": "Royal Society of Chemistry (RSC)"}

d-sorbitol_can_keep_the_viscosity_of_dispersive_ophthalmic_viscosurgical_device_at_room_temperature_

## Abstract: The combination of 3% sodium hyaluronate (HA) and 4% sodium chondroitin sulfate (CS) is used as a dispersive ophthalmic viscosurgical device (OVD) during cataract surgery. For most OVDs containing HA, storage at 2-8 °C is recommended to preserve product characteristics. In order to develop a dispersive OVD that can be stored at room temperature, in this study, we searched additives which can stably maintain the viscosity, a key parameter of OVD, under preservation stability testing at 60 °C. The addition of D-sorbitol to a combination OVD, 3% HA and 4% CS, suppressed the reduction in viscosity compared with other OVDs with or without additives. The addition of D-sorbitol was also effective in improving the residual viscosity of the combination OVD after thermal treatment and light irradiation. Moreover, the OVD containing D-sorbitol can be stored at 25 °C with stably maintaining the initial viscosity for at least 24 months. In conclusion, the new dispersive OVD, 3% HA, 4% CS, and 0.5% D-sorbitol, can be stored at room temperature instead of under cold conditions and may represent an attractive option for clinical use because it is not necessary to bring the product to room temperature prior to use. Sodium hyaluronate (HA) and sodium chondroitin sulfate (CS) are glycosaminoglycans (GAGs), which are widely distributed in the extracellular matrix and on the cell surface of animal tissues 1,2 . HA in particular represents a major lubricating component of the extracellular matrix and has been suggested to enhance sliding between adjacent tissue layers 3 . Unlike most tissues in the eye, the cornea does not contain any blood vessels for nourishment or protection. When corneal endothelium is damaged by mechanical injury or bubble formation during phacoemulsification and aspiration (PEA), the endothelial cells cannot regenerate 4,5 . In such circumstances, ophthalmic viscosurgical devices (OVDs) are indispensable tools for the protection of corneal tissues. Various kinds of HA-based products have been commercialized as OVDs for use in cataract surgery, intraocular lens implantation, and penetrating keratoplasty 6 . OVD products are classified into two major types, cohesive and dispersive, based on the physical properties of their viscoelastic materials 7 . The cohesive OVDs, e.g., Healon ® (Abbott Medical Optics Inc., CA, USA) and OPEGAN Hi ® (Santen Pharmaceutical Co., Ltd., Osaka, Japan), containing 1% HA with an average mass of >2000 kDa, show high cohesive properties . These higher-viscosity cohesive agents are used to protect intraocular tissues from invasive surgical instruments and intraocular lenses during surgery by maintaining a deep anterior chamber 10 . However, this type of OVDs tends to easily flow out of the eye during PEA. The dispersive OVDs, such as Viscoat ® (Alcon Inc., Hünenberg, Switzerland), a combination OVD consisting of 3% HA and 4% CS, are strongly retained in the anterior chamber during anterior segment surgery because of their adhesive nature in comparison with cohesive OVDs 11 , and thus display excellent corneal endothelial protection during PEA 12 . The disadvantages of dispersive OVDs are that it is insufficient to maintain space and is relatively difficult to remove. Additionally, the residual OVD may cause increase of postoperative intraocular pressure 13 . Therefore, if a dispersive OVD is used during cataract surgery, it is recommended to use cohesive OVD at the same time. Both types of OVDs containing GAGs are relatively unstable against light irradiation and thermal treatment. More specifically, such treatments accelerate the depolymerization of GAGs, which reduces OVD viscosity 14,15 . Therefore, to maintain the important viscosity as rheological properties of these OVD products need to be stored in a refrigerator away from light until use. In consequence, prior to use, the temperature of such OVDs needs to be brought back to room temperature (preferably 15 °C to 25 °C), which may take 20 to 40 min 16 . Generally, physico-chemical degradation of HA can occur in one of the five following manners: acid or alkaline hydrolysis, ultrasonic, thermal, oxidation, and photo-degradation 17 . Light irradiation and thermal treatment can accelerate the generation of free radicals, which can elevate oxidative stress and damage GAGs, leading to their depolymerization. For achieving long-term storage of the OVD products at room temperature, formation of oxygen-derived free radicals, which facilitate oxidation including photosensitized oxidation, should be suppressed as far as possible 18,19 . In the present study, we evaluated the effect of various additives on the rheological properties of OVD products. Based on our observations, we have developed a new dispersive-type OVD, Shellgan ® (Santen Pharmaceutical Co., Ltd., Osaka, Japan) containing 3% HA, 4% CS, and 0.5% D-sorbitol. The addition of D-sorbitol to conventional OVD helped maintain stable OVD viscosity compared with additive-free conventional OVD. Subsequently, the D-sorbitol-containing OVD can be stored at room temperature for at least 24 months without reducing its viscosity. ## Results Selection of the additive by preservation stability testing at 60 °C. We evaluated the effect of various additives, i.e., D-sorbitol, glycine (Gly), L-glutamate (Glu), monosodium L-glutamate (MSG), L-methionine (Met), D-glucose (Glc), maltose monohydrate, xylitol, and D-α-tocopherol, on the viscosity of the basal OVD (Sample a) as shown in Table 1. Table 2 shows the time-dependent change in viscosity of the OVD samples stored at 60 °C. Initial viscosity of each sample with additive showed almost the same value as that of the basal OVD (sample a). The viscosity of the basal OVD reduced in a time-dependent manner. The viscosity measured after 14 days was 33.2% of the initial value. Upon adding sugar alcohols to the basal OVD (samples b and i), the progression of their viscosity reduction was decelerated. Particularly, the OVD containing 0.5% D-sorbitol showed viscosity of 25 www.nature.com/scientificreports www.nature.com/scientificreports/ storage at 60 °C for 14 days. By adding D-sorbitol (sample b), the residual viscosity was improved, with the final viscosity being up to 50.5% of the initial value. Addition of Met, Glc, maltose, or D-α-tocophenol had no effect on viscosity (samples f, g, h, and j) and addition of amino acids, except Met, strongly accelerated the viscosity reduction (samples c, d, and e). Further, adding D-sorbitol did not accelerate browning of the basal OVD, judged by visual observation. Based on these results, D-sorbitol was selected as the best additive among the ones tested, and further detail examinations were performed using the new composition containing D-sorbitol. Effect of D-sorbitol concentration on the stability of the basal OVD against heat and light. We evaluated the effect of various concentrations of D-sorbitol on the viscosity reduction of basal OVD induced by thermal treatment or light irradiation. Table 3 shows that the residual viscosity of the basal OVD was 49.05% of the initial value after autoclaving and 85.18% after photostability testing (sample a). Light shielding by using an aluminum foil was effective in preserving the viscosities of all the samples in the photostability testing, although a slight reduction in viscosity was observed in every OVD. The addition of D-sorbitol improved the photostability of the OVD as much as light-shielding. No conspicuous difference in viscosity loss was observed between the 2 D-sorbitol concentrations tested. In thermal stability testing, similar to photostability testing, viscosity loss was observed upon addition of D-sorbitol, but no difference in effect was observed for different D-sorbitol concentrations. Therefore, further studies were performed using OVD containing 0.5% D-sorbitol. Long term preservation stability of the new composition at 5 °C or 25 °C. We evaluated whether addition of D-sorbitol could improve the long-term stability of basal OVD, and the results are shown in Fig. 1. Viscosity reduction of both OVDs was negligible, independent of D-sorbitol addition, over 24 months of preservation testing at 5 °C; their residual viscosity after 24 months was over 87%. When the OVDs were stored at 25 °C, the viscosity of the basal OVD reduced in a time-dependent manner, the residual viscosity at the end of the test became less than 67%, and the value dropped below 35 Pa•s. By adding D-sorbitol, the final viscosity of the basal OVD stored at 25 °C was maintained between 55 and 47 Pa•s for 24 months. The residual viscosity of the new composition at the end-point was over 87%, indicating that addition of D-sorbitol improved the long-term stability of the basal OVD enough for storage at room temperature, as effectively as low-temperature storage. We evaluated the effect of various additives on the viscosity stability of the basal OVD. As shown in Table 2, sugar alcohols were effective in preserving the viscosity of the basal OVD; D-sorbitol was the best additive among the additives tested. The viscosity of OVD containing Gly, Glu, or MSG was extremely reduced after preservation at 60 °C for 2 weeks. Maillard products were generated in those OVDs, owing to an alteration in their color from colorless to pale brown. According to Deguine et al., free radicals generated in the Maillard reaction decrease the viscosity and molecular weight of HA during 1 h of incubation at 37 °C21 . We also demonstrated depolymerization of HA in the OVD by adding amino acids in Fig. 2. Therefore, the drastic viscosity reduction by adding amino acids to the basal OVD was due to the depolymerization of HA caused by the generation of free radicals in the Maillard reaction. The OVD containing D-sorbitol maintained viscometric property highly compared with the basal OVD, against both thermal treatment and light irradiation. It is well known that the generation of free radicals is enhanced by light irradiation or with thermal dependence. The degradation of HA caused by exposure to reactive oxygen species was reported to be inhibited in a concentration-dependent manner by D-mannitol 22,23 . Hence, the antioxidant capacity of D-sorbitol might be one possible mechanism to preserve the viscosity of the basal OVD over a series of preservation tests. L-methionine was not able to suppress the viscosity reduction of the basal OVD in our study, though it was reported that L-methionine preserved the viscosity of HA after exposure to a myeloperoxidase-derived oxidant 24 . Therefore, other mechanisms might be responsible for the viscosity reduction of HA/CS combination products. Figure 3 shows that, surprisingly, the profiles of the molecular weight reduction of HA in both the basal OVD and OVD containing D-sorbitol were hardly different through the preservation testing. These results suggested that D-sorbitol might possess mechanisms other than antioxidant activity to preserve the viscosity of the OVD. The viscosity of the combination of HA and CS significantly increases compared with the sum of the viscosity of HA or CS alone, owing to the mutual interaction of CS and HA to form aggregates via hydrogen bonding 25,26 . D-sorbitol might cause stiffening due to the interaction between HA and CS. Few reports demonstrate that sugar alcohols preserved the viscosity or molecular weight of HA in artificial oxidative conditions. However, there are no reports on D-sorbitol affecting the long-term stability of HA/CS combinations. To fully exploit the abilities of OVDs during surgery, it is desirable to have a sufficient viscosity. For long-term product performance, it is important to manage the preservation temperature in many HA products. Therefore, many distributors recommend that surgeons store the OVD products in a cold place like a refrigerator. Before using the refrigerated OVDs, surgeons should bring the product temperature back to room temperature, as the viscosity depends on the temperature: if the OVD is used when cold, the therapeutic efficiency and usability may be affected. This is the first study to investigate the effects of D-sorbitol on the long-term stability of OVD viscosity. Figure 1 shows that viscosity reduction in the new composition was hardly observed during 24 months of preservation testing at 5 °C, including for the basal OVD. The transition in the viscosity of the new composition stored at 25 °C was maintained in the range of 55 to 47 Pa•s during the testing. The viscosity loss of the new composition stored at 25 °C for 24 months was very similar to the result obtained when the new composition was stored at 5 °C (87%), indicating that D-sorbitol allows the storage of OVDs at room temperature for 24 months. Five HA products containing sugar alcohols have been launched: Synolis VA ® (Aptissen SA, GE, Switzerland), GO-ON ® matrix (Meda AB, Solna, Sweden) and Ophteis ® FR Pro (Rayner Ltd., WSX, United Kingdom), all of which contain 4% D-sorbitol , and Ostenil ® and Visiol ® (TRB Chemedica Int. SA, GVA, Switzerland), both of which contain 0.5% D-mannitol 23,31 . Several reports have demonstrated that D-mannitol or D-sorbitol can improve the effectiveness of intra-articular HA products in the treatment of osteoarthritis without any serious or unexpected effects 22,28,29 . Therefore, we believe that the new combination developed in this study would aid ophthalmic surgeons in administering timely treatment. We demonstrated that adding D-sorbitol to a HA/CS combination can preserve its viscosity at room temperature. Thus, we developed a new OVD formulation, termed Shellgan ® , which is the first dispersive OVD formulation that can be stored at room temperature for at least 24 months (Fig. 4). We believe that the new dispersive OVD product has the potential to be clinically effective by maximizing therapeutic efficacy concomitantly with readiness for intraoperative emergencies, because this product can maintain its viscosity for a long time even when stored at room temperature. ## Materials. Purified HA from chicken combs and CS from shark cartilage were obtained from Seikgaku Corp. (Tokyo, Japan). D-sorbitol, glycine, L-glutamate, monosodium L-glutamate, L-methionine, D-glucose, maltose monohydrate, xylitol, and D-α-tocopherol were purchased from Wako Pure Chemical Industries Ltd. (Osaka, Japan). ## Sample preparation. To prepare basal OVD, powdered HA and CS were dissolved in phosphate-buffered saline to obtain final concentrations of 3% and 4%, respectively (Table 1, a). To evaluate the effect of various additives on the viscosity stability of basal OVD, we selected 9 additives, some of which showed antioxidant capacity determined by 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox) equivalent antioxidant capacity assay (data not shown). Each additive was added to the basal OVD to obtain a final concentration of 0.5% (Table 1, b-j). To evaluate the effect of additive concentration, a sample containing 1.0% D-sorbitol was also prepared (Table 1, k). Every sample was filtered through a 0.22-μm membrane filter before use. ## Determination of molecular weight distribution. The molecular weight distribution of sample was determined by high-performance liquid chromatography (LC Prominence; Shimadzu Co., Ltd., Kyoto, Japan) using a size exclusion column (OHpak SB-806M HQ; Shodex, Co., Ltd., Tokyo, Japan) and a reflective index detector (RID-10A; Shimadzu Co., Ltd.). The flow rate was 0.3 mL min −1 of 0.5 M NaCl at 35 °C. Each sample was

chemsum

{"title": "D-sorbitol can keep the viscosity of dispersive ophthalmic viscosurgical device at room temperature for long term", "journal": "Scientific Reports - Nature"}

the_synthesis_of_1,4-anhydro--d-mannopyranose

## Abstract: We report the first synthesis of 1,4-anhydro--D-mannopyranose. Several unsuccessful approaches were attempted, including azide-mediated cyclization of a 1-O-acetyl-4-O-mesyl derivatives of -D-talopyranose, and N-iodosuccinimide/triflic acid cyclization of a 4hydroxy thiomannoside. Ultimately, base-mediated intramolecular nucleophilic substitution of 2,3,6-tri-O-benzyl--D-mannopyranosyl chloride successfully provided the cyclized product in low yield, which could be deprotected by hydrogenolysis to afford the title 1,4anhydrosugar. ## INTRODUCTION Anomeric anhydrosugars are intramolecular acetals formally derived by the loss of a water molecule from the parent sugar. 1,2 1,2-Anhydropyranoses have been widely exploited in glycosylation chemistry 3 and have been implicated as intermediates in enzyme-catalyzed reactions. 4 1,6-Anhydropyranoses occur in nature both as products of enzymatic reactions and through pyrolysis of cellulose. 2 These compounds have been widely studied owing to their interesting and useful attributes in synthetic carbohydrate chemistry. 2, 1,4-Anhydropyranoses (which may be considered 1,5-anhydrofuranoses) have received relatively less attention, although have been proposed as intermediates in enzymatic transformations, 8 and have had only sporadic investigations into their synthesis and transformations. Studies on this class of sugars experienced an inauspicious start when in 1928, Freudenberg and Braun reported that treatment of 2,3,6-tri-O-methyl--D-glucosyl chloride 1 with sodium in ether afforded the protected 1,4-anhydro--D-glucopyranose 2 (Figure 1, Method A), 14 a result that could not be replicated by Hess and Littmann in 1933. 15 Instead, Hess and Littmann, 15 and later Kops and Schuerch 16 reported that treatment of 2,3,6tri-O-methyl-4-O-tosyl--D-glucose 3 with sodium isopropoxide afforded the protected 1,4anhydro--D-galactopyranose 4 (Figure 1, Method B). These two studies illustrate two of the main approaches used for the synthesis of this class of compounds, namely base-promoted intramolecular nucleophilic substitution by the 4-hydroxyl on an anomeric halide, or basepromoted intramolecular nucleophilic substitution by the anomeric hydroxyl on a sugar bearing a suitable leaving group at the 4-position. 17,18 A third approach, related to the first, involves Lewis acid-catalyzed transacetalization of a 4-hydroxy glycoside 9 or Lewis acidcatalyzed intramolecular glycosylation reactions, as exemplified by Thiem and Weisner through the BF3.Et2O-catalyzed reaction of 2,3,6-tri-O-benzoyl--D-galactopyranosyl fluoride 5 to afford the protected 1,4-anhydro--D-galactopyranose 6 (Figure 1, Method C). 19 A fourth approach involves the intramolecular addition of the 4-hydroxy group to a glycal, promoted by a suitable electrophile, such as in the transformation of 7 to 8 (Figure 1, Method D). 20,21 As the electrophile adds the alkene, this method yields a 2-substituted 1,4-anhydro-2deoxy sugar, and is not suitable for the synthesis of 1,4-anhydro sugars that are oxygenated at C2. In this work we report efforts, which were ultimately successful, to prepare 1,4anhydro--D-mannopyranose 9 (Figure 2). The synthesis of this anhydro sugar has not been reported previously, although it has been invoked as an intermediate in the solvolysis of -Dmannopyranosyl fluoride. ## RESULTS AND DISCUSSION Brimacombe and co-workers 23 reported the synthesis of 1,4-anhydro-2,3-O-isopropylidene--L-rhamnopyranose 10, which was prepared by utilizing an approach based on that outlined in Method B of Scheme 1, however they used sodium azide 24,25 to displace the anomeric acetyl group of 11 to generate an anomeric alkoxide (Figure 3). As this represents the only literature example of a 1,4-anhydrosugar bearing exclusively endo substituents our initial efforts towards 1,4-anhydro--D-mannopyranose were dedicated to a similar approach. As it is unlikely that the 1,4-anhydro system would survive the conditions needed for isopropylidene deprotection, 12 and as it has been shown that non-anomeric ester groups are stable to the reaction conditions, 24 we chose to employ ester protecting groups. Methyl talopyranoside 12 was prepared in 5 steps from methyl -D-mannopyranoside (Figure 4). 26,27 Acetylation with Ac2O and pyridine afforded diacetate 13, which was subjected to acetolysis (2% H2SO4, Ac2O) to afford the triacetate 14. Compound 14 was treated with sodium azide in DMF or DMSO; however, under a range of temperatures up to reflux the desired intramolecular substitution to afford 15 did not occur. Instead, only deacetylated talose mesylates were obtained after extended reaction times. In order to prevent cleavage of the acetates at the 2-and 3-positions, more robust benzoate protecting groups were installed by treatment of 12 with BzCl and pyridine, to afford the dibenzoate 16. Acetolysis of 16 (4% H2SO4, Ac2O) afforded the anomeric acetate 17 in 85% yield. Treatment of 17 with sodium azide in DMF or DMSO at a range of temperatures for extended periods gave no evidence of the 1,4-anhydrosugar 18, and resulted in only decomposition of the starting material. For example, trace amounts of tetrabenzoate 19 were isolated, formed by intermolecular transesterification. Achieving no success in our attempts to generate the desired 1,4-anhydro bridge by Method B, which involves reaction of O1 with C4, we sought to reverse the approach and use O4 as nucleophile to react with C1 under Lewis acidic conditions as outlined in Method C in Figure 1. Accordingly, the thiotolyl -D-mannoside 20 28 was treated with 3.3 equivalents of BzCl at 0 °C in pyridine and allowed to warm to room temperature to afford the tribenzoate 21 in 85% yield (Figure 5). Disappointingly, activation of thiomannoside 21 with Niodosuccinimide and triflic acid at a variety of temperature in a range of different solvents failed to form the 1,4-anhydrosugar 22. We speculate that the deactivating benzoate protecting groups render the 4-hydroxyl insufficiently nucleophilic to effect the cyclization. With these disappointing results, we reconsidered our approach and decided to pursue Method A in Figure 1, in which the nucleophilicity of the O4 can be assured. When investigating the conversion of 2,3,6-tri-O-benzyl--D-glucopyranosyl chloride 23 to the 1,4anhydro sugar 24 using sodium hydride, Sato and co-workers identified a remarkable solvent dependence (Figure 6A). 29 Upon performing the reaction in DMSO mainly 2-benzyloxy-3,6di-O-benzyl-D-glucal 25 was obtained with the desired anhydro sugar 24 isolated in only 8% yield. However, using THF, an impressive 93% yield of 24 resulted. Accordingly, we set about the preparation of the equivalent D-manno configured precursor. Acetolysis of methyl To more carefully investigate this reaction, we developed an alternative synthesis of the chloride 29 (Figure 7). Thus acidic hydrolysis of 26 afforded the pure hemiacetal 30. ## CONCLUSION In summary, we report the first synthesis of the previously unreported 'all endo' 1,4anhydro sugar, 1,4-anhydro--D-mannopyranose. This work is significant as it highlights several unsuccessful approaches to this compound and reveals that conditions suitable for cyclization of 2,3,6-tri-O-benzyl--D-glucopyranosyl chloride 23 to the corresponding 1,4anhydroglucose derivative 24 in high yield, 29 give predominantly the elimination product 25 when applied to the D-manno-configured epimer 29. Nonetheless, this approach does provide small amounts of the desired 1,4-anhydro mannose derivative, which could be deprotected to yield the parent 1,4-anhydro--D-mannopyranose 9. ## General methods 1 H and 13 ## 1,2,3-Tri-O-acetyl-6-O-benzoyl-4-O-methanesulfonyl--D-talopyranoside (14) (i) Methyl 2,3-di-O-acetyl-6-O-benzoyl-4-O-methanesulfonyl--D-talopyranoside (13) A solution of methyl 6-O-benzoyl-4-O-methanesulfonyl--D-talopyranoside 12 26,27 (180 mg, 0.478 mmol) was dissolved in a mixture of acetic anhydride and pyridine (3 mL, 1:2) was stirred for 24 h. The solvent was evaporated under reduced pressure and the residue was dissolved in EtOAc. The organic phase was sequentially washed with 2 M aq. HCl (3), sat. aq. NaHCO3, water, and sat. aq. NaCl, before being dried (MgSO4), filtered, and concentrated under reduced pressure. The crude residue was purified by flash chromatography (acetone/PhMe 1:4) to give the diacetate 13 as a colourless oil (198 mg, 99%), 1 ## Methyl 2,3,6-tri-O-benzoyl-4-O-methanesulfonyl--D-talopyranoside (16) Benzoyl chloride (94 L, 0.810 mmol) was added dropwise to a stirred solution of methyl 6- ## 1-O-Acetyl-2,3,6-tri-O-benzoyl-4-O-methanesulfonyl--D-talopyranose (17) A solution of the methyl talopyranoside 16 (195 mg, 0.336 mmol) in 4% H2SO4/Ac2O (5.0 mL) was stirred at r.t. for 5 h then ice-water was added and stirring continued for 30 min. The mixture was extracted with EtOAc (5  5 mL) and the organic phase sequentially washed with sat. aq. NaHCO3 (5 ), H2O, and sat. aq. NaCl. The organic phase was dried over MgSO4, filtered, and concentrated under reduced pressure to give a residue that was purified by flash chromatography (30 % EtOAc/pet. sp.) affording 17 as a clear oil (173 mg, 85%). [D 19 -71 ° (c 1.00, CHCl3); HCl (2 ), sat. aq. NaHCO3 (2 ), H2O and finally sat. aq. NaCl. The organic phase was dried (MgSO4), filtered, and concentrated under reduced pressure to give a yellow oil which was purified flash chromatography to afford mostly pure tribenzoate 21 (95% by 1 ## 4-O-Acetyl-2,3,6-tri-O-benzyl--D-mannopyranosyl chloride (28) A solution of the diacetate 27 (10.0 mg, 18.7 mmol) in ether saturated with HCl (5 mL) was stirred for 30 min at 0 °C under N2. The mixture was warmed to r.t. and stirred for a further 30 min. The solvent was evaporated under a stream of N2 and Et2O (10 mL) was added. This process was repeated five times, and the dried residue was subjected to flash chromatography (15% EtOAc/pet. sp.) giving the chloride 28 (8.9 mg, 93%). Owing to the instability of this compound it was not further characterized. ## 2,3,6-Tri-O-benzyl--D-mannopyranose (30) (i) From mannoside (26) A solution of the mannoside 26 (2.00 g, 4.31 mmol), AcOH (20 mL) and 1 M aq. HCl (5 mL) was heated in an oil bath at 105 °C for 2.5 h then poured into ice-water and extracted with EtOAc. The organic phase was washed with sat. aq. NaHCO3 then sat. aq. NaCl, dried

chemsum

{"title": "The synthesis of 1,4-anhydro-\uf061-D-mannopyranose", "journal": "ChemRxiv"}

synthesis_of_new_fluorescent_molecules_having_an_aggregation-induced_emission_property_derived_from_

## Abstract: Fluorescent molecules based on a fluorinated isoxazole scaffold were synthesized and investigated for their photochemical properties. The introduction of a fluorine substituent into 3,5-diarylisoxazoles led to an increase of fluorescence intensity and exhibited a redshift in the emission intensity. α-Fluorinated boron ketoiminates (F-BKIs) were also synthesized via a ring-opening reaction of 4-fluoroisoxazoles and exhibited highly fluorescent luminescence and aggregation-induced emission (AIE), showing promise as a new fluorophore. ## Introduction Fluorescence bioprobes based on conventional organic dyes are used for enzyme activity measurements and in bioimaging systems with promising applications in the field of clinical diagnostics . Most of the fluorescence bioprobes are mainly excited with near-ultraviolet or blue light ray and the structures often include fluorescein, rhodamine, or 7-amino-4-methylcoumarin (7-AMC) scaffolds as fluorophores. These fluorophores usually exhibit strong fluorescence in dilute solutions, but most of their emissions are partially or completely quenched in the solid state or in highly concentrated solutions by aggrega-tion-caused quenching (ACQ) . On the other hand, there are molecules that exhibit strong emission even in poor solvents or in the solid state. This property is referred to as aggregation-induced emission (AIE) and has attracted much attention in the field of fluorescence bioprobes . For example, it is presumed that prion disease, which is caused by the accumulation of prion protein aggregates in the brain, plays an important role in the pathophysiological mechanism of prion protein-polymerized oligomers. However, since prion protein oligomers cannot be visualized using fluorescent probes, the use of AIE fluorescent probes is being investigated as a tool for analyzing the causal relationship between prion diseases and prion proteins. The importance of fluorinated heterocyclic derivatives in the pharmaceutical and agrochemical industries continues to grow, with several fluorinated 6-membered heteroaromatic derivatives finding applications in a wide variety of drugs and plantprotective agents . However, there are only a few reports on the synthesis and properties of fluorinated 5-membered heteroaromatic systems, especially those comprising two heteroatoms such as pyrazoles , isoxazoles , and thiazoles . Recently, we reported the selective fluorination of isoxazoles, to give monofluorinated isoxazoles 3 or trifluorinated isoxazolines 4 in moderate to good yields (Scheme 1) . In addition, we reported that the reaction proceeded smoothly by starting with 1,3-diketones (1) to give 3 in excellent yields in a one-pot reaction. As part of a wider research program aimed at the applications of fluorinated 5-membered heteroaromatic systems, in this paper, we report the fluorescent luminescence characteristics of 4-fluoroisoxazoles, the synthesis of α-fluorinated boron ketoiminates (F-BKIs), and their photochemical properties. ## Synthesis and optical properties of 4-fluorinated isoxazoles Although there is a large number of fluorescent molecules, fluorescent probes having an isoxazole scaffold are rare and the limited examples that are available also contain other fluorophores such as styryl, anthranyl, or pyrenyl groups in the molecules. We recently reported the synthesis of 3,5-diaryl-4-fluoroisoxazoles 3 that were found to have planar structures suggesting that they might have the potential to act as a fluorophore . During the synthesis of 3,5-diaryl 4-fluoroisoxazoles 3 according to the previous method (Scheme 2), we noted that 3,5-bis(4-methoxyphenyl)-4-fluoroisoxazole (3b) and 3,5bis(4-trifluoromethylphenyl)-4-fluoroisoxazole (3c) exhibited fluorescent properties by irradiation with a UV lamp. Among the non-fluorinated isoxazoles, only 2c demonstrated fluorescent emission, although it was very weak. Thus, we decided to further investigate the photochemical properties and the results were summarized in Figure 1 and Table 1. Introducing a fluorine substituent into the isoxazole scaffold led to an increasing fluorescent intensity and exhibited a redshift in the emission intensity. Interestingly, the excitation maximum of 3 showed a redshift of approximately 20 nm with the incorporation of a single fluorine atom into the isoxazole scaffold in com- parison with 2c. This observation suggested that the strong electronegativity of fluorine might affect the electron density on the isoxazole ring. ## Synthesis of boron ketoiminates and α-fluorinated boron ketoiminates Boron ketoiminates (BKIs, 6) are one of the new types of boron-chelating dye , their optical properties feature a large Stokes shift and high molar absorption coefficients (ε) that are similar to the corresponding boron diketonates. The synthesis and properties of BKIs have been reported recently and they are easily accessible either from the corresponding 1,3-di-ketones 1 or from isoxazoles 2 through a ring-opening reaction (Scheme 3). Based on the above observations, we attempted to introduce a fluorine atom into BKIs to access the corresponding α-fluorinated boron ketoiminates (F-BKIs, 9). First, we started from 1,3-diketones 1 and reacted them with ammonium formate to give the corresponding enaminoketones 5 in high yields (see entries 1-3 in Scheme 4). Then, compounds 5 were treated with 10 equiv of BF 3 •Et 2 O in anhydrous THF solution in the presence of an excess of Et 3 N to give BKIs 6 in moderate yields. However, when the same conditions were applied to the fluori-Scheme 4: Synthesis of enaminoketones 5 and 8 and their conversion to BKIs (yields refer to isolated yields; a boron complexation of 8a to 9a was not attempted). Scheme 5: Attempted selective fluorination of BKI 6b. nated diketone, 2-fluoro-1,3-diphenylpropane-1,3-dione (7a), the corresponding enaminoketone 8a was obtained in only low yield (Scheme 4, entry 4) and we did not attempt the conversion of 8a towards the α-fluorinated boron ketoiminate 9a. Next, we attempted the selective fluorination of 6b to obtain the desired fluorinated analog 9b. However, in the synthesis of F-BKIs through the selective fluorination of the corresponding BKIs, the use of 1 equiv of Selectfluor did not give any product and performing the reaction with excess amounts of Selectfluor gave rise to the corresponding α,α-difluorinated diketone (Scheme 5). As an alternative method to synthesize F-BKIs 9, we turned our attention to the ring-opening reaction of isoxazoles. The reductive cleavage of the N-O bond in isoxazoles can be achieved by transition metals or their complexes to give the corresponding Scheme 6: Ring-opening reaction of 4-fluoroisoxazoles 3 and their conversion into F-BKIs 9 (yields refer to isolated yields). enaminoketones . Consequently, we examined several conditions for the ring opening of fluorinated isoxazoles 3, and found that using Mo(CO) 6 gave the corresponding α-fluorinated enaminoketones 8 in moderate yields (Scheme 6). With the enaminoketones 8 at hand, the subsequent boron complexation with BF 3 •Et 2 O in the presence of Et 3 N gave the desired F-BKIs 9 in moderate to good yields. ## Optical properties of boron ketoiminates and α-fluorinated boron ketoiminates Chujo and co-workers described that BKIs could be a promising structural motif for having AIE properties . For the purpose of comparison with the photochemical properties of BKIs and F-BKIs, we measured the optical properties of compounds 6 and 9 (Table 2). As shown in Figure 2, the UV-vis absorptions of 6b and 9b in THF decreased upon the addition of H 2 O, and white precipitates formed in samples exceeding 80% of water content. Concurrently, the fluorescent luminescence (FL) of the solutions of 6b and 9b exhibited an increase in the emission intensities with increasing water content. It was interesting to note that the excitation maximum of 9b in the aggregated state showed a red-shift by approximately 20 nm based on the incorporation of a single fluorine atom into the boron ketoiminate scaffold in comparison with 6b. On the other hand, unfortunately, no similar behavior could be observed for the other F-BKIs. This effect of 9b bearing OCH 3 groups on both benzene rings might be attributed to the energy gap between HOMO and LUMO based on the electron-density distribution of boron ketoiminate scaffold induced by the strong electronegative fluorine atom . The FL intensities were lower than that of the corresponding BKIs, a similar tendency to what was also observed in other F-BKIs 9a and 9c. In summary, we found that the F-BKIs described in this report exhibited AIE behavior. ## Conclusion In conclusion, we demonstrated that 3,5-diaryl-4-fluoroisoxazoles exhibited fluorescent luminescence, although, the emissions were not strong. Interestingly the introduction of a fluorine substituent into the isoxazole scaffold led to an increase in the fluorescent intensity in the aggregated state and exhibited a redshift in the emission intensity. We also achieved the first synthesis of α-fluorinated boron ketoiminates (F-BKIs) by the reductive cleavage of the N-O bond in 4-fluorinated isoxazoles and demonstrated that F-BKIs exhibited AIE property similarly to their parent BKI. Further structural modifications of compounds 3 or 9 and applications to fluorescent bioprobes are currently under investigation.

chemsum

{"title": "Synthesis of new fluorescent molecules having an aggregation-induced emission property derived from 4-fluoroisoxazoles", "journal": "Beilstein"}

selective_oxidation_of_pharmaceuticals_and_suppression_of_perchlorate_formation_during_electrolysis_

## Abstract: Many pharmaceutical compounds are excreted unchanged or as active metabolites via urine.They pass through conventional wastewater treatment processes, and present a risk to aquatic ecosystems and humans. Point-source remediation of source-separated urine provides a promising alternative to destroy pharmaceuticals before dilution with wastewater. Electrochemical advanced oxidation processes are one possible option for degrading pharmaceuticals in urine, but they often lead to the formation of oxidation byproducts (OBPs) including chlorate, perchlorate, and halogenated organics at hazardous concentrations due to high background chloride concentrations.Here, we show that the high urea content of fresh human urine suppresses the formation of oxychlorides by inhibiting formation of HOCl/OClduring electrolysis, while still enabling the oxidation of pharmaceuticals by • OH due to the slow rate of urea oxidation by • OH. This results in improved performance when compared to equivalent treatment of hydrolyzed aged urine. This (primarily indirect) electrochemical oxidation scheme is shown to degrade the model pharmaceuticals cyclophosphamide and sulfamethoxazole with surface-area-to-volumenormalized pseudo-first-order observed rate constants greater than 0.08 cm/min in authentic fresh human urine matrixes. It results in two orders-of-magnitude decrease in pharmaceutical concentrations in 2 hours while generating three orders-of-magnitude lower oxychloride byproduct concentrations in synthetic fresh urine as compared to synthetic hydrolyzed aged urine matrixes.Importantly, this proof-of-principle shows that simple and safe electrochemical methods can be used for point-source-remediation of pharmaceuticals in fresh human urine (before storage and hydrolysis), without formation of significant oxychloride byproducts. TOC Figure. ## Introduction Many pharmaceutical compounds are not sufficiently deactivated during typical wastewater treatment and eventually end up being discharged into the environment. These pharmaceuticals may threaten aquatic ecosystems, 1 contribute to the development of antibiotic-resistant bacteria, 2 and can eventually return to human drinking water supplies. 3 Chemical oxidation and advanced oxidation processes (AOPs) are an important means of addressing this growing problem because of their ability to degrade organic contaminants via oxidizing species (i.e., • OH, O3 , etc., in addition to HOCl, Cl • , and Cl2 •in free chlorine or chloride containing solutions) through chemical, photochemical, or electrochemical means. Numerous studies have examined pharmaceutical degradation using AOPs, and there are many proposed strategies ranging from large-scale implementation as a tertiary treatment step at centralized wastewater treatment facilities (WWTFs) to small-scale decentralized implementation to treat point sources. Point-source treatment for pharmaceutical pollutants at homes, businesses, and hospitals is attractive because the compounds may be treated before they are diluted by a factor of 1,000 with other wastewater. However, strategies that employ the addition of chemical agents 14 (Fenton's reagent, peroxides or other oxidants or catalysts) are unsuitable for distributed or in-home use due to safety concerns regarding the handling of oxidants, removal or disposal of wastes from the process, and high-cost. This study focuses on the underlying chemistry relevant to application of electrochemical advanced oxidation processes (EAOPs) for treatment of fresh urine. These EAOPs have the potential to enable point-source approaches to treatment of fresh urine that are safe and easy to operate since they do not require the addition of oxidants or generate new waste streams. The two biggest challenges facing this approach are the relatively low concentrations of pharmaceuticals and the potential formation of oxidation byproducts (OBPs) that can result from oxidation of co-occurring chloride and other halides. Many studies highlight that OBPs formed during AOPs can create serious environmental and human health problems on their own, 15 including chlorate, perchlorate, haloacetic acids, aliphatic halide species, haloacetonitriles, haloacetamides, and nitrosamines. The halogenated OBPs are mostly chlorinated species; however, brominated and iodinated species may be present at much lower concentrations but with higher toxicity. 16 Thus, there is a need to develop novel EAOPs or improved approaches to employing existing EAOPs that enable high efficiency toward pharmaceutical destruction while preventing or mitigating OBP formation. The dominant route for the elimination of non-volatile pharmaceutical excretion is via urine. Of the top 200 prescribed drugs in the U.S., roughly 30% of the ingested pharmaceutical load is excreted unchanged via urine, while ~65% is metabolized and excreted via both urine and feces with the remaining ~5% via biliary elimination. 20 Further, some metabolites retain key pharmacological properties of parent compounds, and many of them will remain active. 21 Thus, targeting treatment of urine (before dilution) is an attractive strategy to reduce pharmaceutical pollution. While the high electrical conductivity of urine (due primarily to chloride salts) is an advantage for EAOPs, the presence of high concentrations of urea or ammonium (over two orders of magnitude greater than pharmaceutical concentrations) could be a major hurdle forcing one to oxidize most of the urea before significant oxidation of the pharmaceuticals. In this study, we show that the unusually slow rate of urea oxidation by the hydroxyl radical ( • OH) (7.9´10 5 M -1 s -1 for urea 22 vs. 2´10 9 M -1 s -1 for cyclophosphamide 23 and 6.2´10 9 M -1 s -1 for sulfamethoxazole 24 ) provides a de facto selectivity for • OH attack on the other organics (Scheme 1, r1 and r3). Scheme 1. Relative rate constants for important reactions during the electrolysis of pharmaceuticals in urine matrixes (with cyclophosphamide as an example pharmaceutical in an undivided cell setup). The formation of chlorate (ClO3 -) and perchlorate (ClO4 -) can be supressed by the presence of high nitrogen concentration (i.e., urea or NH4 + /NH3). Divided cell architectures will block r21 and subsequent reactions. The relative rates of these reactions are important to consider to successfully decrease pharmaceutical concentrations while not generating OBPs. The color coding indicates the relative range of large to small rate constants. It is noteworthy that reactions r11-14 involving urea and NH4 + /NH3 as reactants with high concentration will have high reaction rates, even though the second-order rate constants for these reactions are lower relative to other depicted reactions. The black arrows represent the transport of species. Furthermore, • OH, Cl • and Cl2 •depicted at the anode surface may also desorb and diffuse outward through a relatively stagnant near-anode region. Likewise, Cl -, urea, and the pharmaceuticals diffuse from bulk solution toward the anode as they are consumed at the anode or in the near-anode stagnant region. See Section 3.3 for discussion of the importance of mass transport in galvanostatic experiments. Several studies have examined the application of EAOPs to various urine matrixes and have shown differences in the generation of OBPs. Studies have examined electrochemical treatment of synthetic urine and human-generated stored urine, 16, for multiple purposes including pharmaceutical degradation, distributed wastewater treatment, and nutrient recovery (for nitrogen and phosphorus). Interestingly, the concentrations of the OBPs reported for roughly equivalent oxidation treatments are vastly different. For example, many studies report large amounts (>10 mM) of generated ClO4after 30 A•hr/L of treatment using a boron-doped diamond (BDD) electrode, 16,29,30,35 while a similar report measured ClO4below the detection limit for the same normalized charge passed. 31 One critical difference between these studies that may explain these differences is the composition of the matrix. Fresh urine (authentic and synthetic) has a high concentration of urea -averaging 15 g/L (250 mM), 36 and a pH of about 6. By contrast, stored hydrolyzed authentic urine has a urea concentration of ~0 mM, ammonium concentration of 30-120 mM, bicarbonate/carbonate concentration of 25-30 mM, and a pH of about 9. 32 This is due to naturally abundant bacterial urease which hydrolyzes urea to form ammonium, bicarbonate, and hydroxide (eq 1). This reaction happens rapidly, with one study finding that urea is nearly completely hydrolyzed within 5 hours of storage in a pipe. 37 The concentration of ammonium decreases over time with elevation of pH and volatilization of ammonia if the solution is open to the atmosphere. The concentrations of carbonate species also decrease due to precipitation as CaCO3. 38,39 The hydrolysis of urea to NH4 + and HCO3 -, along with the volatilization of NH3 and precipitation of CO3 2-, can have major impacts on the oxidation pathways during the application of an EAOP. One of the strongest oxidants produced in an EAOP is 2.32 V at pH 7), 40 which reacts rapidly with most organic molecules by either addition or hydrogen abstraction (Scheme 1, r3), with second-order rate constants typically in the range of 10 9 ~ 10 10 M -1 s -1 . 22 However, this important radical is scavenged by NH3 and HCO3 -/CO3 2in hydrolyzed urine much faster than by the urea in fresh urine (eqs. 2-5), with respective pseudo-first-order scavenging rate constants, k' = 2.6 × 10 7 s -1 and 2.0 × 10 5 s -1 (Text S1): 22,41 urea While the amino radical (E 0 ( • NH2/NH3) = 0.6 V) 42 and carbonate radical ( 40 , resulting from reactions of • OH with NH3 and HCO3 -/CO3 2-, may also provide some oxidizing power, their second-order rate constants with most organic molecules are more than 3 orders of magnitude smaller than • OH. Further, • NH2 will be rapidly scavenged by dissolved O2 (Scheme 1, r19). 41 In urine matrixes with high chloride concentrations (~100 mM), reaction of • OH with Cl -(Scheme 1, r6) and direct anodic electron transfer from Cl -(Scheme 1, r5) will also lead to formation of the chlorine radicals, Cl • (E 0 (Cl • /Cl -) = 2.43 V) 40 and Cl2 •-(E 0 (Cl2 •-/Cl -) = 2.13 V) 40 . They may also contribute to degradation of organic molecules (Scheme 1, r4), with rate constants for Cl • typically spanning a similar range as for • OH, and rate constants for Cl2 •typically 1-2 orders of magnitude lower than for • OH. 22,42,46 The rate constants do not appear to have been reported for reactions of Cl • or Cl2 •with urea or NH4 + /NH3. However, observations from previous work suggest that these radicals likely react rapidly with both nitrogen species. Therefore, both urea and NH4 + /NH3 likely serve as scavengers of Cl • and/or Cl2 •-. (Scheme 1, r11 and r12). 27, In addition to their effects on organic contaminant degradation, the concentration of urea or NH4 + /NH3 may affect the sequence of chloride oxidation to free chlorine (FC -primarily HOCl and OCl -, and to a lesser extent Cl2 and/or Cl2O) and higher oxidation states (eg. ClO3and ClO4 - ). In fresh and hydrolyzed urine, FC is formed via radical-driven or direct anodic oxidation of Cl -(Scheme 1, r5-10). 51 As noted above, urea or NH4 + /NH3 appear likely to serve as effective scavengers of Cl • and/or Cl2 •in either fresh or hydrolyzed urine. This could contribute to suppression of FC formation by hindering recombination and/or disproportionation reactions involving Cl • and Cl2 •-(several of which lead to direct formation of Cl2) 52 . In hydrolyzed urine with a pH of 9, NH4 + /NH3 (kHOCl,NH3 = 3.1 ´ 10 6 M -1 s -1 M -1 s -1 ) 53 should react with FC with an apparent second-order rate constant of ~3 ´ 10 4 M -1 s -1 (not corrected for ionic strength) (Scheme 1, r13). It results in an anticipated maximal pseudo-first-order FC scavenging rate constant of ~2 ´ 10 4 s -1 (and t1/2 ~ 4 ´ 10 -5 s) at the 500 mM NH4 + /NH3 level of freshly hydrolyzed urine (with diminishing rate constant values as NH3 volatilizes from the urine during storage). In fresh urine with a pH of 6, urea (kHOCl,urea = 0.63 M -1 s -1 ) 54 should react with FC with an apparent second-order rate constant of ~0.6 M -1 s -1 (Scheme 1, r14). This leads to a lower, but still high pseudo-first-order FC scavenging rate constant of ~0.2 s -1 (and t1/2 ~ 4 s). In either case, the presence of high levels of NH4 + /NH3 or urea should contribute to suppression of the further oxidation of HOCl/OClinto ClO3and ClO4 -(Scheme 1, r16-18). Many reports highlight large amounts of OBPs generated from EAOPs applied to urine. While electrochemical remediation of various OBPs has been demonstrated, including alkyl halides, 55,56 haloacetic acids, 57 nitrosamines, 58 ClO3 -, 59 and ClO4 -, preventing them from forming is a preferred approach. ClO4is the most challenging to remediate due to its high stability among all the OBPs generated during electrolysis of urine. ClO4has a large electrochemical activation barrier for reduction to ClO3of 120 kJ/mole, 63 which makes its reduction very sluggish (Scheme 1, r21). The slow kinetics of ClO4reduction accentuate the need that EAOPs not generate significant quantities of ClO4 -, even if an OBP remediation step is applied after oxidation. In this work, we show the advantages of conducting electrochemical oxidation of pharmaceuticals in fresh urine (at point of generation) as opposed to hydrolyzed (without loss of NH3) or hydrolyzed aged urine (with loss of NH3). We demonstrate that urea (and/or NH4 + /NH3) effectively inhibits the oxidation pathway toward ClO4 -, likely due to scavenging of Cl • and/or Cl2 •-(Scheme 1, r11 and r12) or rapid sequestration of HOCl (Scheme 1, r13 and r14). Meanwhile, the urea in fresh urine inhibits the oxidation of pharmaceuticals much less than the NH4 + /NH3 and HCO3 -/CO3 2in hydrolyzed urine due to urea's lower reactivity toward • OH (Scheme 1, r1, r3 and r20). Accordingly, the electrochemical oxidation rates of the pharmaceuticals investigated are higher in synthetic fresh urine matrixes than in synthetic hydrolyzed and hydrolyzed aged urine matrixes. Thus, this work provides a proof-of-concept for simple and safe point-source oxidation of pharmaceutical compounds in fresh human urine. ## Chemicals and Solutions Ultrapure water (>18.2 MΩ/cm resistance) was used for preparation of all standards and test solutions. Reagent grade (99% purity or higher) (NH2)2CO, NaCl, NaHCO3, NH4OH, NH4Cl, HCl, NaClO3, NaClO4, HgCl2, Na 2 S 2 O 3 , Na 2 SO 4 , NaNO 3 , NaNO 2 , glycine, creatinine, uric acid, and urea obtained from Sigma-Aldrich were used for all standards and test solutions. DPD free and total chlorine test kits were obtained from Sigma-Aldrich. NaOCl (Sigma-Aldrich; 13.5% available chlorine) stock solution was calibrated every month by iodometry. Catalase (Sigma-Aldrich) was vortexed and centrifuged to remove the supernatant, then reconstituted in phosphate buffer and centrifuged again, with the process repeated 5 times to remove thymol. Analytical reference standards of cyclophosphamide (CP) (90% purity) and sulfamethoxazole (SMX) (≥ 98% purity) from Sigma-Aldrich were used. Simplified synthetic fresh urine matrixes consisted of an aqueous solution of 250 mM urea, 100 mM NaCl, and either 1.92 mM CP or 0.39 mM SMX. Synthetic fresh urine matrix solutions consisted of 250 mM urea, 100 mM NaCl, 16 mM Na2SO4, 24 mM NaH2PO4, 13 mM creatinine, 3 mM uric acid, and 1.92 mM CP or 0.39 mM SMX. All prepared synthetic fresh urine matrixes were at pH 6. Synthetic hydrolyzed urine matrixes consisted of an aqueous solution of 250 mM NaHCO3, 100 mM NH4Cl, and 400 mM NH4OH (i.e., [NH4 + ]total = 500 mM) at pH 9.35. Synthetic hydrolyzed aged urine matrixes were adapted from Udert et al. 35 and Hoffmann et al. 16 , and consisted of 250 mM NaHCO3, 100 mM NaCl, and either 140 mM or 33 mM NH4OH at pH 9.02 and 8.88, respectively. Authentic fresh urine was collected at 10 A.M. and blended from 6 people, 3 males and 3 females, with different ethnic background and diets. The blended urine was pH 6.3. Electrolysis of authentic fresh urine was started immediately after the addition of 0.39 mM SMX to 40 mL of the blended urine. ## Electrolytic Cell and Electrodes A custom three-electrode electrochemical cell was used for all oxidation experiments in this work (Figure S1). Detailed schematic dimensions and a photograph of the cell are provided in the SI (Figure S2). The cell was designed to accommodate a large planar working electrode with dimensions 40 mm x 80 mm, which is held in place between the bottom of the cell and a base plate. A seal is formed with a Kalrez gasket exposing a planar surface area of the working electrode of 8.56 cm 2 . Planar anodes (BDD or IrO2) were used for the oxidation experiments. SAE 304 stainless steel tubes were used as the counter electrodes in all oxidation experiments. An Ag/AgCl (3 M NaCl) reference electrode (BASi, West Lafayette, IN, USA) was used in all experiments. In the "undivided cell" setup, the counter electrode was immersed in the same solution as the working electrode. In the "divided cell" setup, the counter electrode was placed inside a glass tube with a 12 mm O.D. porous frit at the bottom (4-8 µm pores, ACE Glass #7209 porosity E). For oxidation experiments, three counter electrodes in fritted glass tubes were used in parallel to minimize resistance and potential drop due to ionic transport through the frits. Planar boron-doped diamond (BDD) electrodes were purchased from Condias GmbH. Planar Ti/IrO2 electrodes were prepared by thermal decomposition of a 250 mM H2IrCl6 precursor solution similar to a method described previously. 35 Ti sheet metal was first sandblasted and cleaned with 1 M oxalic acid for 1 hour at 95 °C. The precursor solution was spray-coated onto the Ti substrates on a hotplate held at 500 °C, and then the substrates were annealed at 500 °C for 1 hr after precursor deposition. ## Electrochemical Methods All cyclic voltammetry, galvanostatic, and potentiostatic experiments were performed with a Princeton Applied Research Potentiostat/Galvanostat Model 263A. Cyclic Voltammetry of Anode Reactions: Each electrode (BDD and IrO2) was sonicated in deionized water then pretreated in 40 mL 100 mM Na2SO4 for 5 minutes at 3.5 V or 2 V respectively. The electrochemical cell was then thoroughly rinsed with deionized water. The matrix of interest was then added to the cell and a single scan was initiated from 0 V to 3.5 V or 2.5 V respectively, and back to 0 V at a scan rate of 50 mV/s in the undivided cell setup. Nitrogen gas was bubbled through the solution for two minutes prior to and blanketed during the cyclic voltammetry. Galvanostatic Oxidation of Pharmaceuticals in Urine: Electrochemical oxidations of pharmaceutical compounds were performed galvanostatically at a current density of 10 mA/cm 2 for 120 min (reaching 4.2 A-h/L) while the solution was stirred at 550 rpm. These experiments were conducted in a supporting electrolyte with the major constituents of synthetic fresh urine (100 mM NaCl and 250 mM urea, pH 6.15) or full synthetic fresh urine matrixes (see above) or synthetic hydrolyzed, aged urine matrixes (i.e., [NH4 + /NH3] = 33 or 140 or 500 mM, [HCO3 -/CO3 2-] = 250 mM and pH 9) and SMX or CP at a concentration of 0.39 mM or 1.92 mM, respectively. Based upon the suggested dosage from U.S. FDA 64,65 and percentage of excretion via urine, 66, 67 SMX and CP will remain in human urine at concentrations of 0.975 mM and 1.92 mM, respectively. While the solubility of SMX is 610 mg/L (2.41 mM) at 37 °C, 68 dissolving this concentration of SMX at room temperature was difficult. Therefore, 100 mg/L SMX (0.39 mM) was used for electrolysis at room temperature. Aliquots were periodically collected, quenched with excess thiosulfate, and analyzed for target compounds and oxidation byproducts. Galvanostatic Oxidation of Urine without Pharmaceuticals: Electrochemical oxidations of various urine matrixes were performed galvanostatically at a current density of 93 mA/cm 2 on BDD electrode for 90 min (reaching 30 A-h/L), stirred at 550 rpm. These experiments were conducted in the supporting electrolytes that simulated different urine matrixes (simplified vs full urine matrixes, fresh vs. hydrolyzed, ammonia loss vs. no ammonia loss). A control experiment was conducted in NaCl solution to yield maximal oxychloride generation as a reference. ## Analytical Methods Aliquots were taken during electrolysis for analyses by ion chromatography (IC) and highpressure liquid chromatography with ultraviolet absorbance and mass spectrometry detection (HPLC-UV-MS) (Text S2). pH was measured by an Apera pH60 pH tester. Free chlorine (FC), total chlorine (TC) and ClO2 were measured by DPD and/or ABTS methods when electrolysis was employed in various urine matrixes (Text S3). ## Advantages of Decentralized Treatment of Urine Point-source electrochemical oxidation of pharmaceuticals in source-separated urine has multiple advantages compared with centralized treatment at a WWTF after it mixes with feces and other wastewater (Table 1). These advantages are: (1) small treatment volumes, (2) high pharmaceutical concentration, (3) absence of non-urine derived background organic carbon or other interfering matrix constituents, and (4) high electrical conductivity. Average human urine production is 1.3 L/(person•day), while average domestic wastewater discharge is two orders of magnitude higher at 148 L/(person•day). 69 The dilution of urine with other waste streams not only decreases the absolute concentration of pharmaceuticals, but also decreases their relative concentration compared to the total concentration of organics in the solution (due to mixing with feces, cooking oils, detergents, etc.). This domestic wastewater may be further diluted by other waste streams containing other organics (such as industrial wastewater, urban runoff, etc.) before reaching a WWTF. The daily per capita load of chemical oxygen demand (COD) to domestic wastewater from all sources (e.g., including feces, urine, greywater, etc.) is more than 6x higher than from urine alone, and the CODpharm/CODtotal is correspondingly 3.8% for domestic wastewater compared to 24.2% for urine (Table 1). Furthermore, the typical conductivity of fresh urine is two orders-of-magnitude higher than domestic wastewater. The combination of large treatment volumes and low conductivity for domestic wastewater streams would correspond to massive ohmic losses in the solution, making electrochemical oxidation as a "polishing" or tertiary WWT step even more costly. a Dilution with other wastewater increases the treatment volume and reduces the matrix conductivity by two orders of magnitude. Pharmaceuticals are at their highest absolute and relative concentration when they are in urine, before they are diluted by other domestic wastewater sources. b To estimate the electricity cost of electrochemically oxidizing all organics in the matrix, chemical oxygen demand (COD) was used. The assumptions for these calculations are an electricity cost of 0.15 $/kWhr, a pharmaceutical concentration of 10 mM, a COD of 5 mols O2/mol pharmaceutical, a total applied voltage of 5 V, a Faradaic efficiency toward oxidant reaction with organics of 30%, and full mineralization of pharmaceuticals to CO2, H2O, etc. ## Reactive Chlorine Species Scavenged by Urea Cyclic voltammetry was performed in three different solutions to reveal the effects of urea on the electrocatalytic oxidation of water (primarily to O2(g) and • OH) and chloride (primarily to Cl2(g), HOCl, Cl • , and Cl2 •-) by boron-doped diamond (BDD) and thermally decomposed iridium oxide (IrO2) anodes. BDD was used as a model "non-active" electrode that only physisorbs • OH, whereas IrO2 was used as a model "active" electrode that chemisorbs • OH, effectively forming a surface hydroxyl group. The three solutions chosen comprised: (1) 100 mM NaClO4 to characterize the water oxidation without chloride oxidation, given that ClO4will not be further oxidized; (2) 100 mM NaCl to observe the additional current from chloride oxidation; and (3) 100 mM NaCl plus 250 mM urea to observe any differences in chloride oxidation that occur with urea present. Figure 1a, b shows cyclic voltammograms (CV) of oxidative sweeps of these three solutions on BDD and IrO2. The CV of BDD with NaClO4 shows the expected large onset potential for O2(g) evolution of 2.4 V vs SHE. 71 With Clpresent, the current onset potential shifts down to 2.1 V vs SHE due to the lower overpotential for Cl2(g) evolution. See Table S1 for a list of relevant thermodynamic standard and formal potentials for these reactions. In contrast to BDD, the CV of IrO2 in Figure 1b showed a difference in the current onset potential and the magnitude of current when Clis present. The IrO2 surface oxidizes according to IrO2 + H2O ⇌ IrO3+ 2H + + 2e -. 72 In the Clonly solution on IrO2, the peak seen in the cathodic sweep from 1.2 V to 0 V likely corresponds to the reduction of oxidized chlorine species such as HOCl, Cl • , or Ir-O-Cl surface groups. Most importantly, the absence of this peak when urea is present in the matrix shows that urea scavenges one or more of these reactive chlorine species (RCS). ## Pseudo-First Order Kinetics and Mass Transfer Most research thus far examining electrochemical oxidation of pharmaceuticals has reported rates of pharmaceutical degradation in terms of observed pseudo-first-order rate constants. As highlighted by Zöllig et al., these first-order kinetics should only be expected in galvanostatic electrolysis for mass-transfer-limited reactions (assuming only heterogeneous reactions). 34 Therefore, the pseudo-first-order rate constants for pharmaceutical degradation are highly dependent on the geometry and the mass transport in the electrochemical setup. A mass-transfer limited observed pseudo-first-order rate constant is directly proportional to the planar surface-area of the anode and inversely proportional to the volume of the solution. In this work, an effort was made to choose reasonable values for the ratio of the electrode surface area to electrolyte volume (A/V) based on scale-up to a practical device. The electrochemical setup used in this work (Figure S1) has an A/V of 0.21 cm -1 and is magnetically stirred. The hydrodynamic flow pattern in the magnetically stirred reactor is similar 73 to the flow pattern in a rotating disk electrode (RDE), 74 where the mass-transfer limited reaction rate is proportional to Re 1/2 (Reynolds number, Re = w r 2 /n). In order to assess if the electrochemical oxidation of pharmaceuticals is mass-transfer limited in our stirred electrochemical setup, we conducted galvanostatic oxidation experiments over a wide range of Reynolds number, 150 < Re < 10,000. We found that the observed firstorder rate constant for CP degradation varied linearly with Re 1/2 (Figure S3). This confirms that the experiments here are mass-transfer limited for CP. All experiments were performed with stirbar radius (r) and rotation rates (w) corresponding to Re = 1,600. This yields a mass-transfer limited electrolysis rate of 0.022 min -1 (i.e., 0.102 cm/min) for CP. To determine what pseudo-first-order rate constants are practical for a real device, we modeled expected degradation rates for various rate constants as shown in Figure S4. A pseudo-first order rate constant of 0.01 min -1 or greater is required to lower pharmaceutical concentrations by at least three orders of magnitude in a 12-hour period, which is a reasonable residence time for a practical at-home situation. This corresponds to a geometry-normalized pseudo-first order rate constant (based on A/V) of at least 0.05 cm/min to degrade pharmaceuticals at a rate reasonable for a scaledup device. ## Electrolysis of Pharmaceuticals Galvanostatic oxidation in fresh-urine matrixes: A series of galvanostatic oxidations were performed with BDD and IrO2 anodes. The experiments were performed on both anodes for two pharmaceuticals, SMX and CP (see chemical structures in Figure 1c, d insets). SMX and CP were chosen as test compounds because of their large differences in reactivity toward free chlorine (FC). SMX has a relatively high bimolecular rate constant with FC of ~10 3 M -1 s -1 at circumneutral pH, 75 while we measured CP to be essentially non-reactive with FC (see Text S4 for experimental details) (Scheme 1, r15). Galvanostatic oxidations were performed in both an undivided electrochemical cell (working electrode (WE) and counter electrode (CE) in same compartment) and divided electrochemical cell (WE and CE separated by a frit) to test if cathodic reactions had any influence on the degradation pathway (see Figure S1a and S1b). The concentrations of SMX and CP were lowered by two orders-of-magnitude after two hours of electrolysis at 10 mA/cm 2 for all conditions (Figure 1c and 1d). This corresponds to geometry normalized pseudo-first order rate constants greater than 0.1 cm/min for all conditions as shown in Figure 1e. As shown in Figure 1f, observed pseudo-first order rate constants of SMX degradation were lower for synthetic fresh urine matrixes and authentic fresh urine. However, the reactions remained sufficiently fast to degrade SMX by four orders of magnitude in ~8 hours as illustrated in Figure S4. The rate constants for oxidation of both pharmaceuticals by both electrodes are similar in simplified synthetic fresh urine matrixes (i.e., urea + Cl -). This suggests that mass transfer to the electrode surface is the dominant factor limiting the degradation rates (see above). For Re = 10,000, pseudo-first-order degradation rate constants up to 0.25 cm/min were achieved. However, stir speeds corresponding to a Re number of 1,600 were used for the experimental data shown in Figure 1, since they are easy to achieve in electrochemical devices. For the divided cell setup, a pH gradient is established because the oxygen evolution reaction (OER) at the anode generates H + while the hydrogen evolution reaction (HER) at the cathode generates OH -. On average, the pH was stabilized to 1.8 for IrO2 and 2.3 for BDD in the anode chamber and 12.5 in the cathode chamber for both electrodes. Furthermore, this gradient is quickly established. Passing 10 mA/cm 2 for 5 mins (~0.2 A•hr/L) was sufficient to establish this gradient of ~10 pH units for both BDD and IrO2. Moreover, the electrolysis rates of CP and SMX did not show significant differences in divided and undivided cell setups. This suggests that there is little to no effect of cathodic reactions (possibly reducing homogeneous oxidants) on the degradation of pharmaceuticals, which is consistent with the above finding of mass-transfer limited degradation kinetics at the anode. In addition, the effect of the bulk solution pH (a drop from 6 in the undivided cell to 2 in the divided cell) had no appreciable impact on degradation rates. This could be due to the fact that the local pH at the anode surface in the undivided cell may be much lower than the bulk pH due to H + generation at the anode. ## Differences between Synthetic Fresh, Hydrolyzed, and Hydrolyzed-Aged Urine Matrixes: The electrolysis rates of CP and SMX were also investigated in synthetic hydrolyzed and synthetic hydrolyzed aged urine matrixes and compared to a reference mass-transfer limited rate (Figure 2a). Sketches of concentration profiles adjacent to the anode surface are shown in Figure 2b to illustrate the differences between kinetic limitation (gradient A), mass-transport limitation due to heterogenous reactions (direct electron transfer and reactions with adsorbed • OH or RCS) (gradient B), and mass-transport limitation due to the combination of heterogeneous reactions and nearanode-surface homogeneous reactions with desorbed • OH, RCS, or other homogeneous oxidants (gradients C and D). As noted above, the mass-transfer limited electrolysis rate for CP was determined to be 0.022 min -1 . Given CP's slow reaction with RCS and the short lifetime of • OH, 0.022 min -1 likely corresponds to the mass-transport limited rate with primarily heterogeneous reactions. Desorbed • OH may exist in solution, but the homogeneous reaction zone where it may react with CP is less than 1 µm in thickness, 76 which is small compared to the stagnant layer thickness of >20 µm. Additionally, the linear relationship between kobs and Re 1/2 also suggests the obtained mass-transfer limiting rate corresponds to gradient B. As shown in Figure 2a, the degradation rates of CP were mass-transfer limited in simplified and full synthetic fresh urine matrixes -likely due to the electro-generation of • OH on the BDD anode. In comparison, the electrolysis rates of SMX exceeded the heterogeneous reaction masstransfer limiting rate (gradient B). This suggests SMX degradation also occurs through homogeneous reactions involving either reactive chlorine species (RCS -a collective sum of FC and chlorine radicals) or other homogeneous oxidants (Figure 2b, Gradient C or D). In contrast, the electrolysis rates of CP and SMX were significantly decreased below the mass-transfer limiting rates in synthetic hydrolyzed urine matrixes amended with varying NH4 + /NH3 levels (to reflect different degrees of aging). This suggests that HCO3 -/CO3 2-(present at the same 250 mM concentration in each hydrolyzed urine matrix) acts as a dominant • OH and Cl • /Cl2 •scavenger in such matrixes. This results in the formation of CO3 •-, which typically reacts with organic contaminants ~10 3 -fold slower than • OH or Cl • and ~10-fold slower than Cl2 •-(see above). Furthermore, the degradation rates of both pharmaceuticals in such matrixes were invariant with changes in the concentration of NH4 + /NH3, consistent with the apparent low reactivity of NH4 + /NH3 toward CO3 •-. 77 This likewise suggests that RCS do not contribute significantly to CP or SMX oxidation in the hydrolyzed urine matrixes, because increasing concentrations (33-500 mM) of NH4 + /NH3 would have been expected to increase RCS scavenging efficiency, and consequently decrease rates of pharmaceutical degradation if RCS were predominant oxidants. Furthermore, the electrolysis of CP and SMX appears to be kinetically limited in hydrolyzed urine matrixes (Figure 2b, Gradient A) for BDD anodes. As a result, it may be advantageous to electrochemically oxidize pharmaceuticals in fresh urine rather than in hydrolyzed urine for faster pharmaceutical removal. (assuming constant bulk concentration). Gradient A: oxidation that is kinetically limited by the heterogenous reactions at the anode surface (i.e., via direct anodic oxidation and surface bound oxidants). Gradient B: the oxidation is mass-transfer limited and the concentration is zero at the anode surface, but increases in solution. Gradient C: the oxidation is mass-transfer limited, but in addition to heterogeneous reactions, oxidants generated at the anode desorb, diffuse outward, and react homogeneously with the pharmaceutical to drive the concentration to zero some distance away from the surface. Gradient D: the oxidation is mass-transfer limited with a larger homogeneous reaction enhancement than gradient C. ## Generation of Oxidation Byproducts The rates of oxidation byproduct generation in full synthetic fresh urine matrixes were compared for each combination of anode and cell configuration as shown in Figure 3. Both ClO4and ClO3were often found to be near or below the detection limits (i.e., 2 μM) for the ion chromatography techniques used in these experiments. The detection limits for both ClO3and ClO4were higher than in pure water due to the high concentration of Clin urine matrixes. No ClO4was measured in matrixes oxidized by the IrO2 anode, which is similar to what has been reported previously. 16,31 ClO4was formed in the matrixes oxidized on the BDD anode, but at much lower concentrations than what has been reported previously. 16,35 Additionally, the ClO4measured on the BDD electrode was at a lower concentration than the ClO3 -, which is consistent with other reports. 16,35 Nitrate (NO3 -) (Figure 3c) and nitrite (NO2 -) (Figure 3d) generation were found to be higher on the IrO2 electrode than the BDD electrode, though in IrO2 divided cell experiments, NO2generation was below the detection limit (5 μM). These oxidized nitrogen anions could come from the oxidation of urea, creatinine, uric acid and/or the oxidation of CP or SMX. Reaction pathways involving radical-driven generation of reactive nitrogen species from NO3or NO2have been shown to lead to potentially harmful nitrated and nitrosated byproducts. 78,79 The suppression of NO2formation when using the IrO2 anode in the divided cell configuration indicates that such pathways would not be active or would at least be diminished and represents a potential advantage of operating in this mode. substantially lower than what have been previously reported in authentic hydrolyzed aged urine. 16,35 Inhibition of Oxychloride Formation Increases with Increased Urea or NH4 + /NH3: The levels of generated OBPs shown in Figure 3 are substantially lower than in previous studies of authentic hydrolyzed aged urine electrolysis. 16,35 Therefore, further experiments were undertaken in several matrixes and at extended electrolysis time to evaluate potential causes of the lower rates of formation of OBPs (Figure 4). Fresh urine has an average concentration of 250 mM urea, 36 whereas hydrolyzed urine should exhibit maximal concentrations of 500 mM NH4 + /NH3 and 250 mM HCO3after urea hydrolysis. Previous studies that have examined the treatment of authentic hydrolyzed aged urine have typically utilized much lower NH4 + /NH3 concentrations (e.g., 34 mM 16 and 109 mM 35 ) indicating substantial ammonia loss during storage (aging). As shown in Figure 4, ClO3and ClO4generation were increasingly inhibited as dissolved nitrogen concentration increased. For example, a 10 3 -fold decrease in ClO3and ClO4generation was observed for a matrix of 100 mM Cland 250 mM urea compared to 100 mM Clalone. A full synthetic fresh urine matrix containing the other primary constituents of urine (citrate, creatinine, uric acid, SO4 2-, and H2PO4 -) also showed exceptionally low ClO3and ClO4generation. A matrix representing hydrolyzed (but not aged) urine (100 mM Cl -, 500 mM NH4 + , and 250mM HCO3 -) yielded a roughly 10 2 -fold lower ClO3concentration than 100 mM Clalone, whereas ClO4levels generated were similar to those observed in the presence of 250 mM urea. The dashed lines in ClO3and (b) ClO4were measured for various matrixes treated using a BDD anode in the undivided setup. For the matrixes containing NH4 + /NH3, the pH was 9-10. For the matrixes containing only Cland/or urea, the pH was 6-7. Galvanostatic treatments were performed for 90 mins at a current density of 93 mA/cm 2 for a final charge passed of 30 A•hr/L. The dashed lines indicate concentrations 100x higher than suggested drinking water limits for each oxychloride species (see accompanying discussion in the main text). ## Mechanism of Inhibition of Oxychloride Formation: In a chloride solution, oxidation of Clto HOCl/OClis expected to proceed through direct oxidation or reaction with • OH at or near the electrode surface (Scheme 1, r5-10). 15,81 Subsequently, HOCl/OClis oxidized at or near the BDD electrode surface via direct electron transfer and/or reaction with • OHads to form ClO3and ClO4 -(Scheme 1, r16-18). 51,82 In this study, we have shown that increasing urea or NH4 + /NH3 inhibits the formation of ClO3and ClO4 -. We hypothesize that high urea or NH4 + /NH3 concentrations effectively prevent oxychloride formation by: (1) scavenging chlorine radical species (Scheme 1, r11 and r12) and/or (2) reacting with HOCl/OCland thereby sequestering chlorine in the form of organic chloramines (N-chlorinated urea) or inorganic chloramines (NH2Cl, NHCl2, and/or NCl3) (Scheme 1, r13 and r14). Either pathway would block key steps in the pathways of Cloxidation to ClO3and ClO4 -. Free chlorine (FC; comprising Cl2O, HOCl, OCland Cl2), total chlorine (TC; comprising FC and chloramines), and chlorine dioxide (ClO2) in bulk solution were quantified in full synthetic fresh urine and various synthetic hydrolyzed urine matrixes to explore the mechanism whereby ClO3and ClO4formation is inhibited (Figure 5). The key observations from these measurements are: (1) ClO2 was not detected at measurable concentrations in any of the matrixes investigated, consistent with the hypothesized action of urea or NH4 + /NH3 on RCS involved in steps preceding formation of ClO2 in the Clto ClO4oxidation sequence (e.g., Cl • , Cl2 and importantly (4) lower concentrations of chloramines were generated in the presence of higher concentrations of NH4 + /NH3 or urea. This last observation indicates that FC generation (followed by chloramine formation) is faster at lower concentrations of NH4 + /NH3. This suggests that at higher NH4 + /NH3 concentrations, the reaction between Cl • /Cl2 without formation of chloramines. A satisfactory explanation must also reconcile the fact that the rates of generation of ClO3and ClO4were found to be similar in NaCl-only electrolyte (no urea, no ammonium, no bicarbonate, etc.) and in synthetic hydrolyzed urine with low NH4 + /NH3 concentration (33 mM NH4 + , 250 mM HClO3 -) (Figure 4), even though bulk-solution FC was effectively sequestered as chloramine in the latter matrix (Figure 5b). Taken together, these observations suggest that: (1) generation of ClO3and ClO4in these systems involves oxidation of FC at or near the anode surface (as opposed to the bulk solution), which is consistent with DFT calculations in previous work; 51,82,85 (2) lower levels of FC formation for high urea and NH4 + /NH3 concentrations are linked to lower levels of ClO3and ClO4formation; and (3) suppression of ClO3and ClO4formation by urea and NH4 + /NH3 derives from their Cl • and/or Cl2 •scavenging effects at or near the anode interface that precede formation of HOCl/OCl -(Scheme 1, r11-12), rather than from FC scavenging effects in bulk solution or at the interface (Scheme 1, r16). Considering the much lower reactivity of urea than NH4 + /NH3 toward • OH (Text S1), scavenging of • OH is unlikely to be the primary mechanism for the suppression of oxychloride formation. This also points to scavenging of Cl • or Cl2 •involved in FC formation as a more likely explanation for the similar effects of urea and NH4 + /NH3 on ClO4formation, and the even somewhat greater effectiveness of urea in suppressing ClO3formation. Finally, the fact that similar amounts of ClO3and ClO4are formed in both NaCl-only and low NH4 + /NH3 (Figure 4) solutions suggests that NH4 + /NH3 becomes depleted near the anode surface at low concentrations. This shuts-down the amino radical pathway (Scheme 1, r11-12) leaving Cl • or Cl2 •to form HOCl (Scheme 1 r6-10), which is then oxidized to form ClO3and ClO4 -. The use of appropriately-selective probe compounds for quantification of Cl • , Cl2 •-, and/or ClO • could provide more definitive evidence of the relative effects of urea and NH4 + /NH3 on chlorine radical scavenging. Unfortunately, the probe compounds most commonly accepted for such uses in advanced oxidation processes, such as nitrobenzene, benzoic acid, 89 and 1,4dimethoxybenzene, 90,91 are unsuitable for electrolysis in either divided or undivided cells due to potential artifacts from their direct oxidation on anodes. Overall, these results indicate that the protective effect of ammonia in mitigating oxychloride formation during electrolysis of hydrolyzed aged urine is likely to be lower than that afforded by urea in fresh urine due to ammonia losses following urea hydrolysis. ## Conclusion This work demonstrates that urea (or NH4 + /NH3 from hydrolyzed urea) at the concentrations present in human urine can suppress the pathway(s) of Cloxidation to the oxychlorides ClO3and ClO4during EAOPs. This fortuitous effect greatly inhibits the formation of the highly stable ClO3and ClO4ions, which are two of the most recalcitrant oxidation byproducts of EAOPs. The data show that pharmaceuticals could be degraded to less than 5% of their starting concentrations with less than 10 µM ClO4and 100 µM ClO3generated following 2 hours of electrolysis in matrixes containing 250 mM urea. These data demonstrate the feasibility of devices that eliminate pharmaceuticals in urine at the source of generation while generating minimal oxidation byproducts. "Non-active" BDD 15 has been the anode of choice for EAOPs because of high oxidizing power, but is prohibitively expensive to use in a practical device. This work demonstrates that "active" IrO2 15 is sufficiently oxidizing to degrade cyclophosphamide (a particularly recalcitrant pharmaceutical) in simplified synthetic fresh urine matrixes at reasonably high rates of ~0.1 cm/min. Therefore, it highlights a large opportunity for the development of both "non-active" and "active" low-cost anodes that have long service lifetime. Device development based on this chemistry could provide an important contribution to mitigating the release of pharmaceuticals and other contaminants into the environment. ## Associated Content Supporting Information. Includes calculated • OH-scavenging rates in various matrixes, analytical methods for IC and HPLC-UV-MS, the procedure used to measure CP/HOCl reaction kinetics, procedures for measurement of free and total chlorine, electrochemical cell schematics, results from mass transfer experiments, modeling of degradation rates for different pseudo-firstorder rate constants, and tables of relevant electrochemical and chemical reactions.

chemsum

{"title": "Selective Oxidation of Pharmaceuticals and Suppression of Perchlorate Formation during Electrolysis of Fresh Human Urine", "journal": "ChemRxiv"}

switchable_silver_mirrors_with_long_memory_effects

## Abstract: An electrochemically stable and bistable switchable mirror was achieved for the first time by introducing (1) a thiol-modified indium tin oxide (ITO) electrode for the stabilization of the metallic film and (2) ionic liquids as an anion-blocking layer, to achieve a long memory effect. The growth of the metallic film was denser and faster at the thiol-modified ITO electrode than at a bare ITO electrode. The electrochemical stability of the metallic film on the thiol-modified ITO was enhanced, maintaining the metallic state without rupture. In the voltage-off state, the metal film maintained bistability for a long period (>2 h) when ionic liquids were introduced as electrolytes for the switchable mirror. The electrical double layer in the highly viscous ionic liquid electrolyte seemed to effectively form a barrier to the bromide ions, to protect the metal thin film from them when in the voltage-off state. ## Introduction An ordinary silver mirror, coated on its back surface with silver, reflects light to produce high quality images by reflection, owing to the high reflectivity of silver. Recently, tunable mirrors, including liquid-liquid interfacial mirrors, have started to gather attention due to their high applicability for use as smart windows, light modulators, and chemical sensors. On the other hand, reversible electrochemical mirrors (REMs) are designed to modulate their reflectance from a highly reflective state, enough to mirror a subject, to a highly transparent state, according to external stimuli such as electricity, light, or heat. A number of REMs have been suggested, including metallic thin flms, 7,8 conducting polymers, metal hydrides, and colloidal electrochemical devices. 15 None, however, has reached widespread practical application because of critical problems such as the poor stability of the mirror state and a lack of bistability in reflectance. Thus, there remain very important challenges for developing switchable silver mirrors with long memory effects, to afford bistable reversible electrochemical mirrors (BREMs). Previous studies have employed electrochemical deposition of a metal (e.g. Cu, Ag, Bi, etc.) onto a transparent conducting substrate to achieve a reflective state. The optical properties of these REMs are switched according to the redox states of the metals, as well as their morphologies. Thus, much effort has been directed towards improving the properties of REMs, especially for Ag flm-based electrochemical devices, 8,20 with the addition of Cu ions that stabilize deposited Ag nanoparticles. 17 However, no studies have yet reported on the stabilization of the mirror state over a long period, either in the electricity-on or -off state. In electrochemical metallic mirrors, the reversible reflectance change originates from the electrodeposition of the Ag flm, to achieve a mirror state, and the dissolution of Ag as an ion into an electrolyte, to achieve a transparent state. While the transparent state is quite stable, the mirror state is unstable because the deposited Ag flm is dissolved into the electrolyte solution as anions diffuse into the metallic flm at the opencircuit state. To maintain the mirror state, it is necessary to apply a reduction voltage continuously to avoid the dissolution of Ag flm into the electrolyte. However, Ag nanoparticles with aggregated structures and poor adhesion onto a substrate (e.g. ITO) generally yield cracks and wrinkles on the metal flms under a prolonged supply of electrical charge. Therefore, it is important to stabilize the metallic flm during the mirror formation. 21 The ultimate goal of these studies is to obtain bistability in switchable mirrors-i.e., stability in both the reflective and transparent states-particularly when the electrical power is turned off. This bistable status is critical in order for new switchable mirrors to have technological promise for applications such as optic devices, memory, as well as energy saving smart windows. The poor bistability of REMs originates from the oxidation of metallic silver to soluble AgBr n (1n) then to Ag(I), as described well in the literature. 22,23 High concentrations of halides increase the rate of Ag dissolution. 24 To protect the metal flm from anions, we attempted to introduce an electrical double layer (EDL), which can block ion diffusion. 25,26 Highly capacitive ionic liquids (ILs) have been successfully used to stabilize metal nanoparticles 27,28 or electrochemical devices. Herein, we report an electrochemically stable and bistable reversible electrochemical mirror (BREM) for the frst time, by using reversible silver deposition on a thiol-modifed ITO electrode in ionic liquids as the electrolyte media. ## Materials Dimethyl sulfoxide (DMSO), silver nitrate (AgNO 3 ), copper chloride (CuCl 2 ), (3-mercaptopropyl)trimethoxysilane (MPTMS), 1-methyl imidazole, 1-butyl imidazole, bromoethane, 1-bromohexane, tetrabutylammonium bromide (TBABr), and polyvinyl butyral (PVB, Butvar® B-98) were purchased from Aldrich. All chemicals were used as received. 1-Methyl-4-hexylimidazolium bromide (MHImBr), 1-butyl-4-ethylimidazolium bromide (BEImBr), and 1-butyl-4-hexylimidazolium bromide (BHImBr) were synthesized and purifed according to previous reports. The structures of synthesized ILs are listed in Table 1. ## Preparation of surface-modied ITO electrodes (TI) The surface-modifed ITO electrode was prepared by a previously reported procedure. 35 In brief, an ITO electrode was cleaned by scrubbing with a soft cloth and sonication in ethanol and acetone for at least 10 min each, and then dried under nitrogen. To get a sufficient amount of the surface treatment reagent on the ITO electrode, hydroxyl groups were formed by treating the bare ITO glass surface with oxygen plasma. The surface of the cleaned ITO glass was modifed by oxygen plasma treatment for 5 min with a power of 6.8 W (CUTE-MP, Femto Science, USA). Then, the plasma-treated ITO electrode was placed in a vacuum chamber with a few drops of MPTMS, and then evacuated for 1 h. ## Preparation of electrochemical silver mirrors The electrochemical mirrors consist of an electrolyte between the surface-modifed ITO glass (TI) and the bare ITO (UI) electrode. The TBABr based electrolyte solution (TBAB) was prepared as follows: 0.5 mmol of AgNO 3 , 0.1 mmol of CuCl 2 and 2.5 mmol of TBABr were dissolved in 11 g of DMSO with 1.2 g of PVB as the host polymer. The ionic liquid-based electrolyte solutions (MHIB, BEIB, and BHIB) were prepared by dissolving 0.5 mmol of AgNO 3 and 0.1 mmol of CuCl 2 in 10 wt% of DMSO and 45 mmol of MHImBr, BEImBr and BHImBr, respectively. The prepared electrolyte was carefully coated on to the bare ITO electrode with a polyimide spacer of 500 mm thickness and a window size of 2.0 2.0 cm 2 , and then assembled with a surface-modifed ITO electrode. ## Measurement Electrochemical measurements for the ionic liquids and prepared REMs were recorded using a universal potentiostat [model CHI 624B (CH Instruments, Inc.)]. UV-Vis. spectra were obtained using a PerkinElmer Lambda750 UV/Vis/NIR Spectrophotometer. The fgures of merit for reflectance and thickness were calculated by dividing reflectance recorded at 650 nm (%) and thickness (nm) by consumed charge density (C cm 2 ). SEM images and EDS mapping were obtained using a JEOL-JSM-7001F, and TEM images were obtained using a JEOL-JEM-2100. TEM sample (Fig. S1c †) of the Ag/Cu alloy particle was obtained by collecting alloy particles from the sonication of the electrodeposited Ag and Cu electrode. The resistances and capacitances of the ILs were determined from an impedance analyzer (Ivium B08016, Ivium technology) as a function of frequency (from 10 to 10 5 Hz) using simulation software (Zview 2.8d, Scribner Associates Inc.). A 50 mm thick IL layer was sandwiched between two ITO electrodes and a laminated top silver contact. For obtaining confocal fluorescence images, the thiol-modifed ITO glass was cut into a 5 cm 1 cm piece. Then the two ITO glasses were separated by 500 mm by a polyimide tape spacer and adhered by a hot-melt adhesive. The electrolytes, TBAB and BEIB, containing a fluorophore (pyrene 1%) were injected between the ITO glasses and then covered with cover glasses. A confocal microscope (Axio Imager Z2, LSM 700, Carl Zeiss) was used for obtaining images of the side view. A laser (wavelength 405 nm) was used as the excitation source for pyrene. ## Electrochemical stability of the REM device A highly reflective mirror with a smooth surface was obtained from an electrolyte solution containing AgNO 3 and CuCl 2 (5 : 1 molar ratio), which was optimized according to the previous study. 20 The other components of the electrolyte were tetrabutylammonium bromide (TBABr) and polyvinyl butyral (PVB) in dimethyl sulfoxide (DMSO), as described in the experimental section. This electrolyte composition is abbreviated as TBAB hereafter. The distribution of metallic nanoparticles, formed upon application of 2.5 V on a 2-electrode REM device, was relatively homogeneous and denser in the presence of Cu 2+ than without it. In the SEM images, along with EDS mapping, the homogeneous distribution of the Ag-Cu bimetallic flms was clearly observed (Fig. 1e). This improved morphology of the metallic thin flm is attributed to the fact that Cu induces aggregation of Ag particles. The Cu content on the electrodeposited Ag-Cu surface was determined as $24% of the Ag content from EDS mapping, which was a smaller amount than in the feed, possibly due to the charge difference between the two atoms (Table S1 ## †). A thiol-modifed electrode (TI) was prepared by anchoring the plasma-treated ITO electrode with (3-mercaptopropyl)trimethoxysilane (MPTMS) (Fig. 1a). The metallic flm was grown on TI upon application of a reduction potential (2.5 V), using the solution of TBAB (Fig. S1d-f †). It was noteworthy that the deposited metallic flm was denser on the surface-modifed ITO (TI) than on the bare unmodifed ITO (UI), which can be observed in Fig. S1d and e † and can be inferred from the fgures of merit for the thicknesses of TITBAB and UITBAB (Fig. 1d). The average roughness of the Ag flm on UI was 66 nm, which is twice that on TI (33 nm), as compared in Fig. S2. † This effect can be explained by the strong bonds between thiol groups and the metal ions. When TI is prepared with MPTMS, the trimethoxysilane groups are anchored onto the surface hydroxyl groups of the ITO electrode, while the terminal thiols are left for metal interaction. These surface thiols could enable the formation of strong interactions with the deposited Ag and Cu metals. Thus, the electrodeposited metal flm on TI is more stable, denser and has lower roughness than that on UI. The REM with TI as the working electrode, using the TBAB electrolyte (TITBAB), showed a characteristic cathodic current as the potential moved from zero to the negative direction (Fig. S3 †). When the reduction potential reached 2.5 V, the formation of a reflective mirror was observed in the device, as metal ions were reduced. On the other hand, with the application of a potential in the positive direction, the anodic current appeared and peaked at +0.2 V. This anodic peak should correspond to the oxidation of the electrodeposited metal particles, because the transmittance of the cell was increased by the dissolution of metallic particles into electrolyte solution. The REM returned to its initial transparent state upon application of >1.0 V for less than 1 min, upon oxidation of Cu 1+ to Cu 2+ , which mediates the oxidation of the Ag. 36 The electro-reflectance change of TITBAB was similar to that of the REM using an unmodifed ITO (UITBAB), which is the same mirror system reported in the literature, 20 except for the electrochemical stability, described below. Although it was possible to switch between the reflective and transparent states in UITBAB and TITBAB, the mirror state disappeared when the electricity was disconnected, possibly because the deposited Ag flm dissolved into the solution immediately. To maintain the mirror state, a continuous reduction potential (2.5 V) should be applied to the REM. As shown in Fig. 1b and c, the Ag-Cu metallic flm formed on TI (TITBAB) showed a dramatic enhancement in its long-term electrochemical stability. The reflectance increased up to 82% within 3 min, and it was maintained constantly with the prolonged application of 2.5 V for 30 min. The metallic flm was also stable, showing a highly reflective state without rupture, as shown in Fig. 1c. On the other hand, the metallic flm grown on the bare electrode (UITBAB) showed instability after a few minutes at 2.5 V. The metallic flms were ruptured, and the reflectance dropped signifcantly, as shown in Fig. 1b, and Movie S1. † Also, the charge density recorded with the application of 2.5 V increased abruptly when the metallic flm was ruptured in the UITBAB, while the charge density in the TITBAB increased linearly. Therefore, the ITO surface modifcation with thiol groups, which is well known to stabilize silver, 21 signifcantly improved the electrochemical stability of the REM device, resulting in a stable mirror status. The reflectance of the TITBAB was higher than that of UITBAB, as shown in Fig. 1b and c, due to the formation of a highly dense metallic flm in the TITBAB compared to the UITBAB, as described above. The result indicates that the surface modifcation afforded a viable method for a dramatic enhancement of long-term stability as well as high reflectance. Thus, the surface-treated electrodes were found to be essential for long-term electrochemical stability, and they were used in further experiments. After metal deposition, the mirror state in the UITBAB disappeared immediately after the applied voltage was shut off. The mirror state in the TITBAB was maintained longer than in the UITBAB, possibly due to the denser flm, but it lasted only 60 s after the electricity was disconnected. In switchable mirrors, the open-circuit memory property-i.e., the ability to maintain the mirror state without further energy consumption-is hard to achieve, although this is commonly observed for at least an hour in other electrochemical color switching systems (e.g., electrochromism). 37-40 ## Bistable reversible electrochemical mirror The low open-circuit memory in switchable mirrors is inevitable because the electrodeposited metal particles are designed to be dissolved in electrolytes containing excess halide ions, to achieve reversible switching. The mechanism for the Ag dissolution is based on the initial oxidation of metallic silver to form soluble AgBr n (1n) then to Ag(I), as described well in the literature. 22,23 There is a possibility that chemisorbed Br ions may catalytically dissolve silver by oxygen dissolved in the electrolyte. 41 However, the effect of this oxidative dissolution is negligible due to the low gas solubility in ionic liquids. 42 The soluble anionic complexes (AgX 2 and AgX 3 2 , where X is Cl, Br, or I) have relative stabilities of I > Br > Cl and their formation is accelerated under high concentrations of halides to increase the rate of Ag dissolution. 24 Therefore, the mirror state rapidly disappears in the open circuit, accompanied by dissolution and diffusion of the electrodeposited mirror whenever Br anions are available. To protect the metal flm in the open-circuit state, we introduced an electrical double layer (EDL), which can block anion diffusion into the metallic flm. A cationic layer was reported to expel anions from the electrode. 25 Therefore, we used an ionic liquid (IL) as an electrolyte for the switching mirrors, taking advantage of the large specifc capacitances of ILs. In addition, the nonvolatile, nonflammable, and highly viscous natures of ILs confer advantages on REMs. 30 The structures of the ILs and their electrochemical properties are summarized in Table 1. After metal deposition, the mirror state in the TIBEIB disappeared immediately upon application of a positive potential (>2.0 V), indicating an electrochemically switchable mirror similar to the TITBAB. The metallic mirror formation was accompanied by concomitant spectral growth from the UV to the IR region, as shown in Fig. 2b for the TIBEIB. Surprisingly, the reflectance of the TIBEIB was maintained after electricity was disconnected (V-off). The reflectance of the TIBEIB, following application of 2.5 V for 30 min, had a slight further increase to 89.3% immediately after switching to the V-off state (Fig. 2c, black solid line). The reflectance increase of the TIBEIB immediately in the V-off state may be attributed to the additional reduction of Ag-Cu due to the immediate polarization change. Importantly, even after 2 hours in the V-off state, the high reflectance was maintained (Movie S2 †). Fig. 2c shows the memory effect (recorded at 650 nm) of the switching mirrors in different ILs having different alkyl chain lengths. In all of the ILs, the reflectance of the mirrors was maintained for longer than 30 min in the V-off state. The reflectance of the TIMHIB, which used MHImBr as an IL, increased to almost 80% immediately after the electricity was disconnected, and the memory effect was longer than 2 h (Fig. 2a) without energy consumption or any physical damage. The mirror switching in an ionic liquid containing BEImBr, abbreviated as TIBEIB, was observed similarly, as expected from its similar CV to the TITBAB (Fig. S3 †). The Cu content on the electrodeposited Ag-Cu surface from TIBEIB was determined as $24% of the Ag content, which was almost the same as that in UITBAB and TITBAB (Table S1 †), as determined from EDS mapping (Fig. S1g †). Although the use of ILs enabled dramatically increased bistability without a loss in the reflectance of the device, the growth of the metallic flm in ILs was slower than with TITBAB, possibly due to the highly viscous natures and high capacitances of ILs (Fig. 2c). But interestingly, the fgure of merit for the thickness of TIBEIB was similar to that of TITBAB. Moreover, the fgure of merit for reflectance was two times larger in the REM with ILs than with TBAB (Fig. 1d). The mechanism for Ag flm deposition can be explained by the reaction of the Ag + ions with Br ions to form AgBr n (1n) (n ¼ 2-4), which is then reduced to give Ag metal on an electrode. It has been reported that immidazolium cations can stabilize AgBr n (1n) and metallic granules. 28,43,44 This stabilizing interaction could generate a dense flm and thus result in a high fgure of merit for reflectance in the mirrors using ILs. Fig. 3a shows a schematic diagram for the working mechanism of a bistable reversible electrochemical mirror (BREM), where Br diffusion onto the metal flm is forbidden due to the electrical double layer (EDL) developed during the switch-on process, allowing the bistability of the switching mirror. The growth of the metallic flm, the switching response, and the reflectance increase at V-off were dependent on the alkyl chain lengths of the ionic liquids. Nonetheless, all of the mirrors in the EDL conditions showed long bistability. The EDL is immediately removed upon application of the reverse potential (oxidation) so that the reversible reaction toward the transparent state can be achieved repeatedly and reliably. Thus, the mirror state of the TIBEIB switched to a transparent state only when an oxidation potential was applied (Fig. S4 and S5 †). The mirror with the IL was reversibly switched between the mirror and transparent states (Movie S3 †) by an alternating potential cycle, as shown in Fig. 2d and e. Table S2 † summarizes the electro-reflectance changes of the mirrors in this study. In order to gain further insight on the bistability, we examined the ion transport in TITBAB and TIBEIB with a confocal microscope (Fig. 3b and c and S6 †). Pyrene was used to elucidate the ion transport, taking advantage of its strong interactions with metal cations, 45 which mean it would move in the same direction as metal ions did. As shown in Fig. S6a, † the fluorescence intensities of the TITBAB and TIBEIB were higher near the working electrode, forming a layer of pyrene crowding when the working electrode was applied with a reduction voltage (3 V). Pyrenes (with metal ions) moved toward the working electrode over time and formed a $150 mm fluorescent layer after 10 min at 3 V, as clearly shown in Fig. 3b and c, and S6a and b. † Interestingly, the fluorescent layer was quite stable in the TIBEIB, while it diffused entirely into the electrolyte in the TITBAB at V-off within 10 min. These results verify that there is no ion transfer from the metallic flm (coated at the working electrode upon reduction) into the electrolyte, and vice versa, in TIBEIB at V-off. On the other hand, the metallic flms are dissolved as ions, which return into electrolyte in the TITBAB at V-off. The localized fluorescent layer in the TIBEIB was fully diffused only when an oxidation potential (+3 V) was applied as seen in Fig. 3c. The growth and diffusion of the fluorescent layer, and thus ion transfer, are more visible in Movies S4 and S5. † ## Conclusions An electrochemically stable and bistable electrochemical mirror was achieved for the frst time by introducing (1) a thiol-modi-fed ITO electrode for the stabilization of the Ag-Cu metallic flm and (2) ionic liquids as an anion-blocking layer to achieve bistability in the switching mirror. Although the timescale on which this transition occurs in the BREM is rather slow (several min) at present, there appears to be considerable room for improvement through the choice of ionic liquids and organically-modifed electrodes. In view of their long memory effects and electrochemical stabilities, the switchable mirrors could fnd numerous applications, such as smart windows for energy saving buildings and automobiles, wireless heaters, reflective displays, and switchable mirrors for optical systems.

chemsum

{"title": "Switchable silver mirrors with long memory effects", "journal": "Royal Society of Chemistry (RSC)"}

chemical_proteomics_reveals_antibiotic_targets_of_oxadiazolones_in_mrsa

## Abstract: Phenotypic screening is a powerful approach to identify novel antibiotics against methicillin-resistant Staphylococcus aureus (MRSA) infection, but elucidation of the targets responsible for antimicrobial activity is often challenging in the case of compounds with a polypharmacological mode-of-action. Here, we show that activity-based protein profiling maps the target interaction landscape of a series of 1,3,4-oxadiazole-3-ones, identified in a phenotypic screen to have high antibacterial potency against multidrug resistant S. aureus. In situ competitive and comparative chemical proteomics with a tailor-made activity-based probe, in combination with transposon and resistance studies, revealed several cysteine and serine hydrolases as relevant targets. Our data showcase oxadiazolones as novel antibacterial chemotype with a polypharmacological mode-of-action, in which FabH, FphC and AdhE play a central role. The emergence of multidrug resistant bacteria in parallel with a dearth of new antibiotic drug approvals may become one of the biggest health care problems of the 21 st century. Among Gram-positive pathogens, methicillinresistant Staphylococcus aureus (MRSA) continues to be the most worrisome. Recent data indicate that in 2019 drugresistant staphylococcal infections, due predominantly to MRSA, were associated with a staggering 750,000 deaths worldwide. 4 New antibiotics with unprecedented modes-ofaction (MoAs) are urgently required to counteract antimicrobial drug resistance. Target-based screening is commonly applied to identify small molecules as chemical starting points (hits) in traditional drug discovery projects, but this strategy is less successful in antibiotic research. 5 Phenotypic screening has instead emerged as a promising approach to identify antibiotics with novel MoAs. A challenging aspect of phenotypic drug discovery is, however, to elucidate the primary targets responsible for the antimicrobial activity observed. Recently, chemical proteomics has emerged as a powerful chemical biology technique to map target interaction landscapes of experimental drugs, including compounds with antibacterial activity. 12,13 Inspired by these established and emerging concepts, we combined phenotypic screening with chemical proteomics to discover new MRSA antibiotics and their interacting proteins. To this end, a focused library of 352 small molecules derived from our in-house drug discovery programs was constructed. This compound set was first screened at 100 µM for antibacterial activity against MRSA USA300 (Figure 1a). This revealed 25 compounds that prevented bacterial growth. Subsequently, the minimum inhibitory concentration (MIC), which is the lowest concentration at which bacterial growth is inhibited, was determined for each of the 25 hits. Benzyl (4-(5-methoxy-2-oxo-1,3,4-oxadiazol-3(2H)-yl)-2-methylphenyl)-carbamate 1 was the most potent antibacterial compound with a MIC of 6.25 µM (2.2 µg/mL). 1 contains an oxadiazolone moiety that previously has been shown to covalently react with catalytically active serine and cysteine residues in enzymes (Figure 1b), 14,15 and has antibacterial activity in Mycobacteria. To determine the structure-activity relationship and optimize the potency of 1, a series of 61 derivatives was synthesized and tested for antimicrobial activity against MRSA USA300 and S. aureus ATCC 29213 strains (Table S1-S6, Supplementary Data 1). The oxadiazolone group and the 2-methylphenyl moiety were both found to be crucial for activity. The benzylcarbamate could be changed to a phenylamide without losing activity. This led to the identification of 2 as a simplified scaffold with comparable antibacterial activity. Subsequent systematic modification of 2 (Figure 1c) resulted in the di scovery of N-(4-(5-methoxy-2-oxo-1,3,4-oxadiazol-3(2H)-yl)-2-methylphenyl)-3-phenoxybenzamide 3 as our lead compound with a 16-fold improvement in potency in both MRSA USA300 (MIC = 0.8 µM/0.3 µg/mL) and the S. aureus ATCC 29213 strain (MIC = 1.6 µM) compared to hit 1. Extended screening of 3 against other clinically relevant pathogens revealed the specific anti-staphylococcal activity of the oxadiazolones (Table 1). 3 was highly potent against a variety of S. aureus strains, including vancomycin-resistant strains and clinical isolates (Supplementary Data 2). Of note, 3 was generally found to be more potent against antibiotic resistant strains compared to wildtype S. aureus. 3 was able to time-dependently kill 99% of bacteria over the course of 24 hours, starting with a 10 6 CFU/mL inoculum (Figure 2a). Furthermore, 3 exhibits a relatively low cytotoxicity and is non-hemolytic (Table S7, Figure S1). Next, we set out to generate strains resistant to 3 to investigate both the rate and mechanism of resistance development. MRSA USA300 was serially passaged daily in the presence of sub-MIC concentrations of compound yielding resistant mutants after 4 weeks (Figures 2b and S2). In comparison, resistance development for the control compound daptomycin, a clinically-used lipopeptide antibiotic, was found to be slower and did not exceed 8x MIC. This is commonly observed in cell membrane targeting antibiotics. 17,18 Of note, after four days the resistance towards 3 stabilized for several weeks before progressing to significantly higher values. This may indicate that multiple mutations are required to fully induce resistance, possibly suggesting a polypharmacological MoA. 3-resistant mutant strains did not show cross-resistance with commonly administered antibiotics (Supplementary Table 8). Together with the high activity of 3 against multidrug resistant S. aureus strains, these observation point to a unique MoA. Having established that the oxadiazolones are potent antibiotics against various pathogenic S. aureus strains, we set out to identify interaction partners using activity-based protein profiling (ABPP). The oxadiazolone moiety covalently reacts to catalytically active amino acids in enzymes, therefore we hypothesized that a strategically positioned alkyne ligation handle on the scaffold of 3 could be used to introduce a fluorescent or affinity tag (e.g., biotin) to visualize small molecule-protein interactions in living systems (Figure 3a). To this end, the meta-phenoxy group of 3 was substituted with an alkyne, resulting in activity-based probe 4 (Figure 3b). The antibacterial activity of 4 was confirmed in MRSA (MIC = 3.1 µM). The probe was subsequently used in an in situ competitive ABPP workflow (Figure 3a). 19 Briefly, MRSA USA300 was cultured until exponential phase (OD600 = 0.7) and treated with competitor 3 or DMSO, followed by labeling with probe 4. Bacteria were lysed and the probe-labeled proteins were conjugated to a reporter tag (fluorophoreazide or biotin-azide) via copper-catalyzed azide-alkyne cycloaddition (CuAAC) ("click") chemistry. When coupled to a fluorescent Cy5 reporter group, ABPP enables visualization of probe-labeled proteins by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and in-gel fluorescence scanning. This resulted in clear labelling by 4 of several proteins, of which most were dose-dependently outcompeted by 3 (Figure 3c). To identify the probe-labeled proteins, we coupled the probe 4-labelled proteins to a biotin reporter group, which allows affinity enrichment and identification of probelabeled proteins by mass spectrometry (MS)-based proteomics. 20 Around 30 proteins were found to be significantly enriched (p < 0.05, > 2-fold enrichment) by probe treatment (Figure 3d, Supplementary Data 3). Pretreatment with 3 significantly inhibited (p < 0.05, > 2-fold inhibition) the labeling of 10 proteins by probe 4 (Figure 3e), suggesting that these proteins are interaction partners of oxadiazolone 3. The Fph proteins (B, C, E, H) were recently discovered and annotated in MRSA as fluorophosphonate binding hydrolases. 21 FphB was found to be a fatty acid metabolizing virulence factor, while FphE activity has been used to phenotypically characterize MRSA through imaging. 22 Target proteins HZ1 and HH9 are reported to have hydrolase activity (Table 2), but their biological function has not been extensively studied. IB7 is a putative acetyl-CoA cacetyltransferase with thiolase activity, 23 while FI2 is an uncharacterized protein. FabH also known as 3-oxoacyl-[acyl-carrier-protein] synthase 3 is an essential enzyme that initiates bacterial fatty acid synthesis 24 , and has recently been explored as a drug target. AdhE is an aldehyde alcohol dehydrogenase, essential in facultative anaerobic organisms in anaerobic conditions. 28,29 Both FabH and AdhE are known to metabolize substrates using an active site cysteine. To confirm the identity of the probe targets with gelbased ABPP, we screened the probe-labeled proteome of nine transposon mutants of MRSA that lack the gene encoding one of the identified target proteins of 3 (Figure 4a). The labeling of AdhE, FphB, FphH, FI2 and HZ1, but not FphC and HH9, could be attributed to specific fluorescent bands on SDS-PAGE (Figure 4b). The lower resolution of gel-based ABPP (overlapping bands) or insufficient sensitivity compared to MS-based ABPP may explain why FphC and HH9 were not identified on gel. Since FabH is essential for MRSA viability, no transposon mutant is available for this protein. Instead, we confirmed the identity of FabH on gel by competitive ABPP using the selective FabH inhibitor Oxa2 (Figure S3). To assess which target proteins were responsible for the antibiotic effect, we hypothesized that the protein inhibition profile of potent oxadiazolones (MIC ≤ 12.5 µM) would be different compared to the interaction profile of their close analogues with no activity (MIC > 50 µM). In a competitive chemical proteomics set-up, we, therefore, compared the interaction profile of three inactive derivatives (5-7) with three active compounds (1-3) (Figures 5a and S4). Strong FphB inhibition was seen in the samples pretreated with 1, but not by the other compounds F12, IB7, HH9 and HZ1 were not significantly inhibited by the bioactive oxadiazolone 3, but did show engagement by the inactive compounds 5, 6 or 7. FphE and FphH were strongly inhibited by all compounds. These observations in combination with the viability of the transposon mutants suggest that FphB, IB7, HH9, F12, FphE, FphH and HZ1 are not essential for the antimicrobial activity of 3. FabH was significantly engaged, but not fully, by all compounds. Since the transposon mutant of FabH is not viable, this implies that partial inhibition of FabH activity could contribute to the bioactivity of the oxadiazolones, however, it was not sufficient to kill MRSA by itself. Finally, significant inhibition of FphC and AdhE labeling was only found by the bioactive compounds, but not by the inactive compounds 5-7 (Figure 5b). Next, we tested whether antibiotic activity of the inactive compounds 5-7 could be induced in the FphC and AdhE transposon mutants. Gratifyingly, it was observed that both 6 and 7 showed increased antimicrobial activity in both transposon mutants (Table 3, Table S10), but not in a FphB transposon mutant, which was taken along as a negative control. Of interest, compounds 6 and 7 did not become as active as 3. Compound 5, which has no activity on AdhE and very weak activity on FphC (< 20%), remained inactive in all individual transposon mutants. Although we cannot exclude that other proteins also play a role, we interpret these data to mean that combined engagement of FphC and AdhE is required for antimicrobial activity of the oxadiazolones. Principal component analysis of the chemical proteomics data confirmed that inhibition of AdhE and FphC activity was associated to a large extent with the antibacterial activity (Table S9, Figure S7). Thus, our chemical proteomics data reveal that multiple targets (in particular, FphC, AdhE and FabH) play a role in the observed antimicrobial activity of the oxadiazolones. To further test our hypothesis, we investigated whether FphC and AdhE activity was changed in two of the 3resistant MRSA strains we generated compared to WT MRSA (Figures 5c, 5d and S5, Supplementary Data 3c, 3d and 4). Using chemical and global proteomics it was observed that AdhE activity was significantly decreased, while its protein abundance was upregulated in the two resistant strains. FphC-activity was also reduced, but to a lower ex-tent. Interestingly, FabH protein levels were significantly increased in the resistant strains, which was accompanied by cross-resistance of these strains to the FabH inhibitor Oxa2 (8x increase in MIC). Taken together, these data suggest that combined inhibition of FabH, FphC and AdhE contributes to the antimicrobial activity of compound 3. To summarize, a phenotypic screen of a focused library led to the identification of oxadiazolones as a new chemotype with antibiotic activity against pathogenic, multidrug resistant S. aureus strains and clinical isolates. A medicinal chemistry program combined with chemical proteomics led to the identification of compound 3 as the most potent antibiotic capable of interacting with multiple bacterial cysteine and serine hydrolases in a covalent manner. Three complementary lines of investigation point to FabH, FphC and AdhE as playing central roles in the antimicrobial activity of 3 and structurally similar oxadiazolones. i) comparative chemical proteomics, ii) gain of function in transposon mutants, and iii) resistance-induced proteomic changes. FabH has previously been identified as a drug target, whereas the function of AdhE and FphC has been less well explored. Recent studies implicate AdhE as a virulence factor in E. coli. 30 FphC is a membrane-bound serine hydrolase with unknown function. Of note, we cannot rule out that other factors, not detected by our chemical proteomics approach, may also contribute to the antibacterial effect of 3, such as non-covalent interactions with proteins or other classes of To conclude, our findings further highlight the value of synthetic compound libraries as an excellent source for antibiotic drug discovery complementary to natural products. By applying comparative and competitive chemical proteomics, using a new tailor-made activity-based probe with a strategically positioned ligation tag, we successfully elucidated the polypharmacological mode-of-action of the oxadiazolones and identified their targets in MRSA. Notably, a target-based approach alone would have not been able to uncover the mode-of-action of the oxadiazolones, thereby showcasing the power of chemical proteomics as a valuable chemical biology technique for antibiotic drug discovery. Future experiments are directed towards understanding the biological role of these targets and further optimization of the compounds as viable drug candidates.

chemsum

{"title": "Chemical Proteomics Reveals Antibiotic Targets of Oxadiazolones in MRSA", "journal": "ChemRxiv"}

radical-polar_crossover_hydroetherification_of_alkenes_with_phenols

## Abstract: We disclose a general electrocatalytic hydroetherfication for modular synthesis of alkyl aryl ethers by utilizing a wide range of alkenes and phenols. The integration of the two involves an electrochemically instigated cobalt-hydride catalyzed radical-polar crossover of alkenes that enables the generation of key cationic intermediates, which could readily be entrapped by challenging nucleophilic phenols. We highlight the importance of precise control of the reaction potential by electrochemistry to obtain optimal chemoselectivity. Notably, this reaction system is pertinent to latestage functionalization of pharmacophores that contain alkyl aryl ethers which has constantly been challenged since traditionally unconventional methods. ## Introduction Alkyl aryl ethers has increasingly shown significant importance in agrochemical and pharmaceutical industries as their structural motifs in synthesis of biorelevant compounds have become ubiquitous (Fig. 1A). Early methods to synthesize alkyl aryl ethers include Williamson ether synthesis, SNAr (nucleophilic aromatic substitution) and Mitsunobu reactions that all suffer from limitations to electronically unbiased or sterically hindered products. In response to these traditional methods, transition metal-mediated C(sp 2 )−O cross coupling of alcohols with aryl halides has been most widely reported upon judicious choice of ligand and base additives to address challenging canonical step in either oxidative addition or reductive elimination. As a complementary strategy in terms of retrosynthetic bond disconnection, C(sp 3 )−O cross coupling of phenols with redox-active esters (RAEs) in lieu of aryl halides has also been presented. On the other hand, alkyl aryl ethers can be formed by acid-catalyzed addition reactions of phenols with alkenes (Fig. 1B). As abundant precursors, alkenes can in principle offer a wide range of alkyl counterparts and are further complementary in forming sterically hindered ethers via highly substituted carbocation intermediates. However, such strong Lewis-or Brönsted acid catalyzed addition is predominantly accessible to only monosubstituted terminal alkenes while internal alkenes are less reactive and prone to isomerization. Moreover, acid-catalyzed additions suffer from competitive chemoselectivity between forming the C−O bond and undesired C−C bond with ortho or para C−H bond of phenols. In this context, catalytic hydroetherification of a wide range of alkenes with phenols as the nucleophile is particularly attractive due to the accessibility to hindered alkyl aryl ethers but remains to be scarce. In addition, this approach has yet been compatible to synthesize key structural motifs in bioactive and pharmaceutical compounds. To this end, we were drawn to the possibility that a catalytic system under mild conditions that can generate diverse cationic alkyl intermediates to their respective olefins and compatible to a wide variety of phenolic nucleophiles. Recently, Pronin, Shigehisa, Hiroya, Zhu and others have successfully expanded Mukaiyama's hydrogen atom transfer from a metal hydride (MHAT) paradigm to nucleophilic alkene hydrofunctionalization under oxidizing conditions (Fig. 1C). These methods however involve the use of chemical oxidants with minimal control over the oxidizing potential that dictates the bimetallic catalyst-controlled oxidation, which consequently necessitates the use of solvent amounts of nucleophiles to engage with unactivated multisubstituted alkenes. Thus, the development of a new catalytic strategy that allows proficient and controllable generation of carbocations from the corresponding alkenes remains a major challenge for hydrofunctionalization with important but challenging nucleophiles such as phenols. Herein we report a cobalt-catalyzed electrochemical radical polar crossover hydroetherification of alkenes in which a controllable electricity drives systematic consecutive oxidations of cobalt catalyst to generate carbocationic species from a comprehensive class of olefins, which would subsequently be entrapped by phenols to afford the desired alkyl aryl ethers (Fig. 1D). By manipulating the oxidation potential at a required minimum, the current system is found to be compatible with oxidatively sensitive phenols including heterocyclic derivatives with exclusive chemoselectivity towards C−O bond formation. Furthermore, this operationally simple protocol also allowed facile and modular access to pharmaceutical ethers via late stage functionalization. ## Results and Discussion To design an electrochemical system for the generation of high valent cationic species, we envisioned a prospective catalytic cycle wherein a Co(II) salen catalyst I would first be anodically oxidized to form Co(III)-hydride III upon addition of Si−H bond of hydrosilane II (Fig. 2A, anodic event 1). This Co(III)hydride III would perform alkene MHAT with simultaneous formation of a metallo/organic radical pair IV, which is known as "solvent-caged radical pair". The second anodic oxidation of either IV or cage Initial exploration of the reaction using allylbenzene (3), 10 mol% of Co(III) salen catalyst Co-1 with 3.0 equiv of methylphenylsilane (2) and TBAPF6 electrolyte in THF upon electrolysis under constant current of 1 mA gave 20% of the desired hydroetherification product 5 as a promising lead result (Fig. 2B). This observation led us to examine the oxidation step involved in the generation of requisite intermediates that display reactivity reminiscent of a carbocation using cyclic voltammetry (CV). CVs were performed using 3 mM cobalt catalyst in the absence (gray line), in the presence of silane 2 (red line) and in the presence of both 2 and 3 (blue line) at a scan rate of 100 mV/s (Fig. 2C, left box). A quasi-reversible feature at −0.05 V shown for the Co-1 (gray line) in the absence of silane and alkene where designated as Co(II)/Co(III) redox couple. The addition of 1.0 equiv of 2 resulted in a significant change, displaying an irreversible oxidation with a peak potential of 0.15 V (red line) which we attributed to the oxidation of Co(III)−H species (Co-2). Notably, the addition of 3 resulted in a decrease in current of this feature as well as a shift in anodic peak potential, with the appearance of an additional redox wave (blue line, Ep,ox = 0.32 V). Accordingly, the CV behavior of Co-1 in the presence of both 2 and 3 was further interrogated under increasing concentration of 3 (Fig. 2C, right box). In this experiment, a sequential increase in current at the second redox feature with a simultaneous decrease at the first oxidation peak was observed as more 3 is added. This is indicative of the generation of putative MHAT complex Co-3, which is proposed to be capable of generating corresponding Co(IV) or carbocation intermediate upon electrochemical oxidation. Indeed, the observed oxidation peak potential of putative Co-3 intermediate is found to be on par with that of previously reported Co(III)-alkyl complex Co-4 (Ep = 0.11 V vs Fc/Fc + ). Based on these voltammetric studies, we hypothesized that the electrochemical oxidation of putative MHAT Co(III)-alkyl intermediate can bypass the bimetallic disproportionation mechanism which has been dictated by the use of chemical oxidants. We examined the viability of this protocol toward more sterically demanding 1,1-disubstituted alkenes for hindered alkyl aryl ether synthesis (Fig. 3A). Accordingly, we chose alkene 6 as the model substrate. The Co(II) analogues were employed as the catalyst precursor due to the ease of preparation and feasibility of being oxidized under anodic oxidation conditions to generate corresponding Co(III)hydrides. Combining the phenol 4 with 6 together with 10 mol % of Co-5 as the catalyst with 3.0 equiv of silane 2 in THF/HFIP (3:1) solution under the application of constant cell voltage of 2.3 V (corresponding to an initial anodic potential Ea,i ∼ 0.5 V vs Fc +/0 ) furnished the hydroetherification product 7 in 35% yield (entry 1). A marginal increase in reactivity was observed with Co-6 as the catalyst (entry 2). Among a variety of solvent systems screened, we found that employing toluene/HFIP co-solvent dramatically increased reactivity (entry 3, see Supporting Information for full data). Further evaluation of cobalt catalysts led to an improvement in reaction efficiency particularly with Co-8 catalyst, which provided 7 in 67% yield (entries 4−6). We note that the yields were improved with a decrease in reaction temperature, providing 92% of isolated yield of 7 under −20 o C (entries 7−8). The increase of voltage input from 2.3 V to 3.0 V (Ea,i from ∼ 0.5 V to 0.8 V vs Fc +/0 ) resulted in decrease in reaction efficiency with poor mass balance presumably due to unwanted side reactions caused by high anodic potential (entry 9). Finally, we also found that the model reaction was smoothly proceeded under reduced catalyst loading (5 mol %) albeit in slight decrease in product yield (entry 10). Subsequently, a set of control experiments was conducted under a series of conventionally employed chemical oxidants without an electrical input for comparative purposes (Fig. 3B). None of these chemical oxidants promoted the reaction efficiency comparable to the current electrochemical protocol, while significant amounts of homocoupled side product 8 were obtained in most cases. This may suggest the use of such chemical oxidants partially obviates the second anodic SET of the reaction, via radical polar crossover, when outcompeted by other side reaction pathways. This highlights employing electrical current as the oxidant circumvents any unwanted pathways and its crucial role in chemoselectivity of our reaction. We then sought to evaluate the reaction scope with respect to both phenols and alkenes (Table 1). Along with 4, uniformly high reactivities were observed upon variation of the para substituent of the phenol (7, 10−17). The substituents on ortho or meta position did not affect the reaction efficiency (18−19). The reaction accommodated naphthols as the oxygen nucleophiles (20−21). It was notable to see that hydroxy-substituted heterocycles were also compatible to afford corresponding alkyl heteroaryl ether products (22−23). An isobutene (24) was also a competent substrate as well, enabling a simple O-tert-butylation of phenol. The hydroetherification also tolerated thiophene moiety to afford product 27. Following our investigations on 1,1-disubstituted alkenes, we further applied this methodology to more sterically congested alkene substrates. We found that the current protocol was effective to both acyclic (28, 30, 32) and cyclic (34) trisubstituted alkenes. Notably, tetrasubstituted alkene 36 also underwent hydroetherification smoothly to produce the corresponding tertiary alkyl aryl ether 37 in good yield. This result further supports the intermediacy of the carbocation generated by an electrochemical radicalpolar crossover mechanism. Besides multisubstituted alkenes, a series of monosubstituted alkenes underwent hydroetherification with phenols. Products 38 and 39, only containing a simple alkyl chain, were afforded in excellent yields. A variety of functional groups including methoxyarene (40), sulfonamide (41), silyl protected alcohol (42), thiophenyl ester (43) or even fructose derivative (44) smoothly participated the reaction without difficulty. This hydroetherification method also proved efficient in engaging styrenes as substrates, which are considered to be more challenging due to the unfavorable homocoupling or reduction caused by intrinsically high stability of benzylic radicals generated by cage-escape mechanism (vide supra, Fig. 2A). Under slightly modified reaction conditions, styrenes bearing a series of functional groups at the para position were converted to the desired ether products in moderate to good yields (45−49). Vinylarenes possessing oxidatively sensitive functional groups on either meta-or ortho-position were smoothly converted into desired products (50−51). Vinyl-substituted pyridine also underwent the hydroetherification (52). The current method could be applied to 1,1-and 1,2-disubstitued aryl alkenes (53−56). In addition, it was found that hydroetherification of ,-disubstituted styrene (57) proceeded at the -position with an exclusive regioselectivity (58), again suggesting the generation of the carbocationic intermediate. It should be noted that the vinylarenes were also compatible with heterocyclic (59) or electronically biased phenols (60) as coupling partners. Lastly, radical clock substrate 61 underwent rupture of the three-membered ring under standard reaction conditions to furnish 62, suggesting key radical intermediacy prior to the generation of carbocation. Finally, the high functional group tolerance of the current hydroetherification was highlighted by its application in late-stage functionalization of natural product structures and densely functionalized pharmacophores (Table 2). A wide range of bio-relevant structures were all transformed into corresponding hydroetherification product with ease (63−71). It is worth noting that fenofibrate (74), an oral medication for dyslipidemia could directly be synthesized in a single hydroetherification step, highlighting the utility of our method as a tool for late-stage drug synthesis and modification. ## Conclusion In summary, we present a modular approach to synthesizing alkyl aryl ethers by integrating a wide range of alkenes and phenols. This protocol involves successive oxidative SET reactions to enable conventional radical polar crossover reactions to allow direct entrapment of cationic alkyl intermediates by challenging nucleophilic phenols. The experimental data suggest that using an electrochemical anode as the oxidant may not only be sustainable but also optimal for high reactivity and chemoselectivity. This versatile hydroetherification method can be manipulated to target complex molecules including bioactive compounds, suggesting that the strategy for nucleophilic phenols as synthetic drug precursors would enlarge a library of potentially new pharmaceutical agents. ## Methods Detailed information of experimental procedures as well as analytical data are provided in the Supplementary Information.

chemsum

{"title": "Radical-Polar Crossover Hydroetherification of Alkenes with Phenols", "journal": "ChemRxiv"}

cochaperones_enable_hsp70_to_use_atp_energy_to_stabilize_native_proteins_out_of_the_folding_equilibr

## Abstract: The heat shock protein 70 (Hsp70) chaperones, vital to the proper folding of proteins inside cells, consume ATP and require cochaperones in assisting protein folding. It is unclear whether Hsp70 can utilize the free energy from ATP hydrolysis to fold a protein into a native state that is thermodynamically unstable in the chaperone-free equilibrium. Here I present a model of Hsp70mediated protein folding, which predicts that Hsp70, as a result of differential stimulation of ATP hydrolysis by its Hsp40 cochaperone, dissociates faster from a substrate in fold-competent conformations than from one in misfolding-prone conformations, thus elevating the native concentration above and suppressing the misfolded concentration below their respective equilibrium values. Previous models would not make or imply these predictions, which are experimentally testable. My model quantitatively reproduces experimental refolding kinetics, predicts how modulations of the Hsp70/Hsp40 chaperone system affect protein folding, and suggests new approaches to regulating cellular protein quality.The discovery of chaperones and their roles in assisting protein folding amended the long-held view that proteins spontaneously fold into their native structures 1-3 . Large, multi-domain proteins may take many hours to fold, or fail to fold properly altogether on their own 2,4 . ATP-consuming chaperones-including Hsp70s-provide critical assistance in the in vivo folding and the biological functions of broad sets of substrate proteins 3 . Extensive experimental studies have firmly established that within the same length of time, more denatured substrate proteins refold in the presence of the Hsp70 chaperones than in their absence 5,6 . Despite tremendous progress in the mechanistic studies of the Hsp70 chaperones 2,7,8 , including the development of theoretical models [9][10][11][12] , it remains unclear why ATP consumption is indispensable to these chaperones; many enzymes catalyze chemical reactions without consuming free energy. Recently it was demonstrated that chaperones such as GroEL and Hsp70 depend on continuous ATP hydrolysis to maintain a protein in a native state that is thermodynamically unstable 13 , but it is unknown how Hsp70 can utilize the ATP free energy to alter the folding equilibrium. In addition, Hsp70s require Hsp40-also known as J proteins 14 -cochaperones in assisting protein folding. It is yet unexplained why cochaperones are absolutely necessary.The Hsp70 chaperones, such as the bacterial DnaK, consist of an N-terminal nucleotide binding domain (NBD) and a C-terminal substrate binding domain (SBD). The Hsp70 SBD adopts an open conformation when its NBD is ATP-bound (I call the Hsp70 to be in the ATP-state), which allows the substrate to bind and unbind at high rates, whereas when the NBD is ADP-bound (ADP-state), the SBD changes to a closed conformation, rendering both binding and unbinding orders-of-magnitude slower 6,15,16 . Hsp70s have low basal ATP hydrolysis activities 17 . The Hsp40 cochaperones, such as the bacterial DnaJ, can drastically stimulate the ATPase activity of Hsp70 using their N-terminal J domain (JD) 18,19 , shared by all Hsp40s (hence the name J proteins) 14 . Hsp40s also have a C-terminal domain (CTD) that can bind to denatured proteins 20,21 . Both Hsp70 and Hsp40 recognize exposed hydrophobic sites 7,22,23 . As a result, they can distinguish different protein conformations using the corresponding difference in the exposed hydrophobic sites. For example, Hsp70 has been shown to bind to both unfolded and partially folded, near native protein structures, but not to native structures 24,25 . Hsp40 and Hsp70 may simultaneously bind to different segments of the same substrate molecule, and the consequent spatial proximity then facilitates the J domain binding to Hsp70 and accelerating its ATP hydrolysis . Following ATP hydrolysis, the chaperone returns from the ADP-state to the ATP-state through nucleotide exchange, which is often catalyzed by nucleotide exchange factors (NEF) such as the bacterial GrpE 29,30 . It is unclear whether Hsp70 can use the free energy from ATP hydrolysis to drive its substrate protein toward the native state, N, and away from the misfolded state, M, such that f N /f M > f N,eq /f M,eq , where f S is the fraction of the substrate in state S at the steady state of Hsp70-mediated folding, and f S,eq is the corresponding fraction at the folding equilibrium in the absence of the chaperone. Previous models 10,31 mostly considered the chaperone as an unfoldase/holdase-which need not consume free energy-that pulls the substrate out of the misfolded state and holds it in an unfolded state. It was proposed that the free energy from ATP hydrolysis was used to achieve ultra-affinity in substrate binding 9,32 . As an unfoldase/holdase, Hsp70 would also pull the substrate out of the native state into the unfolded state; unless Hsp70 has a higher affinity for the native substrate than for the misfolded substrate, which contradicts experimental observations, these models would predict f N /f M ≤ f N,eq /f M,eq . Here I propose a model of Hsp70-mediated protein folding, in which Hsp70 and Hsp40 together constitute a molecular machine that uses the free energy from ATP hydrolysis to actively drive a protein toward its native state, so that f N /f M > f N,eq /f M,eq . It suggests that without Hsp40, Hsp70 alone cannot change the ratio f N /f M from the equilibrium value f N,eq /f M,eq . My model thus answers the question why Hsp70 requires both the Hsp40 cochaperones and ATP consumption in assisting protein folding. My model explains the puzzling non-monotonic dependency of folding efficiency on the chaperone and cochaperone concentrations. It makes quantitative predictions on how protein folding is affected by modulations of the chaperone environment, including changes in the ATPase activity or the nucleotide exchange rate of Hsp70. These predictions may be readily tested by experiments, and inform rational approaches to manipulating chaperone-mediated protein folding. ## Results My model is based on two assumptions supported by experimental observations. The first assumption is that a substrate protein can adopt two additional conformational states besides the misfolded, M, and the native, N, states: the unfolded and misfolding-prone state, U, and the fold-competent state, F. A protein in the F state is unfolded but poised to fold into the native state (Fig. 1a,b). Such intermediate states of folding have been observed experimentally 4 . Conformational transitions can occur between M and U, between U and F, and between F and N (Fig. 1b). The second assumption is that Hsp40's affinity for a substrate protein is higher if the substrate is in the misfolding-prone conformation than if it is in the fold-competent conformation. Experimental observations suggest that the misfolding-prone conformation is less compact and exposes more hydrophobic sites-thus providing more accessible sites for Hsp40 binding-than the fold-competent conformation 4 . Consequently, a substrate in the U state is more likely to be Hsp40-bound than one in the F state. The key idea of my model is a mechanism by which Hsp70 actively drives a substrate toward the native state and away from the misfolded state. Based on the above assumptions, an Hsp70 molecule bound to a substrate molecule in the U state will on average have substantially higher ATP hydrolysis rate-because of the higher probability of cis stimulation by an Hsp40 molecule bound to the same substrate molecule-than if it is bound to a substrate molecule in the F state. If the nucleotide exchange rate is between these two hydrolysis rates, an Hsp70 bound to a substrate in the U state will be driven toward the ADP-state, where it slowly dissociates from the substrate, while an Hsp70 bound to a substrate in the F state will be driven toward the ATP-state, where it rapidly dissociates from the substrate. This enables Hsp70, when bound to substrate molecules, to act like a Maxwell's demon 33 : it quickly releases the fold-competent molecules so that they can fold, but it retains the misfolding-prone molecules to prevent misfolding. By this mechanism, Hsp70 drives the folding along the reac- where S • C • X represents the complex between a substrate in conformation S and the chaperone C bound to nucleotide X = ATP, ADP (Fig. 1b). One ATP molecule is consumed in this reaction path and the free energy is used to compel the substrate into the native state. My model predicts that Hsp70 need Hsp40 in order to alter the folding equilibrium. For Hsp70 to drive substrate folding toward the native state, the above mechanism does not require that more Hsp70 molecules bind to a substrate in the U state than to a substrate in the F state, which is true and reflected in previous models; instead, it requires that an individual Hsp70 molecule, when bound to a substrate, dissociates slower if the substrate is in the U state than if it is in the F state. Hsp40 thus plays a critical role because a substrate-bound Hsp70 distinguishes, probabilistically, between the U and F states of the substrate by sensing whether an Hsp40 is also bound to the same substrate. Defining the excess free energy of folding as where R is the gas constant and T the temperature, it can be shown algebraically (see Methods) and numerically (Fig. 1c) that without cochaperones, ΔΔG = 0. This prediction is consistent with the results from the single-molecule experiment of DnaK-mediated refolding 34 , where DnaK alone in the presence of ATP was unable to alter the ratio of the misfolded and folded fractions. In order to simplify calculations using my model, I assume that a protein's hydrophobic binding sites for Hsp40 and Hsp70-which can be exposed in the U and F states-become entirely buried upon folding (F → N) and misfolding (U → M), thus a protein in the M and N states has zero exposed hydrophobic binding sites, and it does not bind to either Hsp70 or Hsp40 (Fig. 1a). The hydrophobic burial in the misfolded state may be due to intramolecular contacts between incorrectly folded domains within a monomeric protein, or due to intermolecular contacts between different protein molecules as a result of oligomerization or mild, reversible aggregation. A protein may also aggregate irreversibly, and will not refold even in the presence of the chaperone system 5 . A protein in such an irreversibly aggregated state may have a prohibitive kinetic barrier to return to the M or U 2). (b) The transitions between the microscopic states in the chaperonemediated folding pathway. S•C•X represents the complex between the substrate in the conformational state S (=U, F) and the Hsp70 chaperone (denoted as C) bound to nucleotide X (=ATP, ADP). The transitions between S and S•C•X correspond to the chaperone binding to and unbinding from the substrate. The transition of S•C•ATP to S•C•ADP corresponds to ATP hydrolysis, and its reverse, nucleotide exchange. Hsp70 binding stabilizes the substrate in the intermediate states, thus catalyzing the folding reaction. Hsp40 (J) can form a ternary complex with the substrate and Hsp70-thus stimulating ATP hydrolysis-if the substrate is in the U state, but not if the substrate is in the F state. Differential ATP hydrolysis by Hsp70 bound to the substrate in the U and F states drives the refolding through the pathway highlighted in red. The lengths of the reaction arrows are linear with respect to the logarithms of the exemplary rate constants (in 1/s) for the DnaK/DnaJ/GrpEmediated refolding of luciferase at 25 °C (Tables 1 and 2). (c) Without cochaperones, Hsp70 cannot alter the balance between folding and misfolding. DnaK binding to the intermediate states decreases both the native states. It is thus unamenable to Hsp70-assisted refolding and not included in my model. The quantitative details of my model are given in Methods. I applied my model to the analysis of DnaK/DnaJ/GrpE-mediated refolding of luciferase 5 . Most of the relevant kinetic parameters at 25 °C for this bacterial Hsp70 system have been carefully determined experimentally 35 (Table 1). My model quantitatively reproduces the experimentally observed refolding kinetics under various conditions, capturing the slow spontaneous refolding and denaturation of luciferase, the acceleration of refolding with chaperone assistance, and the necessity of GrpE (Fig. 2a). The intermediate conformations U and F in my model in the case of luciferase may correspond to the experimentally identified intermediate conformations I 2 and I 1 of luciferase 4 : the free energy difference between N and F at 25 °C, according to the fitted parameters, is 20 kJ/mol, close to the experimental value of 15 kJ/mol between N and I 1 , measured at 10 °C. Consistent with previous experimental observations 31 , my model suggests that the Hsp70-mediated refolding proceeds in two steps: (1) rapid unfolding of the misfolded substrate, stabilized by the ADP-bound DnaK, followed by (2) slow conversion to the native state (Fig. 3a). Next, I used my model to analyze how the refolding kinetics and yield change with respect to the DnaK concentration, which have been experimentally studied for LucDHis6, a variant of luciferase 31 . The refolding data fits well with the kinetic equation of a first-order reaction, , and the rate constant k is approximately unchanged across different DnaK concentrations 8,36 . DnaK thus affects the refolding kinetics mostly by altering the refolding yield ∞ N( ). My model captures the refolding kinetics and the refolding yield of LucDHis6 in the experimental range of DnaK concentrations (Fig. 2c); although the experiments were performed at the temperature of 22 °C, the predictions of my model using the kinetic parameters derived for 25 °C are nevertheless in quantitative agreement with the experimental data. At the steady state, the reactive flux along the ATP-driven cycle 3b) keeps the protein folding out of equilibrium, elevating the native population above and suppressing the misfolded population below their respective equilibrium values (Fig. 2c). Notably, the refolding yield peaks around [DnaK] = 1 µM, and it decreases at higher DnaK concentrations. This non-monotonic dependence on the DnaK concentration was also reported for the wildtype luciferase while this work was under review 36 . According to my model, the excess free energy at the steady state always increases with increasing DnaK concentrations, but the native population reaches a maximum and then decreases (Fig. 2c), because at high DnaK concentrations, the substrate is trapped in the DnaK-bound state and thus prevented from folding into the native state. I used my model to estimate the ATP consumption in the DnaK/DnaJ/GrpE-mediated folding, which has been experimentally measured for LucDHis6 31 (Fig. 4). My model estimates that, in the initial minutes of refolding, approximately 150 ATP molecules are consumed to refold one LucDHis6 (Fig. 4a), which is reasonably close to the experimental result of ~50 ATP molecules consumed per refolded LucDHis6 when the stoichiometry of DnaK:LucDHis6 is 1:1, significantly higher than the experimental number of ~5 when LucDHis6 is in excess of DnaK, and significantly lower than the estimates of >1000 for many other substrates in other experiments 31, . The discrepancy between the model and the experimental results may be attributable to the approximations in my model and the inaccuracies in the input kinetic parameters. ATP hydrolysis continues at the steady state and the free energy is utilized to promote the native state and suppress the misfolded state (Fig. 4b,c). As [DnaK] exceeds 1 µM, the ATP consumption rate increases rapidly without commensurate increase in the excess free energy. My analysis thus suggests that DnaK may be most free energy efficient at maintaining protein folding out-of-equilibrium when its concentration is in the sub-micromolar range, a prediction that may be tested experimentally. My model suggests that Hsp70 can keep a protein folded even if it thermodynamically tends to misfold and aggregate (Here I only consider reversible aggregation). The chaperone is thus able to play a critical role in maintaining protein conformations, not just in the folding of nascent chains 40 . Higher DnaK concentrations are required to suppress aggregation at increasing substrate concentrations (Fig. 5a) or at decreasing substrate stabilities (Fig. 5b). This may explain how cells that overexpress DnaK can tolerate higher numbers of mutations in the chaperone's substrates 41 . Because the excess free energy plateaus at high chaperone concentrations (Fig. 2c), my results imply a limit on the chaperones' capacity to prevent aggregation, in that there exists a threshold of aggregation tendency (Fig. 5a,b, the black arrows) above which the chaperone can no longer maintain high levels of native concentrations and prevent misfolding/aggregation at the same time. My model suggests that Hsp70 only drives the folding of proteins with sufficiently slow conversion between U and F states (Fig. 5c-e), implying that Hsp70 substrates tend to be slow refolding proteins (Fig. 5d). If the conversion between U and F is too fast, the chaperone diminishes, rather than increases, the native fraction in comparison to the chaperone-free equilibrium. As the conversion slows, the chaperone drives the steady state native fraction higher, but at the price of longer refolding time (Fig. 5e), a trade-off reminiscent of that between speed and specificity in the kinetic proofreading mechanism 42,43 , where the expenditure of free energy (such as from ATP or GTP consumptions) is utilized to increase the specificity of chemical reactions. My model explains the observation that folding is less efficient at both low and high DnaJ concentrations 17 (Fig. 6a). At low DnaJ concentrations, ATP hydrolysis is slow, and nucleotide exchange drives DnaK toward the ATP-state, in which it dissociates from the substrate rapidly and thus unable to prevent aggregation. At high DnaJ (red) and misfolded (blue) populations, but the ratio between the two remains unchanged from its equilibrium value: ΔΔG = 0 (orange, right y-axis). Here, I have taken to show that differential binding of Hsp70 to the substrate in different conformational states does not alter the folding/ misfolding balance. ## Reaction Parameter Luciferase LucDHis6 ## °C 30 °C 22 °C A J U , ( ) 6 1 13 22 54 but are degenerate with respect to the individual rates. I thus fix an arbitrary but plausible value for → k M U and fit → k U M to the experimental refolding data. b For the similar reason as above, I fix → k F N and fit → k N F to the refolding data. c There should be more accessible DnaK binding sites in the U state than in the F state. Here, I arbitrarily set the ratio between the two. Although the values of the other fitting parameters will change accordingly, the quality of the fit and the predictions of my model are insensitive to this ratio (at least for values between 1 and 100). d I assume that the effective distance, L, between the Hsp70 molecule and the J domain of the Hsp40 molecule bound to the same substrate molecule is L = 4.3 nm, the effective concentration is then 3 . The quality of the fit and the predictions of my model are insensitive to this parameter. ## Reaction Parameter Rate Reference . I note that the hydrolysis rates are measured without a protein substrate (indicated by the superscript w/o S), which can further accelerate the hydrolysis. Surface plasmon resonance measurement of DnaJ binding to DnaK yielded , in good agreement with our fit. b Substrate-free DnaK has a basal ATP hydrolysis rate of 0.001 s −1 , but the substrate can further accelerate ATP hydrolysis by up to 9-fold 17,53 . I take the ATP hydrolysis rate of substrate-bound DnaK to be 10-fold higher than the basal rate. The predictions of my model are insensitive to this parameter. c This is the experimental rate for the temperature T = 25 °C. For the higher temperature T = 30 °C, I used an arbitrary but reasonable 5.5-fold higher value of 10 s −1 because I could not find any reported experimental value for this temperature. d The kinetic rates for GrpE binding to DnaK were not determined, but the steady state ratio in the two-step dissociation reaction, was determined. I chose an arbitrary diffusion limited association rate for GrpE binding to DnaK in our calculations. e I note that k C (0) and E a appear too large to be physically meaningful; they should instead be taken simply as numerical parameters that yield an excellent fit of the Arrhenius equation to the experimental data. concentrations, a large fraction of the substrate in the U state is bound to DnaJ. These DnaJ-bound substrate molecules are trapped in the U state, unable to progress toward the F state, resulting in diminished folding. My model also explains the observation that folding decreases at both low and high GrpE concentrations 29 (Fig. 6b). For the chaperone to effectively assist folding, nucleotide exchange should be much slower than ATP hydrolysis when the chaperone binds to a substrate in the U state, but much faster than ATP hydrolysis when it binds to a substrate in the F state, so that the chaperone is driven toward the ADP-state in the former case, and toward the ATP-state in the latter case (Fig. 1b). At low GrpE concentrations, nucleotide exchange is slow, leaving DnaK bound to the substrate in the F state predominantly in the ADP-state-as reflected by the low population of F • C • ATP (Fig. 6b), slowing its dissociation from the substrate and thus preventing the latter from folding to 5 . The dashed lines show the spontaneous refolding and denaturation at the much lower substrate concentration of 0.032 μM (2 μg/ml), as in the corresponding experiments (empty circles and squares) 4 . The experiments of spontaneous refolding and denaturation were performed at 20 °C, lower than the temperature of 25 °C at which most of the kinetic parameters were obtained. In modeling the refolding, I assume that initially all the protein is in the misfolded state, i.e., f M (t = 0) = 1; in modeling the denaturation, I assume that initially all the protein is in the native state, i.e., f N (t = 0) = 1. (b) The refolding of a luciferase mutant, LucDHis6 31 , in the presence of DnaK at various concentrations. (c) The native fraction (f N = [N]/[S], red) and the misfolded fraction (f M = [M]/[S], blue) of LucDHis6 at the steady state of DnaK-mediated refolding at various DnaK concentrations. The corresponding fractions in the chaperonefree folding equilibrium, f N,eq and f M,eq , are shown as dashed lines. The unitless excess free energy ΔΔG/(RT) is shown in orange (right y-axis). The fractions after 80 min of refolding, starting from misfolded LucDHis6, are shown in brown. My model is in good agreement with the experimental data (filled circles), and it suggests that the refolding is still incomplete even after 80 min. The fitting parameters in my model are given in Table 2, and the conditions of the experiments considered in this paper are summarized in Table 3. the native state. At high GrpE concentrations, nucleotide exchange is fast, and DnaK is driven into the ATP-state and does not stay bound to the substrate in the U state long enough-as reflected by the decreasing population of U • C • ADP (Fig. 6b)-to prevent the substrate from aggregation. To maximize substrate folding, higher nucleotide exchange rate should accompany higher stimulated ATP hydrolysis rate (Fig. 6c-e). My model predicts that Hsp70 chaperones with higher Hsp40-stimulated ATP hydrolysis rates can drive substrate folding to higher native fractions (Fig. 6e), at the cost of higher free energy expenditure (Fig. 6f). This result explains a previous experimental observation that a small molecule that enhances ATP hydrolysis by Hsp40-bound Hsp70 can induce higher yields of substrate folding 44 . Modulation of the ATP hydrolysis or the nucleotide exchange rates by small molecules may represent a therapeutic opportunity in the treatment of misfolding-or proteostasis-related diseases 45 . ## Discussion Key to my model is the assumption that Hsp40 has different affinities for a substrate in different conformations, favoring the misfolding-prone conformation over the fold-competent conformation. This is supported by a number of experimental observations. Hsp40 binds to exposed hydrophobic sites using a Zn finger-like domain within its CTD 20 . It has been inferred from experiments that the binding of Hsp40 CTD to substrate is the strongest for unfolded peptides, weaker for partially unfolded proteins, and the weakest for native proteins, and that Hsp40 is able to distinguish the substrate conformations 22 . Hsp40 has been experimentally shown to bind to a number of denatured proteins, including denatured luciferase , but it binds to few native proteins, notably to σ 32 , which adopts a loosely folded and highly flexible conformation 46 , and to RepE, the binding to which depends on the G/F-rich domain (outside CTD) of Hsp40 instead of the Zn finger-like domain 22 . These results suggest that Hsp40 preferentially binds to loose conformations with many exposed hydrophobic sites, which, together with , by the number of refolded substrate molecules after time τ. Here, the DnaK concentration is 0.5 μM, and the other kinetic parameters are given in Tables 1 and 2. The black curve shows the number of ATP molecules hydrolyzed per DnaK molecule after the given time. The brown curve shows the number of ATP molecules consumed per one molecule of refolded LucDHis6 (right y-axis) up to the given time, which increases to infinity at the steady state because no additional LucDHis6 is refolded, yet ATP hydrolysis continues. (b,c) ATP hydrolysis at the steady state. The ATP consumption rate per substrate, at various DnaK concentrations, is shown as the black curve in panel b, and the corresponding native (red) and misfolded (blue) fractions at the steady state are shown as solid lines in panel c. The native fraction is above and the misfolded fraction is below their respective equilibrium values (dashed flat lines). The excess free energy ΔΔG/(RT) is shown in orange (right y-axis) in panel c. I can measure the chaperones' free energy efficiency in maintaining the non-equilibrium by the ratio of the ATP consumption rate to the excess free energy at the steady state (orange, right y-axis, in panel b). The arrow indicates the DnaK concentration at which the chaperones utilize the least amount of ATP per unit of excess free energy. the experimental observation that the fold-competent conformation is more compact with fewer exposed hydrophobic sites than the misfolding-prone conformation 4 , provides support for the assumption. Future experiments may directly test the assumption. My model makes two distinct predictions that subject it to future experimental tests and possible falsification. First, it predicts that some thermodynamically unstable substrates depend on continuous Hsp70 assistance to maintain their native structures, and such a substrate in the steady state of Hsp70-mediated folding will gradually lose its native structure upon disruption of the chaperone system. Second, it predicts that an Hsp70 molecule Top: the native and the misfolded steady state concentrations at increasing total LucDHis6 concentrations, with the DnaK concentration fixed at 1.2 μM. Bottom: the DnaK concentration required to maintain the misfolded protein concentration at or below [M] max = 0.01 µM, as well as the steady state concentration of the native substrate at that DnaK concentration. (b) The chaperones can prevent aggregation at decreasing substrate stability. I vary the protein stability by changing the rate constant of conversion, k N→F , from the N state to the F state; the corresponding change in the folding free energy ΔΔG folding is indicated on the top axis. The native and the misfolded concentrations, as well as the DnaK concentration required to prevent aggregation, are shown as in panel a. (c,d) Hsp70 is more efficient at folding substrates with slower conversion between the U and the F states. Here, I take the kinetic parameters of luciferase folding at 25 °C, and simultaneously scale the forward and reverse rates of the reaction  U F by the same factor, thus changing the kinetics without affecting the folding equilibrium. The times, t 1/2 , for the refolding of the misfolded substrate to reach half of the native fraction at equilibrium (spontaneous refolding) or the steady state (mediated by DnaK/DnaJ/GrpE), as well as the excess free energy (orange, right y-axis), are plotted against the hypothetic rates of conversion in c. The native fractions (red, left y-axis) and the excess free energy (orange, right y-axis) at the steady state are plotted against t 1/2 of spontaneous refolding in d; the equilibrium native fraction is shown as the red dotted line. (e) The time courses of Hsp70-mediated refolding of the misfolded substrate at different hypothetical rates of conversion between U and F (keeping k U→F /k F→U constant). Higher steady state native fractions are obtained at the price of longer refolding times. The optimal GrpE concentration is indicated by the red stars, and the in-plot numbers show the corresponding ratios of the nucleotide exchange rate to the ATP hydrolysis rates in the U and F states. In a and b, the rates of GrpE-catalyzed nucleotide exchange and DnaJ-catalyzed ATP hydrolysis are adjusted for the temperature of 30 °C (see Methods and Table 1). (c) Folding efficiency at different hypothetical rates of nucleotide exchange, for different values of the ATP hydrolysis rate in the U state. Native fractions (solid lines, left y-axis) are diminished at both low and high nucleotide exchange rates. At high rates of nucleotide exchange, the excess free energies (dashed lines, right y-axis) approach zero, indicating that Hsp70 can no longer drive protein folding. (d) The excess free energy as a function of the nucleotide exchange and the DnaJ-catalyzed ATP-hydrolysis rates. The rates used to model DnaK/DnaJ/GrpE-mediated folding at 30 °C are indicated by the red circle. (e) Folding efficiency increases with the DnaJ-catalyzed ATP hydrolysis rate, yielding higher native fractions (solid lines, left y-axis) and larger excessive free energies (dashed lines, right y-axis). (f) Higher ATP hydrolysis rate yields larger excess free energy (orange, right y-axis, top), at the price of higher rate of ATP consumption (red, left y-axis, top). The ratio of the two (bottom) changes only slightly. bound to a substrate molecule will dissociate faster if the substrate is in the fold-competent conformation than if it is in the misfolding-prone conformation, and that this difference will disappear in the absence of Hsp40. In support of the first prediction above, a recent experiment has demonstrated that luciferase at 37 °C can be kept active by the DnaK/DnaJ/GrpE chaperone system when there is sufficient ATP, but it rapidly loses its activity when ATP is depleted by the addition of apyrase 13 . The interpretation of this experiment, however, is complicated because apyrase also affects the luciferase activity assay. Here, based on my model, I propose an alternative experiment, in which Hsp70-mediated maintenance of luciferase activity is disrupted by inhibiting the simultaneous binding of Hsp40 to Hsp70 and to the substrate protein. For example, an isolated J-domain (e.g., DnaJ with its CTD deleted) can be used to compete against Hsp40 in binding to the Hsp70; alternatively, two D-peptides known to compete against substrate for binding to DnaJ, without binding to DnaK, can be used to inhibit DnaJ binding to the substrate 47 . When the J-domain or the D-peptide is added in excess to luciferase kept active by the DnaK/DnaJ/GrpE system, my model predicts that luciferase will lose its activity. The second prediction of my model may be tested by kinetic experiments. Luciferase can be unfolded to different extent at different concentrations of the chemical denaturant guanidinium chloride (GdmCl): luciferase adopts a more compact unfolded structure with fewer exposed hydrophobic sites at lower concentrations of the denaturant than at higher concentrations of the denaturant 4 . My model predicts that, in the presence of Hsp40 and ATP, the dissociation rate of Hsp70 from luciferase denatured by low concentrations of GdmCl will be higher than from luciferase denatured at high concentrations of GdmCl, and that this difference in the dissociate rates will disappear in the absence of Hsp40. Single molecule experiments 25,34 may provide a more stringent test of the second prediction of my model, if one can monitor both the residence time of Hsp70 on a substrate molecule and the probability that the same substrate molecule subsequently folds into the native structure. My model predicts that in the presence of Hsp40 and ATP, the folding probability will be higher if the residence time is shorter, but this correlation will vanish in the absence of Hsp40. Such experiments may be feasible if, for instance, separate fluorescence signals to detect Hsp70-substrate binding and substrate folding become available. ## Methods Model of Hsp70-mediated protein folding. I denote Hsp70 as C, Hsp40 (J protein) as J, and the NEF as E. [Y] denotes the solution concentration of the molecular species Y. There are four types of reactions explicitly considered in my model (Fig. 1b): 1) Hsp70 binding to the substrate. 2) Conformational transitions of the substrate. An Hsp70-free substrate can adopt any of the four conformational states The chaperone-bound substrate can only be in and transition between the U and F states The details of the kinetic rates of the above reactions are described below. Hsp70 binding to the substrate. , in that conformation: 0) , where is the association rate constant of the chaperone binding to a fully accessible binding site, and X = ATP, ADP. The dissociation rate constant of Hsp70 from the substrate, ⋅ k d C X , , does not depend on the sub-strate conformation, but depends on whether nucleotide X = ATP or ADP is bound. Experimentally, . Conformational transitions of the substrate. The transition rates between conformations S and S′ are different between a chaperone-free substrate ( → ′ k ) S S and a chaperone-bound substrate ( ⋅ ⋅ → ⋅ ⋅ ′ k S C X S C X ) (Fig. 1b). The condition of thermodynamic cycle closure dictates that Because Hsp40 has different affinities for different substrate conformations, the transition rates between the conformations will depend on whether the substrate is bound to Hsp40. I treat the effects of Hsp40 on the reactions implicitly by making the affected rate constants dependent on the solution Hsp40 concentration [J] (see below). For the transition ⋅ ⋅ ⋅ ⋅  U C X F C X, I assume that the bound chaperone does not hinder the substrate to go from the F state to the U state, because a binding site available in the F state is most likely also available in the U state (based on the assumption ). Thus I take It follows from thermodynamic cycle closure that the rate of the reverse transition-I use the superscript dagger to indicate that they are influenced by the presence of Hsp40-is I take the rate of (reversible) aggregation to be proportional to the substrate concentration: (for S = F, M) depend on the affinities of Hsp40 for the substrate in different conformational states. For simplicity, I assume that Hsp40 only binds to the substrate in the U state, and consequently only the Hsp40-free substrate can change from conformation U to F or M. The corresponding transition rates are where k U→S is the rate of transition U → S (for S = F, M) for an Hsp40-free substrate, p J is the probability that the substrate is Hsp40-bound, and K A J S , ( ) is the binding constant of Hsp40 for the conformational state S. Hsp40-substrate binding. To keep my model simple, I do not explicitly consider the kinetics of binding and unbinding between Hsp40 and the substrate, and make the approximation that they are always at equilibrium. The key assumption of my model is that the fold-competent conformation F is much less accessible to Hsp40 than the misfolding-prone conformation U, i.e., . To reduce the number of unknown parameters, I take ≈ K 0 A J F , ( ) , i.e., the binding of Hsp40 to the fold-competent conformation is negligible, as in our derivation of the transition rates above. I also neglect subtleties such as that the J domain may bind with different affinities to Hsp70 in the ATP-and ADP-states 48 . The above approximations may contribute to quantitative differences between the predictions of my model and the experimental observations, particularly in predicting how folding changes with Hsp40 concentrations. The binding and unbinding of Hsp40 to the substrate and to Hsp70 can be explicitly included in my model at the cost of greater complexity and additional fitting parameters, but my simplified treatment above is adequate for the key results in this work. ATP hydrolysis. The Hsp40-stimulated Hsp70 ATP hydrolysis rate, k h J ( ) , can be orders-of-magnitude higher than the unstimulated basal rate k h (0) . The ATP hydrolysis rate of Hsp70 bound to the substrate in the F state is simply , following our approximation that no Hsp40 binds to the substrate in the F state. When Hsp70 is bound to the substrate in the U state, its average rate of ATP hydrolysis, given the solution Hsp40 concentration, [J], can be approximated by where e ff , ( ) , with k a,J•C and k d,J•C being the association and dissociation rates of J domain binding to Hsp70, and [J] eff being the effective concentration of a substrate-bound Hsp40 molecule around an Hsp70 molecule bound to the same substrate molecule. The first term on the right hand side is the steady state rate of catalysis, weighted by the probability p J that an Hsp40 is bound to the substrate and thus present to catalyze the hydrolysis. Here I assume a high ATP concentration such that ATP binding to Hsp70 is fast compared to other steps in ATP hydrolysis. 49 , but this effect is not considered in my model for simplicity and lack of experimental parameters. In the absence of the nucleotide exchange factor, the rate limiting step in the reaction is the dissociation of ADP, with the rate constant k d,ADP , whereas when catalyzed by the NEF, the rate limiting step is the conformational change, with rate constant k C 29,50 . The overall rate of reaction at a given NEF concentration, [E], is then approximately , where R = 8.314 J/mol/K is the gas constant. ## NEF Solving the kinetic equations. To simplify the calculations of refolding kinetics, I make the approximation that the solution concentrations of Hsp70, Hsp40, and NEF remain constant throughout the refolding process, which is true if they are in large excess of the substrate-bound chaperone, cochaperone, and NEF. Under this approximation, refolding kinetics is described by a set of linear ordinary differential equations, which are solved by the technique of eigenvalue decomposition of the rate matrix. This simplification allows quick and robust fitting of the folding kinetic parameters to the experimental refolding data. The steady state calculations do not use this approximation.

chemsum

{"title": "Cochaperones enable Hsp70 to use ATP energy to stabilize native proteins out of the folding equilibrium", "journal": "Scientific Reports - Nature"}

the_missing_label_problem:_addressing_false_assumptions_improves_ligand-based_virtual_screening

## Abstract: Ligand-based virtual screening (LBVS) uses machine readable representations of chemicals to learn a mapping function that can predict binding interactions with protein labels. Because it is highly scalable it is increasingly used in drug development in academic and pharmaceutical contexts. We have identified assumptions commonly used in LBVS that are false, which collectively can be described as the missing label problem. Firstly, many of the binding interactions in the bioactivity databases typically used to train LBVS models have never been tested before, but the absence of a label is interpreted by most models as a true negative. Secondly, many proteins have multiple binding sites with unrelated shapes but the associated ligands are grouped together under the one protein label. These assumptions frustrate the ability of the model to learn a correct mapping function. Here we use statistical techniques to predict values for the missing labels and binding sites and show how this improves the ability of LBVS models to rank ligands correctly. In the process we introduce a new technique for removing bias during model evaluation based on data blocking from experimental design theory. All data and code for analysis and generating figures is publicly available on github (https://github.com/ljmartin/Missing_label_problem). ## Introduction Screening of ligands against protein targets is an essential part of drug discovery, target identification, and toxicology. Virtual screening (VS) seeks to predict protein targets using computer modelling, which is both faster and cheaper than in vitro approaches, and is thus increasingly integrated in drug discovery and development 1 . VS approaches can be split into two groups: structure-based VS, which requires a crystal structure to which ligands are fit using a scoring function, i.e. docking; and ligand-based VS (LBVS), which uses similarity to known active drugs to determine possible targets 2 . LBVS often uses molecular fingerprints as machine-readable descriptors. A de facto standard molecular fingerprint is the substructure fingerprint, which uses binary categorical variables -i.e. the presence or absence of substructures -to create a vector representation using multi-hot encoding. Similarly, the protein targets bound by a ligand, here referred to as the 'labels', can be considered as vectors where each entry is a binary variable that indicates the presence or absence of a binding interaction. Thus, common LBVS paradigms use statistical techniques called machine learning (such as regression, neural networks, or decision trees) to learn a function that maps the ligand vectors to the label vectors 3 . This process is sometimes called classification. In the prediction stage, untested molecules are first transformed into a molecular fingerprint ligand vector and are then fed through the mapping function to generate the predicted label vector. We have identified some missing label assumptions commonly used in LBVS that are simple to demonstrate as false. Firstly, treating untested interactions in the label vectors as true negatives learns a potentially incorrect mapping from ligand vector to label vector, as in fact the true nature of these interactions are unknown. Secondly, the grouping of ligands that bind different sites on a single protein under the one label is inconsistent with the similar property principle underlying LBVS, and ignores the fact that many proteins have multiple, discrete binding sites with different shapes. We show that by addressing these assumptions we can substantially improve LBVS performance The LBVS task is closely analogous to 'multi-label learning' and could benefit from some of the approaches used in that field, in particular the use of different scoring functions and addressing missing labels. Multi-label learning is so-called because each instance of the data is associated with multiple labels simultaneously 4 . This contrasts with multi-class learning where each instance is associated with a single label from multiple, mutually exclusive, labels, and with single-label learning where only one label is learned. An extension to multi-label learning, extreme multi-label learning, learns hundreds or thousands of potential labels compared to only tens of labels typically considered in multi-class problems 5 . An example of a multi-label learning problem is scene classification -amongst thousands of potential labels, a given image could be mapped to all of beaches, sunsets, parties and more 6 . Similarly amongst thousands of protein labels any ligand can potentially bind to multiple proteins; for example imatinib, which binds the therapeutic fusion-protein target BCR-ABL but also the off-target C-Abl 7 . One issue shared by these tasks is how best to evaluate predictive performance during the testing stage. Commonly used metrics like precision and recall measure the ability to retrieve the known, true labels. Most training examples, however, are incompletely labelled due to the time-consuming and expensive process of testing every possible label in advance. Thus, in multi-label settings even perfectly valid, newly predicted labels can be scored as false positives if the label was previously unknown. An alternative comes from the multi-label learning literature, in which models are evaluated only by the ability to rank known positives before unknown labels 8 . Ranking can be considered as a generalization of multi-class classification for the multi-label case 9 . A common ranking metric, the ranking loss, rewards prediction of both known positives and unknown positives as long as the known positives come first 4 . This is well-suited to drug discovery or target identification projects, in which a ranked list of predictions is often tested using in vitro assays until some cut-off point, where the predictions are no longer informative. Practically, directly optimizing for ranking thus reduces expense and time in drug development, while theoretically it acknowledges the presence of missing labels. In the training phase, many machine learning algorithms assume the label vectors used as training data are complete and true 10 . As an example, if a ligand is active at a protein label then that position in the label vector is a 1. Conversely if the ligand-pair has not been tested then that position is a 0, which is treated as an explicit, known negative. In reality, the vast majority of label vectors consist of known positives and unknown positives/negatives 6 . The consequence is that, assuming we admit that bioactivity data used for training is incomplete, all LBVS models are learning an incorrect mapping from ligand vector to label vector. The gold standard solution would be to perform binding assays on each individual protein-ligand pair used in training, but this is infeasible at present -a recent count of the data in ChEMBL has 3569 human proteins and ~1.8 million distinct compounds, implying ~6.5 billion possible interaction points 11 . An alternative is to fill in the most likely missing points in the label vectors by treating the set of vectors as a network graph and using co-occurrence trends. This changes the multi-label learning task into a link prediction problem, which has been applied to tasks such as search engine and social network classification 12,13 but, to our knowledge, has not been used to solve the missing label problem in LBVS. Assuming the predicted labels are consistent with the ground truth, i.e. they really come from the same population as the true positive ligands, these new labels should help the models to correctly rank test ligands, providing a simple evaluation of the benefit of this approach. Conversely, if the newly predicted labels are spurious then the ranking loss should become worse. As well as improving ranking for test ligands, filling in the missing labels has the additional benefit of predicting new labels even before training a model and can thus be used as a classifier itself. A related labelling problem that has not yet been addressed in the LBVS literature is the grouping of multiple binding sites for a particular target under the single label for that protein target. In the pharmacology literature the existence of multiple binding sites on a single protein has long been recognised 14 , but this has not yet been acknowledged in LBVS. Extreme examples of proteins with multiple sites are the ligand gated ion channel proteins, which can have as many as 15 or more binding sites 15 . Recently, in silico and in vitro fragment screening results have indicated that GPCRs 16 and enzymes 17 also have multiple binding sites, giving experimental evidence to the suggestion that multiple binding sites at individual proteins is the norm across all dynamic proteins, including the major drug target-types, rather than an exception 18 . Furthermore, even within one single binding site, some ligands with unrelated structures can bind by accessing different conformations of the site known as binding modes 19 . This grouping of binding sites under a single target label violates the similar property principle underlying LBVS, which posits that ligands with the same label have similar structures 20 . Analogously to missing labels, violating this principle is expected to reduce the ability of statistical models to learn the structural determinants of binding by forcing them to learn commonalities between chemical substructures that, in reality, have no similarity. Determining the true binding site for a ligand usually requires either mutation or crystallization, an unscaleable process that cannot realistically be applied to the number of ligand-target interactions in major bioactivity databases. We propose an in silico alternative that uses clustering to group ligands active at a particular target into structurally related subsets. We show that this approach successfully recognises multiple binding sites, recognises scaffold hops, and labels the majority of ligands with their correct site using an example dataset (a GABA A receptor from ChEMBL) where the binding sites for the major classes of active ligands are known and well characterized. Importantly, applying the clustering algorithm to a larger dataset from the ChEMBL bioactivity databse improves the ability of LBVS models to rank ligands. The two goals of our work -predicting missing labels and introducing new labels for binding sites -may seem to abstract the label vectors from the ground truth, making evaluation of the new label vectors difficult. To remedy this, we sought an evaluation method that is robust to possible memorization bias, also known as testtrain leakage. Recent LBVS literature has rightly pointed out that randomly selecting ligands from the training set to create a test set for evaluation can result in overly optimistic performance estimates that do not align with prospective validation . This occurs because the ligands in most bioactivity datasets are not independent and identically distributed -the discovery of one active ligand often leads to many highly similar structural analogues with only a few changed atoms 24 , suggesting that the number of independent data points in most LBVS datasets is far fewer than the actual number of ligands. Predicting the activity of these same-scaffold ligands is trivial for machine learning techniques that are able to memorise the ligand vectors used in training, but it does not generalize to so-called 'out-of-sample' data that have unseen scaffolds. A more realistic evaluation uses time-split cross validation -that is, removing blocks of ligands from a contiguous time period to be the test set . This approach likely originated in financial time-series analysis where it is known as backtesting 28 . Time-split cross validation shows a reduction in memorisation bias but is still susceptible to memorization since, unlike financial time series, molecular bioactivity datasets are non-linear and have uneven distributions of scaffolds after the discovery date as well as the possibility of train-test splits with zero test ligands 27 . We propose a more robust method to avoid bias by randomly choosing test ligands and explicitly removing all highly-correlated, non-independent, structural analogues from each possible training set. This method inherits from blocked experimental design, and we suggest that all LBVS practitioners should be mindful of this statistical background due to the high correlation inherent in bioactivity datasets. The goal of block design is to group the experimental units into "blocks" that are as uniform as possible to avoid measurement of effects orthogonal to the conditions of interest 29,30 . To evenly sample all ligands and account for the non-linearity of ligand sets, we repeat the test set evaluation hundreds of times until the performance metric stabilises in a process inspired by bootstrapping. Removing similar ligands from the training set by a distance metric has been proposed beforesuch as asymmetric validation embedding (AVE) 21 and maximum unbiased validation (MUV) 31 -but these functions search for the global minimum of a bias function rather than rotating all ligands through the test set. As a result the evaluation may be less robust for assessment of out-of-sample data, which is often the intended target of LBVS. Our approach is motivated by an empirically-derived quantitative definition of correlated structural analogues used as a distance cut-off. It removes a single group of highly similar analogues in each case, akin to leave-oneout sampling 32 but instead leaving out blocks of correlated data. It thus eventually samples all scaffolds while maximising the available training data to more closely approximate out-of-sample data and generalizability estimates for prospective LBVS. This leads to a robust evaluation metric for evaluating manipulations of the label vectors. In this research article we show improvements to the de facto standard method of LBVS using a common experimental configuration -molecular substructure-based ligand vectors, protein label vectors from the ChEMBL bioactivity database, and a naïve Bayes classifier for predicting label vectors. In Part 1 we introduce the bootstrapping procedure for robust, bias-free evaluation of model performance, then we use this to show how filling in missing labels (Part 2) and clustering ligands into binding sites (Part 3) both substantially improve ranking. ## The data The dataset used in the following experiments is made up of the ligands that bind to 243 proteins from the ChEMBL bioactivity database, which is an online database that automatically annotates bioactivity data published in several medicinal chemistry journals 11 . The proteins cover a range of protein families, are all 'single protein' records (which accords with common practice in the field), and each has at least 500 but no more than 5500 associated ligands bringing the total to 252,409 ligands. A schematic in Figure 1 describes the LBVS process used here. Morgan molecular substructure fingerprints generated by the rdkit python library 33 are fed into a trained naïve Bayes classifier from the sklearn python library 34 , which then predicts a label vector corresponding to the protein targets for that ligand. In each case, the classifier is trained using the pre-labelled dataset from ChEMBL described above. Rather than run a large number of expensive in vitro assays to evaluate the predictive performance, portions of the dataset are masked during training to use as a test set, as described in Part 1. where each non-zero entry corresponds to a binding interaction with a target protein. A machine learning model learns a mapping from ligand vector to label vector, which can then be applied to unseen ligand vectors to predict new label vectors. ## Part 1. Evaluation using ranking loss and block bootstrapping Evaluation of predictive models generally uses k-fold cross-validation, in which the dataset is separated into k blocks that are, in turn, each used once as a test set by being separated from the training data and then evaluated on prediction metrics (Figure 2a). When some ligands are highly correlated, such as for closely-related molecules resulting from structure-activity relationship analyses 24 , this can lead to highly correlated data in both the training and test sets and thus overly optimistic performance metrics. An alternative that respects the nonlinearity and non-independence of molecular bioactivity datasets is a blocked design in addition to bootstrapping with replacement, in which test sets are repeatedly removed at random until the performance metric converges (Figure 2a). To reduce the bias from highly correlated ligands, in every repeat the test set is made by first choosing a single ligand at random then, in addition to that ligand, selecting the k nearest neighbours up to some distance cut-off. We use the Dice distance as the distance metric between ligands, since it has a wider spread of pairwise distances as compared to some other commonly used metrics (see Figure S1). ## Figure 2. Block bootstrapping can evaluate model performance while also removing bias due to test/train leakage of highlycorrelated data points. a) A common machine learning approach, k-fold cross validation, splits the data randomly such that each data point is in the test fold only once, but assumes that the data is independent and identically-distributed. Backtesting removes all data from a contiguous timeframe, but requires linearly separable data. For molecular data, these assumption are not true. Our bootstrapping-inspired approach repeats the test-train splits multiple times, ensuring that in every split a central ligand and all highly correlated neighbours are removed as the test fold, minimizing inclusion of structural analogues across both the test and train sets (test/train leakage). b) The median ranking loss of 20 randomly-sampled targets at different cut-off values using the bootstrapping sampler. Increasing the cut-off distance used to define the nearest neighbours removes more data from the training set, in turn reducing ability of the classifier to rank correctly. c) Histogram of pairwise Dice distances from the 'Protein kinase C theta' label, showing characteristic bimodal distribution (blue), fit using a Gaussian mixture model with two components (red, smooth). Also shown are the positions of the means of each Gaussian component (red, dashed). d) The means of the highly correlated (red) or different structure (blue) Gaussian components from all 243 protein labels along with the proposed cut-off points (orange), which are defined as the midpoint between the two components for each target. e) Ranking loss from block bootstrapping evaluation of 10 example targets at cut-off d=0.525, which converges at higher number of repeats (coloured lines). Early-stopping criterion reduces computation by stopping when convergence has been reached (coloured points). Increasing the distance cut-off leads to the removal of more nearest-neighbours, reducing the amount of information available in the training set to train the classifier (Figure 2b). Our goal was to identify the optimal cut-off distance that groups highly-correlated analogues, i.e. dependent data, together while maintaining as much independent data in the training set as possible. To our knowledge all previous methods use an arbitrary cut-off or rely on potentially biased, human-defined, definitions of scaffolds such as Murcko scaffolds. We prefer an empirical approach, starting from the intuition that any given pair of molecules are either a pair of analogues or are independent. To find the cut-off that splits these two types of pairs, we fit the pairwise distance distributions from the ligands in each label using a Gaussian mixture model with n=2, reflecting the two possible states for any given pair (analogue or independent), using the a priori knowledge that the distributions are made up of both the analogue pair distances and independent pair distances. Consistent with this, pairwise distance distributions of ligand sets have a characteristic bi-modal shape, one example of which is shown in Figure 2c. Fitting the Gaussian mixture model to all 243 protein targets, and recording the Gaussian means, estimates the mean of the analogue and independent pairwise distances. The Gaussian centres show a clear separation between the two populations at lower and higher Dice distance (Figure 2d). Setting the cut-off to the value in-between the lower and higher distributions should provide the optimal cutoff point to remove structural analogue pairs from every training set. Averaging over the means of the Gaussian fits, and setting the middle value as the cut-off, led to a 95% confidence interval of separating Dice distance of 0.520 to 0.536 (Figure 2d). All of the following evaluations use Dice distance d = 0.525. Visual inspection of 6 ligand pairs, randomly chosen from 6 random targets. with d < 0.525 shows that all have large maximum common substructures, while pairs with d > 0.525 selected in the same way show more diverse structures (see Figure S2 for the comparison of ligand pairs). This cut-off will thus strike a good balance between removal of structural analogue pairs and retaining differently structured pairs, which fulfils our goal of identifying an optimal Dice distance cut-off. Compared to k-fold cross validation the bootstrapping procedure proposed here incurs higher computational costs. This is one of the reasons for using naïve Bayes classifiers, which are highly scalable and thus have short model building times (other reasons are that these classifiers are ideally suited for one-hot encoded binary data 35 ). Bootstrap metrics converge on their expected value in the limit of the numbers of repeats 36 . So, to reduce calculation times, we implemented an early stopping criterion that prevents unnecessary repeats after reasonable convergence, which is defined as the maximum percentage change in median ranking loss over the previous 50 trials of lower than 2.5%. This procedure allows more repeats for labels with higher variance, but shorter calculation time for labels with fast convergence (Figure 2c). The code used to perform the bootstrapping experiments is available as a Python module called the 'VirtualScreeningBootstrapper', which takes as input a ligand vectors matrix, a label vectors matrix, and an sklearn classifier. Part 2: Filling in missing labels improves performance Training most machine learning models requires the input of label vectors that describe the true positive and true negative labels. When the training data is incomplete due to high cost of determining the labels, this process leads to learning an incorrect mapping of the input ligand vectors to the label vectors. A possible solution is using co-occurrence trends in the matrix made up of all the label vectors to predict the missing labels. In this process, the matrix has columns representing each protein target and rows representing the label vectors for each ligand. When a ligand co-occurs at two proteins or more (that is, it has more than one label) it indicates the binding sites at the two proteins have some degree of similarity. Iterating over all ligands with 2 or more labels leads to a label correlation value, calculated as the percentage of ligands for each protein that co-occur at a second label. All proteins in this dataset have at least 500 ligands, which avoids high correlation values occurring by chance from a few shared ligands. The pairwise label correlations are shown as a heatmap in Figure 3a. Clearly some proteins have highly similar binding sites, with correlation values above 80%. A random sample of highly correlated protein pairs is given in Figure 3b as an example. Judging only by the names, these pairs align well with a phylogenetic understanding of the protein labels -the most highly correlated pairs are subtypes that split from a common gene, likely having similar protein sequence and thus similar binding sites. This suggests that the ligands with only one label from a correlated pair are likely to also bind to the second protein, but that this interaction either has simply not yet been tested or is not recognised by the automated annotation process used by ChEMBL. In order to use the correlations between proteins to fill in missing labels, we assumed each protein correlation is independent and calculated a per-protein binding probability for each ligand (see methods). For each ligand with at least two labels, we individually removed that ligands' influence on the correlation matrix and then calculated the probability that each label would be predicted using the label correlation data. Comparing the probability of the known label to the probabilities of the unknown labels in that vector gives the ranking loss. Ligands with only one protein label cannot be used for this evaluation because removing the only label means there are no labels left to calculate a probability. The ranking losses in Figure 3c show a distribution with a mean ranking loss of ~0.012 and median of ~0.002, which is approaching perfect ranking. Using the calculated probability values, we filled in the missing labels using different thresholds, which at lower thresholds predicted several hundred thousand new labels (see Figure S3). To determine if the new labels predicted by this thresholding approach are consistent with the existing labels, we evaluated the label matrices using block bootstrapping, removing only the known ligands as the test set but including the newly predicted ligands as positives in the training set. If the labels predicted by the labelcorrelation-approach are indeed true positives, then their presence in the training set should help the classifier to rank the test set, reducing the overall ranking loss. The relative ranking losses using predicted labels at multiple probability thresholds are shown in Figure 3e, where density to the left of the dotted line indicates improved performance as judged by lower ranking loss. Even at the highest threshold of 90% (which has the fewest newly predicted labels) most, but not all, protein labels perform substantially better when including the newly predicted labels. At lower probability thresholds, the majority of labels still perform better but the spread is increased. This has a remarkable implication: At the lowest threshold of 2%, the label matrix has an approximately four-fold increase in the number of labels, equivalent to 1,142,376 new labels. Most labels perform better when including the newly predicted labels, suggesting most of the predictions are true. The scale of this level of predictions dwarfs the number of predictions ordinarily considered in virtual screening, all while subsequently improving performance of the classifiers. Importantly, the technique is highly adjustable depending on the practitioner's level of acceptable confidence, and a middle ground that shows a good balance between a conservative number of newly predicted labels and improved performance is a threshold of 20%. Ultimately, filling in missing labels using the correlation graph as probabilities, with a 20% threshold, shows an improvement in ranking loss of ~20% averaged over all targets in this dataset. ## Part 3: Clustering into binding sites improves performance The lack of knowledge about protein binding sites can be considered another type of missing label problem, which reduces performance by grouping ligands from different binding sites under the same label. Because there is no expectation that the binding sites on a protein have the same structure (excluding the identical sites on homomeric proteins), by the similar property principle one can deduce that the grouped ligands also have unrelated structure. In order to separate grouped ligands into their binding sites without a priori knowledge from mutation or crystallization evidence, we first use clustering and then merge clusters based on their relative similarity to two distributions of Dice distances. The distributions, shown in Figure 4a, are the pairwise Dice distances from ligand pairs within a single label or between different labels. This serves as a proxy for taking Dice distances from ligand pairs within a single binding site or between different binding sites. To separate ligands into binding sites, first the ligands from within a label are clustered using agglomerative clustering. This generates several proposed binding site groups, which can be merged if they resemble the samelabel distribution more than the different-label distribution. We used the Kolmogorov-Smirnov statistic to determine the distribution similarity. Applying this process to a label from ChEMBL with multiple known binding sites demonstrates the technique's potential (Figure 4b). The protein label used is CHEMBL2093872, which corresponds to a protein complex group of GABA A receptors with 365 active ligands including neurosteroids, insecticide pore-blockers, benzodiazepine or non-benzodiazepine positive allosteric modulators, and orthosteric site ligands such as GABA. Importantly, these different ligand classes have different and well characterised binding sites at GABA A receptors. Our combined clustering and merging technique accurately groups two different scaffolds of pore-blockers at a single binding site (demonstrating it is not simply clustering based on a single scaffold), as well as recognising the neurosteroid site binders. The remaining two clusters largely correspond to either the extracellular allosteric site, which includes multiple scaffolds like diazepam and βcarbolines, or the orthosteric site, which includes multiple scaffolds such as the flavones and muscimol. The clustering isn't quite perfect, with the orthosteric ligands pitrazepine and GABA being grouped in the extracellular allosteric site, but it is a substantial improvement on the original single label. This suggests the technique can be used to improve the label vectors used in LBVS. ## The clustering and merging technique was applied to the whole dataset and split 29 targets (approximately 12% of the number of targets) into either two sites or three sites each. The ranking loss of the new target labels is compared to the original, unclustered labels in Figure 4c, where density to the left of the dotted line indicates improved ranking (lower ranking loss). To compare the performance of ligand sets that are split into multiple labels (one for each proposed binding site) to that using the single original label, the ranking loss of all clusters was calculated separately and then combined using a weighted average with the weights determined by the number of ligands relative to the single original label. All of the clustered ligand sets perform better than their un-clustered counterpart. This represents a substantial improvement on its own and, as with filling in missing labels, is highly adjustable in that practitioners can choose which targets to keep clustered based on the relative performance. In addition to this, the clustering approach gives access to large numbers of new labels in ChEMBL: Currently the most commonly used protein labels are the 'single protein' records because it is assumed all the ligands will be structurally related since they bind to a single protein but, with clustering into binding sites, other protein type records such as 'protein complex', 'protein complex group', 'chimeric protein', 'protein family', and 'protein-protein interaction' can also be used, increasing the amount of training data and thus the chances of determining the correct label for novel test ligands. Ultimately, filling in missing binding site labels leads to a mean improvement in the ranking loss of approximately 30% for proteins with multiple binding sites. ## Discussion Many LBVS approaches use assumptions that may have arisen out of necessity or simplicity but limit performance. In particular: • The use of incomplete label vectors to train machine learning models, which treat unknown interactions as true negatives when some are actually positive • Grouping different binding sites under the same label, which forces models to learn similarities between unrelated chemical substructures Collectively, these two assumptions can be viewed as missing label problems. Ultimately, they stem from a single assumption: that the bioactivity data used for training is correct and complete. If this were true, then no more interactions between proteins and ligands present in ChEMBL need be tested. In reality both assumptions are false in many cases. Continued updates of the ChEMBL database is an implicit recognition of this -here we make this recognition explicit, addressing the missing label problem using a more complex but more correct approach to LBVS. As we showed, this improves the performance of the trained models, which suggests that the newly predicted, formerly missing, labels are accurate. The common assumption of independent and identically distributed data has previously been recognised as problematic . So, to show improvements in addressing missing labels, here we also developed a new method of model evaluation called block bootstrapping that is more computationally intensive but explicitly avoids biased evaluation by removing structural analogues from each and every test/train split. The key to addressing the simplifying assumptions in LBVS is borrowing techniques from other fields. As a precedent, the proposal for time-splits to replace random-splits for cross validation may have been inspired by financial time-series analysis where it is known as backtesting 28 , although we aren't aware of an explicit reference to that field. Our proposal for block bootstrapping improves upon backtesting by handling the nonlinearity of ligand bioactivity datasets at the cost of increased computation. While the exact cut-off used to separate highly-correlated pairs from independent pairs will differ by the choice of molecular fingerprint, the use of Gaussian mixture models to determine the cut-off is applicable to any dataset. A previous approach to de-biasing, the AVE bias, initially uses randomly selected training and test sets followed by a genetic algorithm to reduce bias between the sets 21 . This technique appears to effectively reduce bias, but the global minimum of the bias function may be unique and so, by definition, does not including all scaffolds in the test set over the course of evaluation. Our method, using a blocked design, is akin to leave-one-out sampling but instead each test set leaves out a block of highly-correlated analogues. After enough repeats, it thus selects all scaffolds to be in the test set and converges on a performance value that uses all of the available data. The multi-label learning field has addressed the shift to ranking as a generalization of multi-class learning, which in contrast commonly uses precision and recall as metrics 9 . When some labels are unknown rather than true negatives, as in LBVS, precision penalizes the prediction of new labels that may be true. At the same time, recall may be optimistic because there are unknown positives that aren't counted. An extension of precision to the multi-label case is precision at k (p@k) . For search engines, as an example, this metric measures the proportion of relevant, i.e. true positive, results returned in the top k results, where k refers to the number of items an average user might reasonably browse 8 . For LBVS, the value of k is impossible to define without biasing against highly selective ligands (using large k) or in favour of highly non-selective ligands (using small k). To remedy this we used ranking loss, which is equally applicable to both selective and non-selective ligands as well as forgiving when models predict unknown positives as true positives -as long as they are ranked below the known positives. The approach to filling in missing labels using the label correlation graph was inspired by Tan et al. 40 , which is also an extreme multi-label learning setting based on the number of labels. In that work, when items have multiple labels the probability score is the sum of the correlations with each other label. Clearly this can lead to probabilities above 1, which we found unintuitive. Instead, we assumed that each correlation is independent of the others meaning the correlations can be multiplied to generate a final probability. This is akin to the 'naïve' independence assumption used in a naïve Bayes classifier and, while not totally correct, appears also to be an effective assumption 41 . A possible improvement on this technique is to take dependence relations into account when calculating probabilities. Since clustering into binding sites is a highly task-specific problem, we found no analogy in other fields. Nevertheless, the use of distributions of Dice distances between ligands to merge pairs of clusters is reminiscent of the pioneering similarity ensemble approach (SEA) to LBVS 42 . In SEA, the distance scores of two groups of ligands were fit using an extreme value distribution to generate expectation values. In comparison, our approach uses a non-parametric fit. The good performance on a GABA A receptor dataset as well as improved ranking loss and simple implementation recommends this approach. Finally, we comment on the multi-label setting. This work solely uses 'binary relevance', which is a multi-label learning term meaning that a single classifier is fit for every label 43 . This is a convenient problem setting that allows for our block bootstrapping technique, but more advanced problem settings are possible. In particular, random forests, neural networks, and various problem transformation techniques can fit multiple labels at once to take advantage of the similarities between labels and improve performance 43 . Binary relevance was necessarily used here to demonstrate the benefit of addressing missing labels, since the block bootstrapping technique is not applicable to other multi-label problem transformations. Nevertheless, improvements to the label matrix should be classifier-and problem transformation-agnostic and benefit all multi-label learning approaches. Here we have shown how several simplifying assumptions used in LBVS are false. We provide solutions that address these assumptions and substantially improve the predictive performance of LBVS machine learning models. To do this we developed a new method for evaluating LBVS models based on bootstrapping that is guaranteed to avoid scaffold memorization since it explicitly removes similar structural analogues from every training set. Using this method, we showed how filling in missing labels in the label vectors not only corrects the unsatisfying situation where statistical models are built using knowingly incomplete label vectors that are falsely assumed to be complete, but also improves the ultimate performance of those models. Similarly, the labels of proteins with multiple binding sites have been improved by clustering the ligands into their binding sites, removing the spurious grouping of unrelated ligands under the one label. ## Methods All data and analysis scripts, along with IPython notebooks describing the process, are available at https://github.com/ljmartin/Missing_label_problem Data Protein ligand interaction data was downloaded from ChEMBL24 11 . All proteins are 'single protein' records, with activity at a protein defined as a pchembl value greater than 5, the equivalent of <10μM. Proteins were kept based on having greater than 500 but less than 5500 active ligands. Ligands were stored and manipulated in python by converting the SMILES strings into molecule objects from the python rdkit library 33 . The ligand set was sanitized by removing ions and ligands with molecular mass less than 90amu or greater than 80. All ligands were featurized as Morgan fingerprints with radius=2 and folded to 256-bit binary vectors using the GetMorganFingerprintAsBitVect method in the rdkit library. These formed the instance vectors used for model training. The label vectors were generated using the MultiLabelBinarizer method available in the sklearn python library 34 . ## Model training Predictive modelling of protein/ligand interactions was performed using a Bernoulli naïve Bayes classifier as implemented by the sklearn python library 34 , using default settings i.e. Laplace smoothing parameter of 1. This type of classifier is designed for single-label classification tasks, while the label matrix is inherently multi-label. In order to use a naïve Bayes classifier, model training was performed using the 'binary relevance' problem transformation technique, which splits the labels into multiple single label learning tasks. As a result, each protein target is evaluated individually, and separately from the others. The full ligand set was used as available testing and training data for each single-label learning task. ## Model evaluation by block bootstrapping Model evaluation was performed by calculating a performance metric on test sets made up of ligands that were masked from the model during fitting. After fitting, the model was used to generate predicted probabilities for a label for each of the masked ligands, which were then compared to the ground truth labels for evaluation. The performance metric was label ranking loss as implemented by the sklearn library 34 . Due to the binary relevance problem transformation, the ranking loss was calculated using the ranked probabilities of the test ligands binding to a single protein target, thus measuring the ability of the model to rank true positives before unknowns for each single target. Evaluation used our block bootstrapping approach, which evaluates the model on multiple training and test sets until the median of the performance metric measurements converges. In each iteration of this process the test positives are selected by first choosing a true positive ligand at random and then masking that ligand as well as all of its nearest neighbours up to some cut-off. Nearest-neighbours were defined by the Dice distance between the molecular fingerprints. For each iteration, the test negatives were chosen by randomly selecting 10% of the ligands that are not true positive. The error of a bootstrap measure converges to zero in the limit 36 . To save computation time, satisfactory convergence was defined by thresholding the maximum change in the median of the performance metric over several previous trials. A threshold of 2.5% was used so that when the performance metric had changed by less than 2.5% in either direction over the previous 50 trials, no further iterations were performed. ## Gaussian mixture model fitting A Gaussian mixture model was fit to the set of pairwise distances of the true positive ligands associated with each target to determine the most likely same-scaffold and different-scaffold distances. Pairwise distances were calculated using ligand vectors and the pairwise_distances method available in the sklearn Python library 34 , with metric='Dice'. Gaussian mixture models were fit using all pairwise distances, except for the distances to self, using the GaussianMixture method in sklearn, with n_components=2. One mixture model was fit for each target, and the means of the Gaussians were recorded with the lower-valued mean attributed to same-scaffold pairs and the larger-valued mean attributed to different-scaffold pairs. The cut-off separating the means was determined as the value at the midpoint between the two Gaussian means for each target. The confidence interval of the cut-off was calculated using bootstrapping, taking 10,000 samples of the set of cut-offs at random with replacement. The 95% CI is calculated by recording the 250 th and 9750 th values of the ranked means of these bootstrapped samples. ## Correlation graph Percentage correlations between each pair of targets was calculated using the set of label vectors with 2 or more labels (label vectors with a single entry don't offer any information on the correlation between any pair of labels). The correlation matrix measures the percentage of ligands of each target that are shared with another target. The percentages in this correlation matrix are then used as probabilities to determine new labels. The probably 𝑝𝑝 𝑖𝑖 of an unknown label for the ith protein target being a positive is calculated as: where 𝑛𝑛 = the number of true positive labels present in the label vector being considered, and 𝑐𝑐 𝑖𝑖𝑖𝑖 = the percentage correlation of the ith target with the jth target. In more simple terms and using the rolling of dice as an analogy, this is equivalent to asking "given n dice rolls, what is the probability that any one of them is a six?", which is equal to one minus the probability that none of them are six. Values in the label matrix were set to '1', indicating a true positive, if the probability score was greater than some threshold value. All existing ground truth labels were retained. The new label matrices, using threshold 𝑝𝑝 𝑖𝑖 values of 0.02, 0.05, 0.2, 0.5, or 0.9 were then assessed for their effect on ranking using the block bootstrapping approach. To maintain relevance to the ground truth, the test set true positives were chosen only from the original, ground truth, while test set negatives were chosen only from the remaining labels that are not true positives. This means the newly predicted labels are always present in the training set. If the newly predicted ligands have ligand vectors that are consistent with the true positives in the test set, then they should improve ranking of the ground truth true positives, and vice versa. To measure the effect of including the newly predicted labels on ranking, for each target the log10 of the relative ranking loss, compared to the original label vectors, is reported. ## Clustering into binding sites Ligand sets for each target were clustered in order to group them into candidate binding sites. These clusters were subsequently merged based on statistical analysis. Clustering used the AgglomerativeClustering method in the sklearn library, with a precomputed distance matrix recording the Dice distances as input, n_clusters=None (because the number of binding sites is a priori unknown), affinity='precomputed', linkage='average' and distance_threshold=0.8. The distance threshold hyperparameter has not been optimized, but was chosen to slightly overcluster, generating a number of candidate binding sites that can then be subsequently merged. Merging the clusters involved comparison to two distributions of Dice distances. One distribution was made up of pairwise distances within single target labels, representing Dice distances most likely to be from a single binding site, and another distribution was made up of pairwise distances from between different target labels, representing Dice distances most likely to be from different binding sites. These distributions were generated by 30 repeats of selecting two targets at random, and calculating pairwise Dice distances from ligands within or between the two target labels. Candidate binding site clusters generated by agglomerative clustering were then compared to these two distributions by calculating a Kolmogorov-Smirnov (KS) statistic using the ks_2samp method available in the scipy python library 44 . The KS statistic is a measure of the distance between two distributions by comparing their cumulative distribution functions. It was preferred to other non-parametric techniques because it captures the influence of distribution shape, and the same-label distribution has a characteristically different shape to the different-label distribution (refer to Figure 4A). Thus, it can measure whether the given two clusters have greater likeness to the same-binding site distribution or the different-binding site distribution. The pair of clusters that have greatest likeness to the same-binding site distribution are merged until all possible pairs of clusters resemble the different-binding site distribution or until there is only a single cluster left.

chemsum

{"title": "The missing label problem: Addressing false assumptions improves ligand-based virtual screening", "journal": "ChemRxiv"}

radical_cations_and_dications_of_bis[1]benzothieno[1,4]thiazine_isomers

## Abstract: Bis[1]benzothieno [1,4]thiazines (BBTT) are particularly electron-rich S,N-heteropentacenes and their radical cations and dications can be relevant intermediates in charge transport materials. All three regioisomers of N-p-fluorophenyl-BBTT (syn-syn, syn-anti, and anti-anti) were studied. A reliable preparation of radical cations and dications using antimony pentachloride as an oxidant gives deeply colored salts.The electronic structure of the radical cations was assessed by EPR spectroscopy, whereas dicationic structures were characterized by NMR spectroscopy. In addition, a deeper insight into the electronic structure was experimentally and computationally obtained by UV/Vis spectroscopy as well as UV/ Vis spectroelectrochemistry and DFT and TDDFT calculations. BBTTs can be considered as highly polarizable donor π-systems with significant charge transfer contributions in neutral, cationic and dicationic state.Scheme 1 Regioisomeric BBTTs resulting by formal thieno expansion of phenothiazine. † Electronic supplementary information (ESI) available. See ## Introduction Molecular electronics based upon organic molecules has become increasingly important over the past decade and the quest for tailor-made functional small molecules is an ongoing task for chemistry. 1,2 Since the efficiency of electronic devices correlates with electronic properties of the constituting molecules, a rational design of novel molecule based semiconductors commences with high lying HOMO levels, which in turn correspond to low oxidation potentials. Electron-rich heterocycles, such as phenothiazine, are prominent structural motifs used in electroactive materials for batteries, OLEDs, and DSSCs. They possess low oxidation potentials with fully chemically reversible one-electron oxidations, low band gaps, and luminescent properties. In addition to their convenient accessibility, their intrinsic properties are easily fine tunable by synthetic functionalization. 22,23 Furthermore, the remarkable stability of their radical cations is crucial for providing efficient charge transfer as hole conductors. 24,25 Recently, we successfully proposed the two-fold thieno expansion of phenothiazine to bis benzothieno thiazines (BBTT) that can be considered to be novel electron-enriched phenothiazine congeners (Scheme 1). 26 The resulting compounds likewise possess discretely adjustable properties that depend on the mode of anellation, as for dithienothiazines, 27 as well as on the substitution pattern. Different steric hindrance unequivocally causes the absence of measurable luminescence in syn-syn BBTTs (bis benzothieno[2,3-b:3′,2′-e] thiazines), whereas inverse anellation of benzo [b]thiophene units leads to luminescence for syn-anti BBTTs (bis benzothieno[2,3-b:2′,3′-e] thiazines) peaking for anti-anti BBTTs (bis benzo-thieno[3,2-b:2′,3′-e] thiazines) with fluorescence quantum yields drastically increased to 45%. Furthermore, anti-anti BBTTs represent the first known fused thiazines with antiaromatic character in the crystalline solid state. Likewise, being identified as Weitz-typeredox systems, 31 BBTTs are outscoring their progenitors, i.e. phenothiazines, by significantly lower oxidation potentials with reversible two step oxidations, forming radical cations and dications. 26,28 Herein, we report the first synthesis of BBTTs' radical cations and dications by chemical and electrochemical oxidation as well as the detailed investigation of structural and electronic properties. ## Results and discussion Synthesis and ground state structure N-p-Fluorophenyl-BBTTs were prepared according to our previously published procedure 28 by twofold Buchwald-Hartwig amination (Scheme 2). Starting from respective dibrominated dibenzo[b]thienyl sulfides 1, 2, or 3 28, and p-fluoroaniline (4), all regioisomeric BBTTs 5-7 were reliably synthesized. The original syntheses were upscaled by a factor of ten and even upon reducing catalyst loading the yields for compounds 5 (49%), 6 (45%), and 7 (67%) were maintained. Fluorinated compounds were chosen due to their synthetic accessibility of the neutral BBTT, as well as the opportunity to monitor conversion and easily identify the products by 19 F NMR spectroscopy. As supported previously by CV measurements, the BBTTs' radical cations (5: E 0 0/1+ = 0.21 V, 6: E 0 0/1+ = 0.05 V, 7: E 0 0/1+ = 0.01 V; against ferrocene E 0 0/1+ = 0.00 V) and dications (5: E 0 1+/ 2+ = 0.68 V, 6: E 0 1+/2+ = 0.72 V, 7: E 0 1+/2+ = 0.75 V; against ferrocene E 0 0/1+ = 0.00 V) can be readily generated in solution and they are stable over short periods. 26,28 In addition, the radical cations reveal high semiquinone formation constants K SEM (5: 9.2 × 10 7 , 6: 2.3 × 10 11 , 7: 3.5 × 10 12 ) 28 that claim remarkably stable radical cations against disproportionation into reduced BBTT and dication. 28,31 Consequently, we attempted the preparation of the oxidized species of all three BBTTs. Analogously to dithieno thiazine (DTT) 27,30, chemical oxidation was carried out with antimony pentachloride as oxidant (Scheme 3). 41 By choice of solvent, temperature and equivalents of antimony pentachloride the oxidation stage can be reliably controlled. BBTTs 5-7 are soluble in toluene and dichloromethane, whereas the corresponding hexachloroantimonate salts of the radical cations 5 •+ , 6 •+ , and 7 •+ precipitated from toluene and thereby removed from the reaction equilibrium. Additionally, over-oxidation was prevented by a stoichiometric deficiency of Scheme 2 Selective synthesis of regioisomeric N-p-fluorophenyl-BBTTs 5, 6, and 7 prepared by cyclizing Buchwald-Hartwig amination (for the synthesis of compounds 5 and 6 DCPF (1,1'-bis(d̲ ic̲ yclohexyl-p̲ hosphano)f̲ errocene) was employed as a ligand, and for compound 7 DPPF (1,1'-bis(d̲ ip̲ henylp̲ hosphano)f̲ errocene) was used as a ligand). 28 Scheme 3 Controlled preparation of oxidized BBTT-compounds 5 •+ , 6 •+ , 7 •+ , and 5 2+ , 6 2+ , 7 2+ as hexachloroantimonates using antimony pentachloride as an oxidant. oxidant. If dichloromethane was used as a solvent, the hexachloroantimonate salts of the radical cations remained soluble, forming dication salts upon increasing the amount of antimony pentachloride. The desired oxidation products precipitated from the respective reaction mixture. For assessing the electronic structure, the molecules' geometries were calculated on the DFT level of theory using the uB3LYP functional 42,43 and the 6-311G** basis set 44,45 with SCRF (IEFPCM, CH 2 Cl 2 ) as implemented in the Gaussian 09 program package. 49 Indeed the BBTT units are entirely planarized with a perpendicular orientation of the N-p-fluorophenyl substituents for all three radical cations 5 •+ , 6 •+ , and 7 •+ . Depending on the anellation mode, different spin density distributions were found, however, with highest spin density located on the nitrogen atoms (Mulliken atomic spin density at N-atom: 5 •+ : 0.284, 6 •+ : 0.248, 7 •+ : 0.228). The highest delocalization of spin density is anticipated for the anti-anti derivative 7 •+ , whereas the lowest delocalization is predicted for the syn-syn derivative 5 •+ (Fig. 1). The radical nature of the obtained salts 5 •+ , 6 •+ , and 7 •+ was proven by X-band EPR spectroscopy and by comparison with the simulated spectra (Fig. 2). Spectra with hyperfine coupling patterns of three equidistant lines reflect the coupling of the unpaired electron with the nitrogen nucleus in each case. Furthermore, the range of the g-factors (5 •+ : 2.0106, 6 •+ : 2.0050, 7 •+ : 2.0078) indicates the predominant localization of the radicals on the nitrogen atoms. 41,50 The Mulliken atomic spin density on the nitrogen atoms decreases from syn-syn over syn-anti to anti-anti derivative, which is in accordance with the descending order of hyperfine coupling constants. Hence, an increasing spin delocalization across the BBTT units for anti-anellated benzo[b]thiophene wings can be deduced. The stronger delocalization matches with an increased stability of the corresponding radical cation 7 •+ , which is consistent with the experimentally determined semiquinone formation constants K SEM , which also give largest values for anti-anti radical cation The delocalization of spin density of the radical cations is unequivocally supported by the calculated Wiberg bond orders (Fig. 3). 51 The bond orders of the systems should gradually converge upon oxidation due to increasing orbital overlap by planarization resulting in an increased resonance stabilization. All three regioisomers show the most significant changes in bond orders in the 1,4-thiazine core, where CN and CS bond orders increase while CC bond orders decrease. This effect is most pronounced for the syn-syn regioisomer 5/5 •+ while the antianti isomer 7/7 •+ is most marginally affected. However, antifused benzo[b]-thiophene wings result in higher bond order changes on the fused thiophene in turn. This also confirms the previously observed increased spin delocalization upon introduction of anti-anellated wings. Nevertheless, even with anti-anellation, the spin density on the benzo ring is not high enough for causing observable coupling with the protons' nuclear spins. Although the calculated spin density distributions (Fig. 1) suggest this conclusion, no hyperfine coupling patterns of the protons are experimentally found. For hexachloroantimonate salts of dications 5 2+ , 6 2+ , and 7 2+ , DFT calculations claim formation of singlet dications being thermodynamically favored over triplet dications by energy differences of −92 kJ mol −1 (5 2+ ), −89 kJ mol −1 (6 2+ ), and −90 kJ mol −1 (7 2+ ). Consequently, EPR-silent but NMRactive target compounds should be present. NMR spectra of the bis(hexachloroantimonate) salts of 5 2+ , 6 2+ , and 7 2+ were recorded and compared to the spectra of the native compounds 5, 6, and 7, thus, confirming the formation of diamagnetic dications, i.e. in their singlet ground states (Fig. 4). As expected, the majority of the multiplets shift towards lower field, underlining the electron-deficient dicationic nature of the compounds. Due to structural differences of the regioisomers, parts of the signals appear shielded in high field. DFT-calculated geometry optimizations show planar BBTT units with orthogonally twisted N-p-fluorophenyl substi-tuents, in analogy to the radical cations. syn-Anellated wings place benzo protons into the anisotropy cone of the N-aryl substituent's diamagnetic ring current. Thereby, their signals are shifted to high field. ## Photophysical properties and spectroelectrochemistry For further insight into electronic transitions of oxidized species, i.e. radical cations and dications, spectroelectrochemical measurements were performed in an OTTLE (Optically Transparent Thin Layer Electrochemical) cell, monitoring the in situ formation of radical cations and dications from the neutral precursors 5, 6, and 7 (Table 1). Hence, spectroelectrochemical UV/Vis absorption spectra were recorded by steadily increasing the applied potential in steps of 0.1 V. Initially, UV/Vis spectra of all three parent regioisomers 5, 6, and 7 were recorded showing multiple absorptions bands in the UV region with the longest wavelength maxima at 309 nm (5) and 311 nm (6). For compound 7 the longest wavelength maximum is found in the visible at 424 nm (Fig. 5). All oxidation processes proceed without formation of intermediates, as supported by the occurrence of isosbestic points. Moreover, the oxidations from the neutral BBTTs to the radical cations reflect the largest changes in the UV/Vis spectra and consequently in the electronic structure (vide infra). This originates from major geometry changes from butterfly structures to the planar alignments of the BBTT units (cf. phenothiazines and DTT). 38,52,53 Consequently, oxidized compounds yielded by chemical as well as electrochemical oxidation are deeply colored (Fig. 6). By gradual increase of the potential the formation of radical cations is accomplished, revealing significantly bathochromically shifted absorption bands (Fig. 7). Therefore, compound 5 •+ shows an intense maximum at 680 nm. A strongly broadened absorption band ranging from 450 to 850 nm with shoulders at 566 and 617 nm in conjunction with a maximum at 694 nm is found for radical cation 6 •+ . The two prominent absorption maxima for oxidation product 7 •+ appear at 534 and 717 nm. Further increase of the potential allows monitoring of the formation of dications (Fig. 8). For dication 5 2+ , the longest wavelength absorption maximum at 680 nm shifts slightly hypsochromically, resulting in a band at 672 nm while simultaneously a second band of lower intensity appears at 526 nm. In addition, the most intense band appears at 346 nm. A com- parable absorption band is found at 324 nm for dication 6 2+ . The shoulders observed for radical cation 6 •+ merge into a new maximum at 595 nm, whereas the maximum at 694 nm decreases to a shoulder at 709 nm. Upon oxidation of radical cation 7 •+ to dication 7 2+ , the two intense bathochromically shifted maxima of the radical cation result in the formation of two new maxima at 561 and 657 nm for the dication. Likewise, an intense band appears in the UV at 337 nm. Finally, the spectra from spectroelectrochemical measurements were compared with the spectra of the synthesized hexachloroantimonate radical cation salts (Fig. 9). The superposition shows that the salts unambiguously correspond to the spectroelectrochemically generated specimen and a good overall fit is obtained. Differences in the intensity of the absorption bands in the UV region of the spectra (230-380 nm) can be accounted to counter ion effects. 25 Likewise, the longest wavelength absorption maxima for the hexachloroantimonate salts of the radical cations are only slightly bathochromically shifted (5 ). As known for Weitz-type redox systems, the radical ions are remarkably stable due to their high semiquinone formation constants K SEM (vide supra). 31 However, the dications could not be successfully measured due lacking stability under prolonged measuring conditions. ## Electronic structure of the radical cations and dications For a deeper insight into the electronic structure of the absorption behavior of neutral, radical cation and dication BBTTs TDDFT calculations 54,55 ((u)B3LYP 43,44,56 /6-311++G**, IEFPCM CH 2 Cl 2 ) were performed. First, the optical transitions of the experimentally observed absorption bands of the parent compounds 5, 6, and 7 were assigned (Table 2, Fig. 10). The longest wavelength absorption bands are reasonably well reproduced by transitions at 339 nm (5) and 422 nm (7). They predominantly consist of HOMO → LUMO transitions representing a charge transfer (CT) from the thiazine core to the benzo[b]thiophene wings. The calculated longest wavelength absorption band of 6 appears at 331 nm and can be predominantly assigned to a HOMO → LUMO+1 transition. Again, this transition is accompanied by a CT from the central thiazine to the benzo[b]thiophene units, with a dominant shift of coefficient density due to anti-anellation. Likewise, for assigning the observed UV/Vis absorption of the radical cations, TDDFT calculations with the uB3LYP/6-311++G** functional were performed to prevent spin pairing and to cause separate orbitals occupation (Table 3, Fig. 11). This leads to a singly occupied molecular orbital (SOMO). The longest wavelength absorption bands are almost accurately reproduced by HOMO−1 → SOMO transitions at wavelengths of 665 nm (5 •+ ), 714 nm (6 •+ ) and 701 nm (7 •+ ). For cation 7 •+ , the measured second absorption band at 534 nm is best represented by a HOMO−3 → SOMO transition at 538 nm, while thiophene wings to the central thiazine core, while coefficient density on the N-aryl substituents is absent. Finally, TDDFT calculations (uB3LYP/6-311++G**) were also carried out to assign the absorption bands of the dications 5 2+ , 6 2+ , and 7 2+ (Table 4, Fig. 12). The longest wavelength absorption bands are best represented by calculated HOMO → LUMO transitions at wavelengths of 736 (5 2+ ), 709 (6 2+ ), and 747 nm (7 2+ ) that can be interpreted as CT transitions from the outer benzo[b]thiophene part to the central thiazine unit. The following experimental bands in the visible are assigned to transitions arising from HOMO−4 → LUMO at 517 nm (5 2+ ), HOMO−1 → LUMO at 617 nm (6 2+ ), and HOMO−2 → LUMO at 579 nm (7 2+ ). These transitions are characterized by CT contributions from the N-aryl moiety to the thiazine core for structure 5 2+ , whereas for structure 6 2+ by a shift of coefficient density from the benzo[b]thiophene parts to the thiazine core. For structure 7 2+ , the transition can be accounted as a combination of both preceding coefficient density shifts. The preceding calculations reveal that BBTT are strongly polarizable systems with large CT contributions. Accordingly, neutral BBTT are potent donors due to their electron richness, whereas the oxidized compounds are accompanied by strong acceptor properties. ## Aromaticity of the dications Furthermore, the dications' thiazine cores should inhere aromatic character due to their Weitz-type nature. Proving this, [nm] syn-syn 5 starting with the previously mentioned optimized planar geometries, NICS (0) and NICS (1) (above (+1) and under (−1) the plane) calculations were performed. Using the GIAO protocol with B3LYP functional and 6-311+G** basis set 44,45 calculations were carried out for all five rings of the BBTT core, giving NICS values indicating diatropic ring current stating aromaticity (Table 5). In a dry Schlenk tube with a magnetic stir bar under Ar atmosphere were placed dibrominated sulfane 1, 2, or 3 (2.3 g, 5.0 mmol), Pd(dba) 2 (0.14 g, 5.0 mol%) and ligand (10 mol%). By glovebox technique sodium tert-butoxide (1.4 g, 15 mmol) was added into the reaction tube. Subsequently, p-fluoroaniline (0.47 mL, 5.0 mmol) und dry toluene (30 mL) were added by syringe before stirring the reaction mixture at 110 °C (oil bath) for 65 h. After cooling to room temp the reaction was terminated by addition of saturated aqueous sodium sulfite solution (400 mL). The aqueous phase was extracted with dichloromethane (3 × 300 mL), the combined organic phases were dried with anhydrous magnesium sulfate and the solvents were removed under vacuum. After adsorption to Celite® the residue was purified by chromatography on silica gel. After recrystallization from n-hexane the products 5, 6, or 7 were isolated as crystals. Deviations from this general procedure are individually highlighted. 2) and 1,1′-bis (dicyclohexyl-phosphano)ferrocene (0.29 g, 0.50 mmol) as a ligand and after chromatography (n-hexane) and recrystallization from n-hexane (550 mL) compound 6 (0.92 g, 2.3 mmol, 45%) was obtained as yellow crystals, Mp 201 °C. 1 H NMR (300 MHz, acetone-d 6 ) δ 6.92-6.98 (m, 1 H), 7.17 According to the general procedure with bis(2-bromobenzo[b] thiophen-3-yl)sulfane (3) and 1,1′-bis(diphenyl-hexylphosphano)ferrocene (0.29 mg, 0.50 mmol) as a ligand and after chromatography (n-hexane/dichloromethane/triethylamine 100 : 1 : 3) and recrystallization from n-hexane (450 mL) compound 7 (1. 35 ## General proceduresynthesis of bis[1]benzothieno[1,4] thiazine radical cations (BBTT •+ ) In a dry Schlenk tube with a magnetic stir bar the N-4-fluorophenyl-BBTT 5, 6, or 7 (0.12 g, 0.30 mmol) was dissolved in dry toluene (60 mL). Under vigorous stirring, a 1 M solution of antimony pentachloride in dichloromethane (0.30 mL, 0.30 mmol) was added dropwise to the solution at room temp. After 45 min the precipitate was filtered off and washed with dry toluene (5 × 10 mL) and dry n-pentane (2 × 10 mL). Deviations from this general procedure are individually highlighted. 12-(4-Fluorophenyl)benzo 12-(4-Fluorophenyl)benzo thieno[2,3-b]benzo thieno [2,3-e] thiazinyl-12-ium hexachloroantimonate (syn-anti 6 6-(4-Fluorophenyl)benzo thieno [3,2-b]benzo thieno[2,3e] thiazinyl-6-ium hexachloroantimonate (anti-anti 7 In a dry Schlenk tube with a magnetic stir bar was dissolved N-4-fluorophenyl-BBTT 5, 6, or 7 (0.12 g, 0.30 mmol) in dry dichloromethane (9 mL). Under vigorous stirring a 1 m solution of antimony pentachloride in dichloromethane (0.60 mL, 0.60 mmol) was added dropwise to the solution at −78 °C (dry ice/acetone). After 45 min the mixture was allowed to warm to room temp. The precipitate was filtered off under Ar atmosphere and washed with dry dichloromethane (5 × 10 mL). Deviations from this general procedure are individually highlighted. 12-(4-Fluorophenyl)benzo ) 2 (0.10 g, 48%) was obtained as a dark blue solid. 1 12-(4-Fluorophenyl)benzo thieno[2,3-b]benzo ## Conclusions The electronic properties of oxidized species of all three anellative bis benzothieno thiazine (BBTT) isomers were exten-sively characterized by experimental and computational methods. All three regioisomeric N-p-fluorophenyl-BBTTs (synsyn, syn-anti and anti-anti) were successfully transformed into the corresponding radical cation and dication salts by using antimony pentachloride as an oxidant on a preparative scale. By EPR spectroscopy the radical character was unambiguously proven and the predominant localization of the unpaired spin on the central 1,4-thiazine moiety was supported by the resulting hyperfine coupling with the nitrogen nucleus. Moreover, experimentally determined hyperfine coupling constants and computationally determined spin density distributions and Wiberg bond orders underline that a stronger delocalization arises from anti-alignment of the benzo[b]-thiophene units. The diamagnetic singlet character as well as geometric structure of the dications was assigned by NMR spectroscopy, especially supported by high-field shifted signals in diamagnetic anisotropy cones of syn-aligned wings due to planarization of the BBTT core. Spectroelectrochemical measurements by in situ generation of the oxidized species allowed characterization of the electronic absorption bands. Finally, the experimentally observed absorption bands could be reproduced TDDFT calculations and assigned to the underlying molecular orbital transitions. The calculated electronic structure of the radical cations and dications underlines for all transitions a high polarizability of BBTT upon photonic excitation, as revealed by pronounced charge transfer from the benzo[b]thiophene moieties or the N-aryl unit to the central 1,4-thiazine core, stating a strong acceptor character of oxidized BBTT. Further studies on the synthesis of functionalized BBTTs focusing on enhancing their neutral antiaromatic character as well as addressing their electroactive and electro-optical applications are currently underway.

chemsum

{"title": "Radical cations and dications of bis[1]benzothieno[1,4]thiazine isomers", "journal": "Royal Society of Chemistry (RSC)"}

discovery_of_xl01126:_a_potent,_fast,_cooperative,_selective,_orally_bioavailable_and_blood_brain_ba

## Abstract: Leucine Rich Repeat Kinase 2 (LRRK2) is one of the most promising targets for Parkinson's Disease. LRRK2 targeting strategies have primarily focused on Type 1 kinase inhibitors, which however have limitations as the inhibited protein can interfere with natural mechanisms which could lead to undesirable side effects. Herein, we report the development of LRRK2 Proteolysis Targeting Chimeras (PROTACs), culminating in the discovery of degrader XL01126, as an alternative LRRK2 targeting strategy. Initial designs and screens of PROTACs based on ligands for E3 ligases VHL, CRBN and cIAP identified the best degraders containing thioether-conjugated VHL ligand VH101. A second round of medicinal chemistry exploration led to qualifying XL01126 as a fast and potent degrader of LRRK2 in multiple cell lines, with DC50 values within 15-72 nM, Dmax values range from 82-90%, and degradation half-lives span from 0.6h to2.4h. XL01126 exhibits high cell permeability and forms a positively cooperative ternary complex with VHL and LRRK2 (α=5.7), which compensates for a substantial loss of binary binding affinities to VHL and LRRK2, underscoring its strong degradation performance in cells. Remarkably, XL01126 is orally bioavailable (F=15%) and can penetrate the blood brain barrier after either oral or parenteral dosing in mice. Taken together, these experiments qualify XL01126 as a suitable degrader probe to study non-catalytic and scaffolding functions of LRRK2 in vitro and in vivo and offer an attractive starting point for future drug development. ## Introduction Around 10 million people worldwide are living with Parkinson's disease (PD) 1 , a progressive neurodegenerative disorder characterized by both motor (e.g. bradykinesia, resting tremor, postural instability, rigidity) and non-motor (e.g. memory loss, hyposmia) disabilities. Current PD treatment is limited to motor symptoms management with dopamine replacement or by enhancing the activity of the remaining dopaminergic neurons. No known therapy is available that can slow down the progress or prevent the onset of the disease. Furthermore, PD cases are growing at a fast ever speed and are projected to increase to over 17.5 million by 2040 due to the fast-growing aging population 2 . While aging remains to be the major risk factor of PD, >20 genes have been identified to be associated with the onset and progress of PD 3 , suggesting the potential of discovering disease-modifying PD treatments. Leucine-rich repeat kinase 2 (LRRK2), encoded by LRRK2 gene, is a large (286 kDa), multi-domain protein that, in addition to its kinase domain, possesses a second enzymatic guanosine triphosphatase (GTPase) domain and several other domains and motifs that are involved in protein-protein interactions 4 . Pathological mutations in the kinase domain and GTPase domain of LRRK2, such as G2019S and R1441C/G/H mutations, can increase the kinase activity of LRRK2 and eventually lead to pathogenic hallmarks associated with PD, such as ciliogenesis inhibition 5,6 , defective mitophagy and autophagy , and mitochondrial dysfunction 10 . Increased LRRK2 kinase activity, independent of LRRK2 mutations, has also been reported in idiopathic PD patients 11 . Conversely, LRRK2 knockout or pharmacological inhibition of LRRK2 kinase activity are neuroprotective in cellular and in animal models . These observations provide strong rationale for targeting LRRK2 to treat PD. Over the past years, several LRRK2 kinase inhibitors have been developed, including LRRK2-IN-1 16 , HG-10-102-01 17 , MLi-2 18 , PF-06447475 19 , and DNL201 and DNL151 which are the first two LRRK2 kinase inhibitors in clinical trials 20 . However, all these inhibitors are ATP-competitive type 1 kinase inhibitors which preferably bind to the closed active conformation of LRRK2, leading to dephosphorylation of Ser935 and other biomarkers sites, LRRK2 aggregation and mislocalization to microtubules 21,22 . These unintended effects may interfere with vesicle trafficking and could underlie undesirable on-target side-effects observed on lungs and kidneys 23,24 . Alternative LRRK2 targeting strategies, such as G2019S LRRK2 selective inhibitors 25,26 , LRRK2 dimerization inhibitors 27 , GTPase inhibitors, antisense oligonucleotide 28 , type 2 LRRK2 kinase inhibitors 29 , and LRRK2 proteolysis targeting chimeras (PROTACs) , have therefore been proposed and are under active exploration. As one of the most promising disease-modifying targets, LRRK2 lies at the nexus of an emerging signaling network of high relevance for understanding and developing treatments for PD 34 . Although three LRRK2 targeting therapies 28,35,36 are already in clinical trials, the exact mechanism by which LRRK2 mutations and its kinase activity contribute to the development of PD is still under investigation. Rab GTPases implicated in vesicular trafficking have been identified as bona fide physiological substrates of LRRK2 37 , but many components involved in the upstream and downstream wiring of LRRK2 signaling pathways are yet to be discovered, and the question remains as to whether LRRK2 kinase inhibitors will have beneficial disease-modifying effects in PD patients. More in-depth LRRK2 target validation is therefore warranted. Induced target protein degradation is a paradigm-shifting drug discovery approach. Heterobifunctional degraders (also known as PROTACs) can induce target protein degradation by recruiting an E3 ubiquitin ligase in proximity to the target protein, resulting in the polyubiquitination and subsequent degradation of the target protein by the proteasome . More than 15 PROTAC degraders are in or approaching the clinic currently , against a variety of targets, including hormone receptors (e.g. AR and ER), transcription factor (e.g. STAT3), anti-apoptotic protein (e.g. BCL-XL), kinases (e.g. BTK and IRAK4), and epigenetic proteins (e.g. BRD9). PROTAC is not only an emerging drug discovery modality but also offers new chemical tools for target identification and validation, and for deciphering target biology 43,44 . For example, PROTAC-mediated degradation can reveal non-catalytic activity of protein kinases 45,46 . Herein, we report the discovery and characterization of XL01126, a von Hippel-Lindau (VHL)-based, fast, potent, cooperative and selective LRRK2 PROTAC degrader that is also orally bioavailable and blood brain barrier (BBB) permeable. XL01126 qualifies as a chemical probe to study LRRK2 biology, further validate the target as a therapeutic concept in PD, and usher future drug development. ## Identification of initial VH101 thioether-linked PROTACs as moderate LRRK2 degraders We began our efforts by designing and synthesizing a small set of PROTACs aiming to maximize sampling of chemical space and target-PROTAC-E3 ternary complex pairing. HG-10-102-01(Figure 1), a BBB penetrant type 1 LRRK2 inhibitor, was chosen as the LRRK2 ligand, on the basis of its small molecular size and favorable physicochemical properties 17 . According to homology modeling of HG-10-102-01 with LRRK2, the morpholine ring is pointing towards solvent 17 , suggesting of a suitable exiting vector for PROTAC linkage. We converted the morpholine ring to piperazine to facilitate linker attachment. For the E3 ubiquitin ligases, we decided to recruit Cereblon (CRBN), Cellular Inhibitor of Apoptosis (cIAP), and VHL, which have readily available ligands with known PROTACable sites 47 (Figure 1). After converting both the warhead and E3 ligase ligands into "PROTACable" intermediates, they were tethered together through linkers and a small library of first-generation compounds containing 12 LRRK2 PROTACs (Figure S1) were generated (Schemes S3, S4, and S5, ). Figure 2. Screening of the first-generation PROTACs in WT and G2019S LRRK2 MEFs. (A) Representative Western blots monitoring total LRRK2, LRRK2-pSer935, Rab10-pThr73, total Rab10, and Tubulin levels following the treatment of WT and G2019S MEFs with the indicated compounds at 33 nM, 1 µM, or DMSO for 4h. (B) Quantitative analysis of the relative LRRK2 protein and Rab10-pThr73 levels, which are presented as ratios of total LRRK2/Tubulin or Rab10-pThr73/total Rab10, normalized to the DMSO treated sample. Data were obtained from two biological independent experiments. These PROTACs were then biologically evaluated in mouse embryonic fibroblasts (MEFs) by Western blotting. Briefly, MEFs were treated with compounds at 33 nM and 1 µM for 4 h (Figure 2) and 24 h (Figure S2) separately, and the intracellular level of LRRK2, phosphorylated LRRK2 at Ser935, and phosphorylated Rab10 (pRab10) at Thr73 were determined. Rab10 5,48 is one of the bona fide substrates of LRRK2, whose phosphorylation status is directly affected by LRRK2 kinase activity and protein level. Phosphorylation of LRRK2 at Ser935 is a well-studied biomarker site used to assess the efficacy of type 1 LRRK2 inhibitors 29 . HG-10-102-01 based PROTACs can potentially dephosphorylate LRRK2 at Ser935 through both LRRK2 degradation and inhibition. Among the first-generation PROTACs, compounds SD75, SD82, and SD100 (Figure 1) degraded 30-70% of G2019S LRRK2 at 1 µM/4h (Figure 2) and achieved 70-85% G2019S LRRK2 degradation after 1 µM/24h treatment (Figure S2). These three compounds also showed substantial dephosphorylation of LRRK2 and Rab10, with 75-90% pRab10 dephosphorylated after 1 µM/24h treatment in G2019S LRRK2 MEFs (Figure S2). A fourth compound, SD13, also looked promising as it degraded 60% G2019S LRRK2 at 33 nM/4h treatment (Figure 2) and 68% G2019S LRRK2 at 33 nM/24h (Figure S2). However, less G2019S LRRK2 was degraded upon 1µM treatment by SD13, compared to the 33 nM treatments, suggestive of the "hook effect" 49 . Although SD75, SD82, and SD100 showed only moderate LRRK2 degradation, they did not show any sign of the "hook effect" at 1µM concentration. Notably, all three compounds share the same E3 ligase and ligand (VHL, VH101) and exit vector out of the tert-leucine group, suggesting a potential hot-spot of ternary complex formation between VHL and LRRK2. We therefore decided to focus further medicinal chemistry optimization on this chemical series with the goal to further improve the compounds' fitness as LRRK2 degraders. Given the modular nature of PROTAC molecules, the structural modification of the second generation of LRRK2 PROTAC degraders focused on modifying the LRRK2 ligand, the linker, and the VHL ligand (Figure 1 and Figure 3), separately. To best assess which structural modification would confer the most significant activity improvement, we designed molecular match pairs of SD75, SD82, and SD100 by changing one structural moiety at a time. XL01078B, XL01072, and XL01070B were designed (Figure 3) and synthesized (Scheme S3) where the 5-chlorine substitution on the aminopyrimidine ring of HG-10-102-01 was replaced with -CF3 substitution which was reported to improve binding affinity to LRRK2 50 . XL01119, XL01118 and XL01120 (Figure 3) were molecular match pairs of SD75, SD82, and SD100 respectively, by harboring an extra methyl group on the benzylic position of VHL ligand, which was introduced to increase binding affinity to VHL E3 ligase 51 . Fluorine substitution was introduced on the phenyl group of the VHL ligand of XL01123, XL01122, and XL01121 (Figure 3) attempting to fine-tune physicochemical properties at a permissible site 51 . The linker length, composition, and rigidity, which can significantly affect the physicochemical and pharmacokinetic properties of PROTACs, as well as their ternary complex formation and activity 51,54,55 , were explored as represented by compounds XL01131, XL01140, XL01111, XL01126, XL01134, and XL01076 (Figure 3). In an attempt to improve the drug-like properties and reduce molecular size, we designed XL01145, XL01149 and XL01168 (Figure 3). These compounds are derived from truncated HG-10-102-01 with the morpholinoamide moiety removed as its absence retains binary binding affinity to LRRK2 50 . These 18 new compounds were synthesized as outlined in Schemes 1, S1, S2, S3, S6, S7, S8 and S9 and were also screened via Western blotting (Figure 4 and Figure S3). Quantitative analysis of the Western blots (Figure 4B and Figure S3B) revealed that at 33 nM/4h treatment, XL01126 and XL01134 were the most effective optimized compounds that degraded 20-30% of WT LRRK2 and 50-60% of G2019S LRRK2 (Figure 4B). Accordingly, these two compounds were also the most potent in decreasing pRab10 in both WT and G2019S LRRK2 MEFs, with > 60% pRab10 inhibited in G2019S LRRK2 MEFs at 33 nM/4h (Figure 4B). In contrast, the first-generation degraders SD75, SD82, and SD100 induced little to no degradation of LRRK2 at 33 nM/4h treatment (Figure 4) and showed weak (<40%) degradation at 33 nM/24h treatment (Figure S3), at which XL01126 and XL01134 degraded 50-60% of WT LRRK2 and 70-80% of LRRK2 G2019S (Figure S3). Most of the compounds exhibited substantial WT LRRK2 and G2019S LRRK2 degradation (30-80%) at 1 µM/4h or 1 µM/24h treatment (Figure 4 and Figure S3), leading to potent and almost complete pRab10 inhibition in WT MEFs and G2019S LRRK2 MEFs, respectively. Multiple new compounds, including XL01078B, XL01119, XL01123, XL01131, XL01126 and XL01134, surpassed SD75, SD82 and SD100 in degrading WT LRRK2 and G2019S LRRK2 at 1 µM/4h and 1 µM/24h treatment, suggesting modifications at the warhead (XL01078B), the E3 ligase ligand (XL01119 and XL01123), and the linkers (XL01131, XL01126, and XL01134) can all improve the degraders' fitness to some extent. Nonetheless, the significant improvement exhibited by XL01126 and XL01134, which are isomers of each other, encouraged us to characterize them further. Figure 4. Screening of the second-generation PROTACs in WT and G2019S LRRK2 MEFs. (A) Representative Western blots monitoring total LRRK2, LRRK2-pSer935, Rab10-pThr73, total Rab10, and Tubulin levels after treating WT and G2019S MEFs with the indicated compounds at 33 nM, 1 µM, or DMSO for 4h. (B) Quantitative analysis of the relative LRRK2 and Rab10-pThr73 levels, which are presented as ratios of total LRRK2/Tubulin or Rab10-pThr73/total Rab10, normalized to the DMSO treated sample. Data were obtained from two biological independent experiments. ## Identification of XL01126 as a potent and fast LRRK2 degrader To characterize XL01126 and XL01134 and compare them with the top first-generation degrader SD75, a dose dependent degradation assay was carried out in WT and G2019S LRRK2 MEFs (Figure 5). SD75 dose-dependently degraded LRRK2 following 24h treatment in WT and G2019S LRRK2 MEFs (Figure 5A). However, the degradation of LRRK2 was only partial with Dmax reached at 3 µM (Dmax,24h = 51% and 58% for WT and G2019S LRRK2 respectively). Dose-dependent LRRK2 pSer935 and pRab10 dephosphorylation, which account for both LRRK2 inhibition and degradation, were also observed after SD75 treatment, with EC50 = 2270 nM and 379 nM for the dephosphorylation of Rab10 in WT and G2019S LRRK2 MEFs, respectively. XL01134 and XL01126, the top LRRK2 degraders from the second-generation compounds, showed more extensive LRRK2 degradation after a significantly shorter treatment time (4h) when compared to SD75 (Figure 5B and 5C). XL01134 degraded G2019S LRRK2 (DC50, 4h = 7 nM) more potently than WT LRRK2 (DC50, 4h = 32 nM), with maximum LRRK2 degradation reached at 300 nM and Dmax values against WT LRRK2 and G2019S LRRK2 are 59% and 81%, respectively. However, at concentrations above 300 nM, a strong "hook effect" was observed (Figure 5B). XL01126 also degraded G2019S LRRK2 (DC50, 4h = 14 nM) and WT LRRK2 (DC50, 4h = 32 nM) at nano-molar concentrations, but achieved more complete degradation than XL01134, with Dmax,4h = 82% in WT MEFs and Dmax,4h = 90% in G2019S LRRK2 MEFs, achieved at around 1 µM. Moreover, no "hook-effect" was observed with XL01126 at higher concentrations (Figure 5C). Due to the potent LRRK2 degradation capabilities, XL01134 and XL01126 resulted in more pronounced pRab10 dephosphorylation (Figure 5B and 5C) than SD75. XL01134, at 4h, showed 30-fold more potent pRab10 inhibition than SD75 (at 24h) in both WT MEFs and G2019S LRRK2 MEFs. XL01126 (at 4h) is 40-fold more potent than SD75 (at 24h) in inhibiting Rab10 phosphorylation in WT MEFs, and 25-fold more potent in G2019S LRRK2 MEFs. Figure 5. Dose dependent LRRK2 degradation, LRRK2 dephosphorylation, and Rab10 dephosphorylation by SD75, XL01134, and XL01126 in WT and G2019S LRRK2 MEFs. Representative Western blots of total LRRK2, LRRK2-pSer935, Rab10-pThr73, total Rab10, and Tubulin levels after treating WT and G2019S LRRK2 MEFs with SD75 (A), XL01134 (B), or XL01126 (C) at the indicated concentrations for indicated time period. The relativeLRRK2 protein and pRab10 levels were obtained by quantifying the ratios of total LRRK2/Tubulin or Rab10-pThr73/total Rab10, respectively, and the ratios were normalized to the DMSO treated samples. The relative total LRRK2 and pRab10 levels were plotted against the compounds concentration and fitted against "non-linear regression, one site-fit LogIC50" in GraphPad to obtain the DC50 and EC50 values. Data were obtained from two to three biological independent experiments. To further compare the degradation profiles of XL01134 and XL01126 with that of SD75, a time-dependent degradation assay was performed in MEFs using Western blotting (Figure 6). SD75 was shown to degrade WT LRRK2 and G2019S LRRK2 at 1 µM in a time-dependent manner with moderate Dmax (52% for WT LRRK2 and 81% for G2019S LRRK2) and half-lifes (T1/2) against WT LRRK2 (5.1h) and G2019S LRRK2 (1.4h). In contrast, XL01134 and XL01126 degraded LRRK2 at higher rates, achieved higher Dmax values at only 300 nM, a concentration at which SD75 barely degraded LRRK2. Remarkably, XL01126 presented an improved profile (Dmax-WT = 82%, Dmax-G2019S = 92%, T1/2-WT = 1.2h, T1/2-G2019S = 0.6h) when compared to XL01134 (Dmax-WT= 75%, Dmax-G2019S = 82%, T1/2-WT= 2.7h, T1/2-G2019S = 1.4h). With the shortest degradation half-lives and highest degradation percentage, XL01126 emerged as the most efficient and fastest degrader among the three. The time-dependent pRab10 dephosphorylation correlates well with the LRRK2 degradation (Figure 6A, 6B and 6C). XL01126 dephosphorylated pRab10 the fastest with T1/2, pRab10 at 0.7h and 0.3h in WT and G2019S LRRK2 MEFs, respectively. This was followed by XL01134, which induced 50% reduction in Rab10 phosphorylation after 2.1h and 0.3h in WT and G2019S LRRK2 MEFs, respectively. In contrast, SD75 exhibited the slowest inhibition of pRab10 (T1/2, pRab10 = 6.7h on WT MEFs and 1.1h on G2019S LRRK2 MEFs). Figure 6. Time-dependent LRRK2 degradation, LRRK2 dephosphorylation and pRab10 dephosphorylation by SD75, XL01134, and XL01126. Representative Western blots of total LRRK2, LRRK2-pSer935, Rab10-pThr73, Rab10 total, and Tubulin levels after treating the WT and G2019S LRRK2 MEFs with SD75 (A), XL01134 (B), or XL01126 (C) at the indicated concentrations for the indicated period of time. The relativeLRRK2 protein and pRab10 levels were obtained by quantifying the ratios of total LRRK2/Tubulin or Rab10-pThr73/total Rab10, respectively, and the ratios were normalized to the DMSO treated samples. The relative LRRK2 and pRab10 protein levels were plotted against the treatment time and were fitted against "non-linear regression, one phase decay" in GraphPad to obtain the half-life (T1/2) values. Data were obtained from two independent biological experiments. XL01126 surpassed its warhead and negative PROTAC cis-XL01126 in inhibiting downstream signaling in G2019S LRRK2 MEFs. Representative Western blots of total LRRK2, LRRK2-pSer935, pRab10, Rab10 total, and Tubulin levels following the treatment of G2019S LRRK2 MEFs with HG-10-102-01 (A), XL01126 (A and B), and cis-XL01126 (B) at the indicated concentrations for 4h. The relative LRRK2 protein and pRab10 levels were obtained by quantifying the ratios of total LRRK2/Tubulin or Rab10-pThr73/total Rab10, respectively, and the ratios were normalized to the DMSO treated samples. The relative LRRK2 and pRab10 protein levels were plotted against the compounds concentration and fitted against "non-linear regression, one site-fit LogIC50" in GraphPad to obtain the DC50 and EC50 values. Data were obtained from two independent biological experiments. The potent and fast degradation of LRRK2 and inhibition of the Rab substrate phosphorylation by XL01126 prompted us to question if our PROTAC could surpass its warhead (HG-10-102-01) in dephosphorylating the substrate of LRRK2 and how much of the substrate dephosphorylation results from the protein degradation. This is of particular relevance for this project because the warhead ligand itself is a strong LRRK2 inhibitor with nanomolar kinase inhibition activities (Figure S5) 17 , and is a general challenge with PROTACs against protein kinase. As expected, the warhead HG-10-102-01 did not degrade LRRK2, but potently inhibited LRRK2 phosphorylation and Rab10 phosphorylation (EC50 = 110 nM on G2019S LRRK2 MEFs, EC50 = 214 nM on WT MEFs) (Figure 7 and Figure S4). In contrast, XL01126 dose-dependently degraded both WT LRRK2 (Fig- ure S4) and G2019S LRRK2 (Figure 7A). Crucially, XL01126, showed around 3-fold more potently inhibited Rab10 phosphorylation in WT MEFs than HG-10-102-01 (Figure S4A), and 6-fold more potently inhibited Rab10 phosphorylation in G2019S LRRK2 MEFs (Figure 7A). These observations suggest that converting HG-10-102-01 to a PROTAC degrader not only improves downstream signaling inhibition, but also increases selectivity for G2019S LRRK2 over WT. Cis-XL01126 (Scheme 1), a non-degrading distomer control of XL01126 where the stereochemistry at the hydroxyl group of hydroxyproline is inverted to abrogate VHL binding 56 , showed no degradation of WT LRRK2 (Figure S4B) and G2019S LRRK2 (Figure 7B), but inhibited Rab10 phosphorylation at a similar potency as HG-10-102-01 in both WT MEFs (Figure S4) and G2019S LRRK2 MEFs (Figure 7). However, due to the lack of LRRK2 degradation, cis-XL01126 was around 7-fold less potent than XL01126 in inhibiting Rab10 phosphorylation (116 nM vs 15 nM), further demonstrating the potency boost in downstream functionality achieved from LRRK2 degradation over and above kinase inhibition. Figure 8. XL01126 degrades LRRK2 in human peripheral blood mononuclear cells (PBMCs) derived from healthy donors and mouse bone marrow derived macrophages (BMDMs). Representative Western blotting of total LRRK2, LRRK2-pSer935, pRab10, Rab10 total, and GAPDH levels following treating the PBMCs with XL01126 and cis-XL01126 at the indicated concentrations for 4h (A) and 24 h (B). The relative LRRK2 protein and pRab10 levels were obtained by quantifying the ratios of total LRRK2/GAPDH or Rab10-pThr73/total Rab10, respectively, and the ratios were normalized to the DMSO treated samples. The relative LRRK2 and pRab10 protein levels were plotted against the compounds concentration and fitted against "non-linear regression, one site-fit LogIC50" in GraphPad to obtain the DC50 and EC50 values. Data points are presented as mean ± SEM from three biological independent replicates. (C) Representative Western blotting of total LRRK2, LRRK2-pSer935, pRab10, Rab10 total, and GAPDH levels following treating the PBMCs with 300 nM of XL01126 and cis-XL01126 for the indicated time periods. The relative LRRK2 protein and pRab10 levels were obtained by quantifying the ratios of total LRRK2/ GAPDH or Rab10-pThr73/total Rab10, respectively, and the ratios were normalized to the DMSO treated samples. The relative LRRK2 and pRab10 protein levels were plotted against the treatment time and were fitted against "non-linear regression, one phase decay" in GraphPad to obtain the half-life (T1/2) values. Data points are presented as mean ± SEM from three biological independent replicates. (D) Representative Western blotting of LRRK2 total and Tubulin levels after treating BMDMs with XL01126 and cis-XL01126 for 4h. The relative LRRK2 levels were obtained by quantifying the ratios of total LRRK2/Tubulin, and normalized to the DMSO treated samples. The relative LRRK2 levels were plotted against the compounds concentration and fitted against "non-linear regression, one site-fit LogIC50" in GraphPad to obtain the DC50 values. To scope and assess the degradation activity of XL01126 on other LRRK2 mutants and cell lines, dose-dependent degradation assays of XL01126 were carried out in R1441C LRRK2 MEFs (Figure S6), bone marrow-derived macrophages (BMDMs), and human peripheral blood mononuclear cells (PBMCs) (Figure 8). XL01126 exhibited potent LRRK2 degradation in all cell types, with significant differentiation observed between XL01126 and cis-XL01126 in terms of Rab10 dephosphorylation (Table 1, Figure S6, and Figure 8). The fast (T1/2, 300 nM = 2.4h) and potent (DC50, 4h = 72 nM, DC50, 24h = 17 nM,) degradation of human LRRK2 in PBMCs suggests the potential of applying XL01126 to additional human cell lines. ## XL01126 induces cooperative ternary complex formation As the top two degraders from the second generation, XL01126 and XL01134 are epimers of each other, the only difference being swapped chirality at one of the two tertiary carbons of the cyclohexyl ring in their linkers. This small difference in chemical structure gives rise to very different degradation profiles for the two compounds (Figure 5 and Figure 6). These two epimeric PROTACs also exhibited strikingly different binding affinities to VHL as revealed by a fluorescence polarization (FP) displacement binding assay (Figure 9A) 49,57 and a VHL target engagement assay (Figure 9B) 58 . XL01126 has >10-fold weaker binary binding to VHL than XL01134 and also was found to be the weakest LRRK2 binder amongst the compounds tested (Figure 9C) PROTACs have previously been shown to tolerate weakened binary binding affinities to either their E3 ligase 59,60 or target protein 61,62 such that, despite the weak binding, they are able to induce potent protein degradation at a concentration well below the weakened Kd. Conversely, PROTACs made of more potent target ligands do not necessarily guarantee for more potent degraders 55,61 These studies together illustrated a now well-established feature with PROTACs, that is the extent of target degradation does not necessarily correlate with the PROTAC's binary binding affinity to E3 ligase or target protein. The ternary binding affinity, cooperativity and stability of the ternary complex, can instead play critically important roles in PROTAC induced protein degradation . To test whether our PROTACs can induce cooperative ternary complex formation and illuminate the relationship between the degradation potency and ternary complex formation, a ternary binding affinity assay and a ternary complex formation assay are warranted. However, we could not implement the mostly commonly used biophysical techniques such as fluorescence polarization 57 and surface plasma resonance 67 for these assays, due to the lack of sufficient recombinant expressed LRRK2 in hand. We therefore turned to endogenously expressed LRRK2 and developed a NanoBRET-based ternary binding affinity assay and ternary complex formation assay in HEK293 cells (Figure 10). In the NanoBRET-based ternary binding affinity assay, LRRK2-NanoLuc was transiently expressed in HEK293 cells as the BRET donor and LRRK2 tracer which is prepared by conjugating HG-10-102-01 with a fluorophore (BOD-IPY 576/589 ) (Figure 10A and Scheme S11) was introduced as the acceptor. Titration of PROTAC degraders to the lysed cells and LRRK2 tracer in the presence or absence of recombinant VCB protein (VHL complexed with elongin B-elongin C) gives ternary and binary binding affinities of PROTACs against LRRK2 respectively. Similarly, the ternary complex formation assay also used LRRK2-NanoLuc transiently expressed in HEK293 as the BRET donor, but the acceptor was recombinant VCB protein labeled with BODIPY 576/589 via NHS ester-activated crosslinking reaction. PROTACs that can bridge LRRK2 and VCB together will produce BRET signal (Figure 10E). In line with the degradation potency, XL01126 induced the most cooperative ternary complex as indicated by its positive cooperativity (α = 5.7) (Figure 10B) and the highest maximal level of ternary complex formation (Figure 10F). In contrast, XL01134 induced significantly lower cooperativity (α = 1.4) and SD75 has a negative cooperativity with VHL and LRRK2 (Figure 10) In the NanoBRET-based ternary complex formation assay, SD75, although a less potent degrader than XL01134, induced higher ternary complex than XL01134. However, it should be noted that this assay was carried out in the permeabilized HEK293 cells, and SD75 is likely to induce less intracellular ternary complex formation given its relatively lower permeability comparing to XL01134 (Figure 9D). The relative permeability (intracellular availability) of each compound was obtained by querying VHL engagement or LRRK2 engagement under live-cell and permeabilizedcell conditions 58,68 (Figure 9). XL01126 induced LRRK2 degradation is selective and dependent on the ubiquitin proteasome system To assess the degradation selectivity of XL01126 and identify potential off-targets at the proteome level, we performed unbiased quantitative tandem mass tag (TMT)based global proteomic profiling in WT MEFs. Over 8000 proteins were quantified in the cell lysate samples from WT MEFs that were treated with 300 nM XL01126, cis-XL01126, or DMSO for 4h (Figure 11). The data corroborate a significant chemical knockdown of LRRK2, as validated by Western blotting (Figure S7). LRRK1, the closest homologue of LRRK2, and other LRRK2-related proteins such as VPS35 and Rab-specific phosphatase PPM1H remained unaffected. The proteomic data also revealed a small (~30%) depletion in protein levels of phosphodiesterase 6d (PDE6D) (Figure 11). PDE6D has a deep hydrophobic ligand-binding pocket, and has been shown to be degradable via PROTACs 43,69 Curiously, PDE6D was also found as adventitious off-target degradation of PTK2 PROTACs previously 70 . Inspection of chemical structures highlighted that the PTK2 PROTAC and XL01126 share a similar aminopyrimidine warhead at the target ligand end, a moiety known to be critical to the high binding affinity in PDE6D inhibitor Deltasonamide 69 , suggesting a potential off-target degradation due to adventitious PROTAC binding to PDE6D. Dose-dependent degradation of PDE6D in both WT MEFs and LRRK2 KO MEFs as shown via Western blotting (Figure S7) indicated that XL01126 induced PDE6D degradation is LRRK2-independent and excluded it being a downstream consequence of LRRK2 degradation. A study examining the mechanism of LRRK2 degradation demonstrated that degradation by XL01126 is mediated by the ubiquitin-proteasome system as XL01126-induced degradation can be blocked by VHL ligand (VH101), neddylation inhibitor (MLN4924), and proteasome inhibitor (MG132) pretreatments in both WT MEFs (Figure S8) and G2019S LRRK2 MEFs (Figure 12). However, the LRRK2 dephosphorylation and Rab10 dephosphorylation are not completely rescued by VH101, MLN4924, and MG132 pretreatments owing to the kinase inhibition effect of XL01126 as also evidenced in our kinase inhibition assay (Figure S5). ## XL01126 increases mitophagy in immortalized mouse embryonic fibroblasts cells With a potent, fast and selective LRRK2 degrader in hand, we next established XL01126 cellular functionality in bioassays that report on LRRK2 activity. Mitochondrial dysfunction is one of the pathophysiological hallmarks of PD 71 and can be rescued by mitophagy, a quality control mechanism whereby damaged or unnecessary mitochondria are delivered to lysosomes for degradation through membrane trafficking 8 . It has been shown that increasing mitophagy with inducer agents has the potential as a PD therapy 72 . Previous studies have shown that LRRK2 kinase activity impairs basal mitophagy and that LRRK2 knockout or pharmacological inhibition of LRRK2 with kinase inhibitors was able to rescue the mitophagy level 8 . Utilizing XL01126 as a chemical degrader tool and using cis-XL01126 as a non-degrader, kinase inhibitor control, we found that both XL01126 and cis-XL01126 induced mitophagy level dose-dependently (Figure 13) in mito-QC MEFs, a mCherry-GFP-mitochondria reporter cell model developed previously 73 . Although XL01126 and cis-XL01126 act on LRRK2 through different mechanisms, they shared similar potency in inducing mitophagy at 10-100 nM, indicating that mitophagy level is indeed LRRK2 kinase-dependent, and that other domains or motifs of LRRK2 are not involved in regulating mitophagy. ## XL01126 is orally bioavailable and can penetrate blood brain barrier To qualify XL01126 as both a cellular and in vivo suitable degrader probe, and to assess its drug development potential, we next evaluated the physicochemical and ADME properties (and Figure S9), as well as the in vivo pharmacokinetic profiles of XL01126 (Figure 14). Due to the high molecular weight and lipophilicity, XL01126 has low solubility in PBS and moderate solubility in Fed State Simulated Intestinal Fluid (FeSSIF) (Table 2), which, however, are all well above its DC50 values (14 -72 nM). The high stability (half-life at 108.29 min) of XL01126 in mouse plasma indicates XL01126 might be suitable for in vivo studies and we reasoned that plasma protein binding may account for its stability as protein binding can decrease the amount of free compound available for enzymatic metabolism. The protein binding also affects the potency of XL01126 in cells as shown by the significant potency shift of XL01126 in MEFs in the presence and absence of 10% fetal bovine serum (FBS) in the culture media (Figure S10). To further qualify XL01126 as appropriate for in vivo studies, we assessed its PK profiles in mice (Figure 14 and Table 3). Following a single dose of XL01126 via intravenous (IV, 5 mg/Kg), intraperitoneal (IP, 30 mg/Kg), and oral gavage (PO, 30 mg/Kg), the concentrations of XL01126 in plasma, brain tissue, and cerebrospinal fluid (CSF) were determined. XL01126 showed fast absorption in both IP and PO injection with Cmax (7700 ng/mL and 3620 ng/mL for IP and PO separately) reached at 0.25 min and 2h for IP and PO dosing, respectively. High plasma concentrations were achieved in all routes of administration, and were maintained at levels way above the DC50 values for XL01126 in the experimental time period. The metabolism of XL01126 seems slow in all administration routes, probably because of high protein binding. Strikingly, XL01126 was also detected in brain tissues and CSF (Figure 14B and 14C), suggesting that XL01126 is capable of penetrating the BBB regardless of its unfavorable in vitro ADME properties and violation of Ro5 and/or RoCNS 74 . To the best of our knowledge, this is the first-time report of a VHL-based PROTAC that is both oral bioavailable (F=15%) and BBB permeable. Further investigation of XL01126 will focus on its in vivo pharmacodynamics and PD-related functional studies. ## Discussion Although LRRK2 is a sought-after target for PD, the exact signaling pathways that link LRRK2 with PD pathology are unknown. LRRK2 is a large (286 KD), multi-domain protein that has two enzymatic domains and several other moieties involved in protein-protein interactions. However, LRRK2 kinase inhibitors are the most frequently used, if not the only, pharmacological tools for the study of LRRK2 biology, leaving the GTPase domain and protein-protein interaction domains of LRRK2 underexamined. The fast, potent, and selective LRRK2 degrader that we have developed and characterized in this study offers a new chemical tool for deciphering the biology of LRRK2. Employing the target protein degradation strategy to treat neurodegenerative disease can be revolutionary as protein aggregates are among the major pathologies and many attempts to modulate these diseases with conventional smallmolecule drugs have not been successful. Significant effort has already been made to target neurodegenerative disease related proteins with either peptide-based or small molecular PROTAC degraders 75 . However, achieving favorable PK profiles with oral bioavailability and BBB penetration have been the major obstacles for central nervous system (CNS)targeted PROTACs. Amongst the only successes reported to date, Wang et al. developed a tau-targeting PROTAC (C004019) that can penetrate the BBB after subcutaneous injection and induce tau protein degradation in the brain. 76 Herein, we disclose the identification of a LRRK2-targeting PROTAC that exhibits remarkable oral bioavailability and BBB penetration. Both CC004019 and XL01126 are VHLbased PROTACs with multiple violations of Ro5 and/or RoCNS. Their capability of penetrating the BBB challenges the Ro5-and RoCNS-based pre-conceptions and dogma and has expanded the chemical space of CNS targeting drugs. PROTAC is an emerging drug discovery modality, yet the development of an active and efficient degrader is still a laborious and unguided process. Structure-guided PROTAC design 62,66 is an attractive strategy, but solving the crystal structure of a target protein:PROTAC:E3 ligase ternary complex is a challenging feat. The step-by-step PROTAC development strategy we used here provides an empirical and generalized roadmap for developing PROTACs against LRRK2 and other challenging targets. The ternary binding affinity assay and ternary complex formation assay we developed here successfully circumvented the use of recombinant full-length LRRK2 protein which is challenging to express and purify. These two assays can potentially be applied to PROTAC or molecular glue development for other challenging targets as well. Further optimization of XL01126 and related LRRK2 degraders may result in compounds that exhibit improved activity or drug-like properties, improved selectivity for a particular LRRK2 mutant, decreased off-target degradation to PDE6D, and improved cooperativity, allowing further enhancement of the degradation vs inhibition window to achieve enhanced therapeutic performance. ## Chemistry Chemicals that are commercially available were purchased from Apollo Scientific, Sigma-Aldrich, Fluorochem, and Enamine and were used without further purification. All solvents use for reactions are anhydrous. LC-MS was carried out on Shimadzu HPLC/MS 2020 equipped with a Hypersil Gold column (1.9 μm 50 × 2.1 mm), photodiode array detector and ESI detector. The samples were eluted with a 3 min gradient of 5-95% acetonitrile in water containing 0.1% formic acid at a flow rate of 0.8 mL/min. Flash column chromatography was performed on Teledyne ISCO Combiflash Companion installed with disposable normal phase RediSep Rf columns (230-400 mesh, 40-63 mm; SiliCycle). Preparative HPLC purification was performed on Gilson Preparative HPLC system equipped with a Waters X-Bridge C18 column (100 mm × 19 mm and 5 μm particle size) using a gradient from 5 to 95% of acetonitrile in water containing 0.1% formic acid over 10 min at a flow rate of 25 mL/min. Compound characterization using NMR was performed either on a Bruker 500 Ultra shield or on a Bruker Ascend 400 spectrometer. The 1 H NMR, 13 C NMR and 19 F NMR reference solvents used are CDCl3 -d1 (δH = 7.26 ppm/δC = 77.16 ppm), CD3OD-d4 (δH = 3.31 ppm/δC = 49.00 ppm) or DMSO-d6 (δH = 2.50 ppm/δC = 39.52 ppm). Signal patterns are described as singlet (s), doublet (d), triplet (t), quartet (q), quintet (quint.), multiplet (m), broad (br), or a combination of the listed splitting patterns. The coupling constants (J) are measured in hertz (Hz). HRMS was performed on a Bruker MicroTOF II focus ESI Mass Spectrometer connected in parallel to a Dionex Ultimate 3000 RSLC system with diode array detector and a Waters XBridge C18 column (50 mm × 2.1, 3.5 µm particle size). All final compounds are >95% pure by HPLC. ## tert-butyl (2S,4S)-4-hydroxy-2-((4-(4-methylthiazol-5yl)benzyl)carbamoyl)pyrrolidine-1-carboxylate (2) To a solution of compound 1 51 (1.2 g, 3.94 mmol) in DCM (7.9 mL) was added 4N HCl in 1,4-dioxane (7.9 mL). After stirring at room temperature overnight, the mixture was concentrated under reduced pressure, washed with ethyl ether and dried to give a light yellow solid (902 mg, 95% yield). To a suspension of the solid (500 mg, 2.08 mmol) in DCM (10 mL) was added TEA (0.962 mL), (2S,4S)-1-(tertbutoxycarbonyl)-4-hydroxypyrrolidine-2-carboxylic acid (480 mg, 2.08 mmol), and HATU (830 mg, 2.18 mmol). After stirring at room temperature overnight, the mixture was diluted with DCM, washed with water and brine, dried over sodium sulfate, filtered, and condensed to afford a residue which was purified via flash column chromatography on silica gel (0-10% methanol in DCM) to give compound 2 as a solid (560 mg, 65% yield). To a solution of compound 2 (568 mg, 1.36 mmol) in DCM (6.8 mL) was added 4N HCl in 1,4-dioxane (6.8 ml). The resulting mixture was stirred at room temperature overnight and condensed to afford a solid (530 mg, 100% yield). To a solution of the obtained solid (200 mg, 0.57 mmol) and TEA (236 µL, 1.70 mmol) in DMF (5 ml) was added dropwise with a mixture of Fmoc-S-trityl-L-penicillamine (329 mg, 0.54 mmol), HATU (215 mg, 0.57 mmol) and TEA (79 µL, 0.57 mmol) in DMF (5 mL). After stirring at room temperature overnight, the mixture was diluted with DCM, washed with water and brine, dried over sodium sulfate, filtered and condensed to afford a residue which was purified with flash column to afford a residue as an amine compound 3 (120 mg, 32% yield for two steps, LC-MS, ESI -, 689.4 [M-H] -) which was used to the next step. To a solution of the amine compound 3 (60 mg, 0.087 mmol) in DMF (1.5 ml) was added TEA (24 µL, 0.174 mmol), HATU (35 mg, 0.092 mmol), and 1-fluorocyclopropane-1-carboxylic acid (9 mg, 0.087 mmol) separately. After stirring at room temperature for 4h, the resulting mixture was diluted with ethyl acetate, and washed with water and brine, dried over sodium sulfate, filtered and condensed to afford crude product which was purified via flash column chromatography on silica gel to give 4 (57 mg, 85% yield) as white solid. 3.34 (dd, J = 11.0, 4.0 Hz, 1H), 3.22 (d, J = 10.9 Hz, 1H), 2.43 (s, 3H), 2.12 (d, J = 14.0 Hz, 1H), 2.08 -1.97 (m, 1H), 1.30 -1.12 (m, 4H), 1.07 (s, 3H), 1.01 (s, 3H). 13 To a solution of compound 4 (57 mg, 0.073 mmol) in DCM (1.6 mL) was added triisopropylsilane (0.08 mL) and TFA (0.08 mL) at 0 °C. The resulting mixture was stirred at 0 °C for 30 min and condensed to afford a residue which was purified through flash column chromatography (0-10% methanol in DCM) on silica gel to yield compound 5 (36 mg, 92% yield). ## 4-((5-chloro-4-(methylamino)pyrimidin-2-yl)amino)-3methoxybenzoic acid (7) To a solution of 2,5-dichloro-N-methylpyrimidin-4-amine 17 (4.39g, 24.65 mmol) in a mixture of dioxane and water (70 ml :70 ml) was added 4-amino-3-methoxybenzoic acid (4.13 g, 24.70 mmol) followed by 4N solution of HCl in dioxane (6.18 ml, 24.72 mmol) at room temperature. After refluxing the reaction mixture at 100°C overnight, the mixture was cooled down to precipitate white solid. The solids were filtered, washed with water, dried under vacuum to afford compound 7 as white solid (5.95 g, 19.32 mmol, 78% yield). ## Tert-butyl 4-(4-((5-chloro-4-(methylamino)pyrimidin-2yl)amino)-3-methoxybenzoyl)piperazine-1-carboxylate (8) To a solution of 7 (2.1 g, 6.08 mmol) in DMF (25 mL) was added HOBt (0.98 g, 7.29 mmol), EDCI (1.39 g, 7.29 mmol), 1-Boc-piperazine (1.19, 6.38 mmol), and DIPEA (4.23 mL, 24.33 mmol) separately at room temperature. The mixture was stirred at room temperature for 16 h, then diluted with water (50 mL) and extracted with EtOAc (200 mL). The organic layer was washed with water and brine, dried over sodium sulfate, and concentrated to give a residue which was purified by flash column chromatography on silica gel (0% to 100% of EtOAc in DCM) to give compound 8 as white solid (2.52 g, 5.28 mmol, 87%). 1 To a solution of 8 (2.52 g, 5.28 mmol) in a mixture of DCM and MeOH 9:1 (30 ml) was added 4N solution of HCl in dioxane (5.28 ml, 21.12 mmol) at room temperature. After stirring at room temperature overnight, the mixture was diluted with Et2O (200 ml) to precipitate a solid which was filtered, washed with Et2O (100 ml) and dried overnight to give Boc-deprotected product (2.13 g, 5.17 mmol, 98% yield) as a HCl salt. ] + . To a suspension of the solid (25 mg, 0.06 mmol) in acetone (3 mL) was added K2CO3 (42 mg, 0.30 mmol) and trans-1,4-bis(bromomethyl)cyclohexane (50 mg, 0.185 mmol) (see Scheme S2 for synthesis). After stirring at 50 °C for 2 days, the mixture was diluted with DCM, washed with water and brine, dried over sodium sulfate, filtered, and condensed to afford a residue which was purified with flash column chromatography on silica gel to give compound 9 (10 mg, 29% yield). 2S,4S)-1-((R)-3-((((1R,4R)-4-((4-(4-((5-chloro-4-( To a solution of compound 9 (13 mg, 0.023 mmol) in THF (1.5 mL) was added compound 5 (10 mg, 0.019 mmol) and DBU (0.016 mL, 0.11 mmol). After stirring at room temperature overnight, the mixture was condensed and purified with preprative HPLC under acidic condition (5-95 % CH3CN in 0.1 % aq. HCO2H) to give cis-XL01126 (11.9 mg, 62% yield) as a white solid. 49, 148.71, 147.50, 137.38, 131.58, 131.42, 131.36, 129.77, 128.24, 120.26, 116.82, 109.65, 105.60, 79.21, 71.22, 65.42, 60.27, 58.63, 55.98, 55.89, 53.90, 47.56, 43.66, 38.44, 35.48, 34.98, 32.77, 32.66, 31.46, 28.23, 25.74, 25.34, 16.26, 13.92, 13. To a solution of compound 9 (8 mg, 0.014 mmol) in THF (1.5 mL) was added compound 10 77 (7.6 mg, 0.019 mmol) and DBU (0.012 mL, 0.085 mmol). After stirring at room temperature overnight, the mixture was condensed and purified with preparative HPLC under acidic condition (5-95 % CH3CN in 0.1 % aq. HCO2H) to give XL01126 (7.7 mg, 53% yield) as a white solid. 20, 158.73, 157.98, 152.82, 150.36, 148.69, 147.56, 138.18, 131.72, 131.45, 131.18, 129.66, 128.18, 120.30, 116.92, 109.74, 105.67, 78.4 (d, J = 261.9 Hz), 70.29, 65.41, 58.99, 56.67, 56.38, 56.02, 53.94, 47.66, 43.26, 38.54, 36.76, 35.41, 35.10, 32.83, 32.76, 31.48, 28.22, 25.79, 25.43, 16. Generation of mouse embryonic fibroblasts (MEFs) Primary MEFs were generated as described in a previous study 78 . Briefly, the uterine horn was collected from adult female mice at day E12.5 and transferred to a 10 cm tissue culture dish containing cold PBS. Two forceps were used to tear the yolk sacs to isolate each embryo. Forceps were cleaned thoroughly with 70% ethanol between each embryo isolation. The embryos were culled, and a tissue piece was collected in a PCR tube for genotyping. The red tissue of the embryo was removed, and the remainder was minced with a scalpel blade and incubated with 7.5 ml trypsin-EDTA solution for 10 minutes in a 37°C, 5% CO2 tissue culture incubator. The dish was removed from the incubator and checked under a light microscope for single cells. 7.5 ml complete media was added to the trypsinised cells and the cell suspension was transferred to a 15 ml Falcon tube, and centrifuged at 1200 rpm for 5 minutes at room temperature. The trypsin was aspirated, the cell pellet was resuspended in 5 ml fresh complete media, and the cell suspension was plated in a 60 mm tissue culture dish and incubated in a 37°C, 5% CO2 tissue culture incubator. The MEFs at this stage were considered as passage 0 and were passaged and expanded for experimental use once the genotype was confirmed by allelic sequencing and immunoblotting. MEFs were cultured in Dulbecco's modified Eagle medium (DMEM) containing 10% (v/v) fetal bovine serum (FBS), 2 mM L-glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin supplemented with 1X non-essential amino acids and 1 mM sodium pyruvate. ## Generation of bone marrow-derived macrophages (BMDMs) Macrophages were cultured in complete media containing DMEM, 10% (v/v) heat inactive FBS, 20% (v/v) L929 preconditioned medium, 2.5% (v/v) HEPES, 2 mM L-glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin, 2% sterile-filtered β-mercaptoethanol, 1X non-essential amino acids and 1 mM sodium pyruvate. Bone marrow isolation and macrophage differentiation was modified from 79 , employing L929 preconditioned medium as the source of M-CSF for differentiation. Briefly, scissors and forceps were used to dissect femurs and tibiae from adult mice, and muscle tissue was carefully removed from bones. Clean femurs and tibiae were placed in a tissue culture dish containing complete media. The ends of each bone were cut with scissors to expose bone marrow. Bone marrow was flushed with a 25gauge needle attached to a 10 ml syringe containing complete media. Media containing bone marrow was passed through a 70 µm cell strainer and precursor cells were plated on non-tissue culture treated 10 cm bacteriological plates containing 10 ml complete media. This was marked as day 0 of isolation. On day three post-isolation, macrophages were topped up with 5 ml fresh complete media. On day seven post-isolation, macrophages were rinsed once with PBS and incubated with versene for 5 minutes in a 37°C 5% CO2 tissue culture incubator. Macrophages were detached with cell scrapers and were centrifuged at 1200 rpm for 5 minutes at room temperature. The versene was aspirated and the remaining cell pellet was resuspended in complete media. The cell suspension was counted, and cells were seeded for experimental analysis in a 6-well format, in tissue culture treated dishes at a final cell density of one million cells per 6-well. PBMC cells separation and treatment PBMC cells were separated from human blood from healthy volunteer donors following existing protocol 80 and pelleted by centrifugation at 1000 g for 2 min. The supernatant was discarded and the PBMC pellet was resuspended in PBS containing 2% FBS for washing. The suspension was centrifuged at 1000 g for 2min again and the PBMC pellet was resuspended in RPMI-1640 (Gibco) media supplemented with 10% FBS. The cells were then seeded into 6-well plates and treated with testing compounds at indicated concentrations and time period. After treatment, the cells were collected into 2-ml eppendorf tube and centrifuged at 500g for 2 min to pellet the cells, the supernatant was discarded, and the pellet was resuspended in 1 ml PBS and centrifuged at 500 g for 2min again. The PBMC pellet was lysed with 60 µL of lysis buffer containing 50 mM Tris-HCl, pH 7.5, 1% (v/v) Triton X-100, 1 mM EGTA, 1mM sodium orthovanadate, 50 mM NaF, 0.1% (v/v) 2-mercaptoethanol, 10 mM 2-glycerophosphate, 5 mM sodium pyrophosphate, 0.1 µg/ml mycrocystin-LR (Enzo Life Sciences), 270 mM sucrose, 0.5 mM DIFP (Sigma, Cat# D0879) in addition to complete EDTAfree protease inhibitor cocktail (Sigma-Aldrich Cat # 11836170001). DIFP is highly toxic and must be prepared in a fume hood to a stock solution of 0.5 M in isopropanol. The lysed cells were then centrifuged at 1500 g for 15 min at 0 °C. The supernatants were collected for analysis by quantitative immunoblotting. For long term storage, the supernatant was flash frozen and stored at -80°C. Protein concentrations of cell lysates were determined using Pierce™ BCA Protein Assay Kit (ThermoFisher). Cell culture, treatment, and lysis Culturing and passaging of adherent cell lines were carried out using aseptic technique in CL1 or CL2 (for PBMC isolation) biological safety cabinets. All cells were incubated in a 37°C incubator with 5% CO2. Cell lines were regularly tested for mycoplasma contamination. For western blot assay, the cells were seeded in 6-well plates. Cells were treated with the indicated compounds such that the final concentration of DMSO was 0.1%. Following the treatment of cells with compounds at indicated concentrations and time periods, the media was removed and the cells were washed with PBS and lysed in 100 µl ice-cold complete lysis buffer containing 50 mM Tris-HCl pH 7.4, 1 mM EGTA, 10 mM 2-glycerophosphate, 50 mM sodium fluoride, 5 mM sodium pyrophosphate, 270 mM sucrose, supplemented with 1 μg/ml microcystin-LR, 1 mM sodium orthovanadate, complete EDTAfree protease inhibitor cocktail (Roche) and 1% (v/v) Triton X-100. The cells were immediately placed on ice and were scraped and collected into 1.5 ml Eppendorf tubes. Cell lysates were incubated on ice for 10 minutes prior to centrifugation at 15,000 g at 4°C for 15 minutes. The cell pellet was discarded, and supernatant was collected for analysis by quantitative immunoblotting. For long term storage, the supernatant was flash frozen and stored at -80°C. Protein concentrations of cell lysates were determined using the Bradford assay. All experiments with human peripheral blood were performed in guidance with local standard operating procedures, in line with the Human Tissue Act 81 and good clinical practice 82 for research. Non-clinical local ethical approval was in place and donors gave written informed consent. ## Quantitative immunoblotting Cell lysates containing a quarter of a volume of 4X NuPAGE LDS sample buffer (NP0007) supplemented with 5% β-mercaptoethanol, were heated at 95°C for 5 minutes. 15 to 20 μg of samples were loaded onto pre-cast 4-12% Bis-Tris midi 20W or 26W gels (Thermofisher Scientific, Cat# WG1402BOX or WG1403BOX) and resolved at 130 V for 2 hours with NuPAGE MOPS SDS running buffer (Thermofisher Scientific Cat# NP0001-02). Proteins were electrophoretically transferred onto a 0.45 µm nitrocellulose membrane (GE Healthcare, Amersham Protran Supported 0.45 mm NC) at 90 V for 90 min on ice in transfer buffer (48 mM Tris base and 39 mM glycine supplemented with 20% methanol). The transferred membrane was blocked with 5% (w/v) skim milk powder dissolved in trisbuffered saline with tween (TBS-T) (50 mM Tris base, 150 mM sodium chloride (NaCl), 0.1% (v/v) Tween-20) at room temperature for 1 hour. Membranes were washed three times with TBS-T and were incubated in primary antibody overnight at 4°C. Prior to secondary antibody incubation, membranes were washed three times for 15 minutes with TBS-T. The membranes were incubated with secondary antibody for one hour at room temperature, protected from light. Thereafter, the membranes were washed with TBS-T three times with a 15-minute incubation for each wash, and protein bands were acquired via near-infrared fluorescent detection using the Odyssey CLx imaging system and quantified using Image Studio software. Graphs were generated using Graphpad Prism version 8 software. ## Antibodies Monoclonal rabbit LRRK2 Ser935 (Cat# UDD2) was purified by MRC PPU Reagents and Services at the University of Dundee and was used at a final concentration of 1 μg/ml. Total LRRK2 (C-terminus) was from Antibodies Inc./Neuromab (Cat# 75-253) and was diluted 1:1000. The MJFF monoclonal rabbit Rab10 pThr73, which was characterized previously 83 was purchased from Abcam Inc. (ab230261) and diluted 1:1000. Mouse monoclonal alpha-tubulin (#3873) was purchased from Cell Signaling Technology and used at 1:1000. The mouse monoclonal anti-Rab10 total antibody was purchased from Nanotools (#0680-100/Rab10-605B11) and was used at a final concentration of 1 μg/ml. Mouse monoclonal Hif-1α was purchased from R&D Systems (Cat# MAB1536) and was diluted 1:1000. Mouse monoclonal Ubiquitin was purchased from Biolegend (Cat# 646302) and was diluted 1:1000. Rabbit polyclonal PDE6D antibody was purchased from Novus Biologicals and was used at a final concentration of 1:500. All rabbit and mouse primary antibodies were diluted in 5% (w/v) bovine serum albumin (BSA) dissolved in TBS-T (50 mM Tris base, 150 mM sodium chloride (NaCl), 0.1% (v/v) Tween 20). Goat anti-mouse IRDye 800CW (#926-32210), goat anti-mouse IRDye 680LT (#926-68020), goat anti-rabbit IRDye 800CW (#926-32211), and goat anti-rabbit IRDye 680LT (#926-68021) IgG (H+L) secondary antibodies were from LI-COR and were diluted 1:10,000 in 5% (w/v) milk in TBS-T. Total proteome sample preparation and MS analysis Wildtype MEFs were seeded in 10 cm tissue culture dishes, at a density of two million cells per dish. Cells were treated with 0.1% DMSO, 300 nM XL01126, or 300 nM cis-XL01126 for 4 hours prior to harvest in 400 µl complete lysis buffer, supplemented with 1 μg/ml microcystin-LR, 1 mM sodium orthovanadate, complete EDTA-free protease inhibitor cocktail (Roche) and 1% (v/v) Triton X-100. Cell lysates were incubated on ice for 10 minutes, then underwent three rounds of high energy sonication for 15 cycles (30 seconds on, 30 seconds off) using the Diagenode Bioruptor. Cell lysates were centrifuged at 15,000 g at 4°C for 15 minutes. Cell pellet was discarded, and supernatant was collected for protein quantification using a BCA protein assay kit (Pierce #23225). 100 µg cell lysate was employed for total proteomic analysis. Proteins in cell lysate were reduced with 0.1 M Tris(2-carboxyethyl)phosphine (TCEP) diluted in 300 mM triethylammonium bicarbonate (TEABC) to a final concentration of 10 mM. Samples were incubated on a Thermomixer for 30 minutes at 60°C at 800 rpm then cooled down to room temperature and underwent alkylation with 0.04 M iodoacetamide (IAA) freshly dissolved in water. Samples were then incubated in the dark on a Thermomixer at room temperature for 30 minutes at 800 rpm. Alkylation was quenched with the addition of 0.1 M TCEP dissolved in 300 mM TEABC at a final concentration of 5 mM. Samples were incubated on a Thermomixer at room temperature for 20 minutes at 800 rpm. Sodium dodecyl sulfate (SDS) was added at a final concentration of 5% (w/v) from a 20% (w/v) stock. 12% (v/v) phosphoric acid was then added to a final concentration of 1.2% (v/v). Samples were diluted in 6 times the sample volume of S-trap wash buffer containing 90% (v/v) methanol diluted in 100 mM (v/v) TEAB pH 7.1. ## S trap cleanup and digestion Samples underwent S-trap cleanup to remove detergents and other impurities with S-trap mini columns (PROTIFI Cat# MSPPC02-MINI-80) placed in 2 ml Eppendorfs. The protein mixtures were added to columns and centrifuged briefly (1000 g / 1 minute / RT). Columns were washed with 400 µl S-trap buffer 4 times, centrifuging after each wash at 1000 g / 1 minute / RT. Columns were placed in fresh 2 ml Eppendorfs and 100 µl of 5 µg Trypsin/Lys-C freshly dissolved in 50 mM TEAB, pH 8.5 was added. Columns were centrifuged briefly (200 g / 1 minute / RT) and Trypsin/Lys-C mixture was pipetted back onto the column. 100 µl 50 mM TEAB, pH 8.5 was added directly to the 2 ml Eppendorfs to cover any digested peptides remaining in the tube. The S-trap columns in 2 ml Eppendorfs were incubated at 47°C without shaking for 1.5 hours, then at RT overnight. 80 µl 50 mM TEAB was added to S-trap columns, which were centrifuged, and eluates were collected in new 1.5 ml Eppendorf tubes. 80 µl 0.2% (v/v) formic acid was added to columns, which were centrifuged, and second eluates were pooled with first eluates. 80 µl 50% (v/v) acetonitrile diluted in 0.2% (v/v) formic acid was added to columns, which were centrifuged, and third eluates were pooled with previous eluates. 500 ng digested peptides were set aside to vacuum dry separately to verify that digestion efficiency by calculating the zero and single missed cleavages was >98%. The remaining peptides were divided in half (50 µg peptides each tube) and vacuum dried and stored in -80°C prior to continuation with tandem mass tag (TMT) labeling. TMT labeling 800 µg TMT mass tag reagents were dissolved 80 µl 100% (v/v) anhydrous acetonitrile to obtain final concentrations of 10 µg/µl. Resuspended TMT reagents were incubated at RT for 10 minutes, then vortexed and centrifuged briefly (2000 g / 2 minutes / RT). 50 µg lyophilized peptides were resuspended in 50 µl of a mixture containing 42 µl 50 mM TEAB and 8 µl 100% (v/v) anhydrous acetonitrile. Resuspended peptides were sonicated for 10 minutes, then centrifuged at 17,000 g for 10 minutes at RT. Peptides were transferred to fresh protein low-bind 1.5 ml Eppendorf tubes. 20 µl of 10 µg/µl TMT reagent were added to solubilized peptides, vortexed, centrifuged briefly (2,000 g / 1 minute / RT) and incubated on a Thermomixer for 2 hours at 800 rpm at RT. 50 µl of 50 mM TEAB was added to each reaction, followed by vortex, brief centrifugation (2,000 g / 1 minute / RT) and incubation on a Thermomixer at 800 rpm at RT for an additional 10 minutes. 5 µl of each TMT labeled sample was set aside, vacuum dried and injected on MS to confirm that labeling efficiency was >98%. The remaining reactions were stored in -80°C until labeling efficiency was verified. TMT samples were thawed to RT and labeling reactions were quenched with the addition of 5 µl 5% (v/v) hydroxylamine (dissolved in water from a 50% (v/v) stock solution). Samples were incubated on a Thermomixer for 20 minutes at 800 rpm at RT. Quenched TMT labeled samples were pooled, vacuum dried and subjected to High-pH fractionation as described previously 84 , 96 fractions were collected and were concatenated into 48 fraction. Pooled fractions were vacuum dried and stored in -20 freezer until the LC-MS/MS analysis. LC-MS/MS analysis: High-pH fractions were solubilized in 60 µl of LC-buffer (3% ACN (v/v), 0.2% Formic acid (v/v) by placing them on a Thermomixer at room temperature for 30 minutes with an agitation at 1800 rpm. 7 µl of each fraction was transferred into LC-vail inserts for mass spectrometry analysis. LC-MS/MS analysis was carried out on a Thermo Lumos ETD Tribrid mass spectrometer inline with 3000 ultimate RSLC nano-liquid chromatography system. Sample was injected into pre-column (C18, 5µm, 100Ao, 100µ, 2cm Nano-viper column # 164564, Thermo Scientific) at 5 µl/min flow rate and subsequently loaded onto the analytical column (C18, 5µm, 50cm, 100Ao Easy nano spray column # ES903, Thermo Scientific) for the separation of peptides using nano-pump operated at 300 nl/min flow rate. 85 min non-linear gradient was applied (5% Solvent B (80 %ACN v/v in 0.1% Formic acid v/v) to 22% B for 70 min and increased to 35% B for another 10 min for a total of 100 min run time. The eluted peptides were electrosprayed into the mass spectrometer using easy nano source. The data was acquired in a data dependent acquisition (DDA) mode in SPS MS3 (FT-IT-HCD-FT-HCD) method and was acquired using top speed for 2 sec for each duty cycle. The Full MS1 scan was acquired at 120,000 resolution at m/z 200 and analyzed using Ultra high filed Orbitrap mass analyzer in the scan range of 375-1500 m/z. The precursor ions for MS2 were isolated using Quadrupole mass filter at 0.7 Da isolation width and fragmented using normalized 35% Higherenergy collisional dissociation (HCD) of in Ion routing multipole analyzed using Ion trap. Top 10 MS2 fragment ions in a subsequent scan were isolated and fragmented using HCD at 65% normalized collision energy and analyzed using Orbitrap mass analyzer at 50,000 resolution in the scan range of 100-500 m/z. Database search and data analysis: Raw MS data of 48 High-pH fractions were searched using MaxQuant search algorithm (Vesion 2.0.3.0) 85 against Uniprot Mouse database (Release version May 20021 containing 25,375 sequences). 10 plex TMT reporter ion MS3 workflow was loaded and used following search parameters. Trypsin as a protease was selected by allowing two missed cleavages, deamidation of Asn and Gln; Oxidation of Met were used as variable modifications and Carbamidomethylation of Cys as a fixed modification. The default mass error tolerance for MS1 and MS2 (4 ppm and 20 ppm) were used. Min of 2 unique+razor peptides were selected for the quantification. The data was filtered for 1% PSM, peptide and protein level FDR. The output protein group .txt files were further processed using the companion Perseus software suite (version 1.6.15.0) 86 . Decoy hits, contaminants, proteins identified by sites and single peptide hits were filtered out. The data was then log2 transformed and T-test was performed between the sample groups and the p-values were corrected using 5% permutation-based FDR to identify the differentially regulated protein groups. Fluorescence polarization assay FP competitive binding assays were performed following the method described previously 49,57 . All the measurements were taken on a PHERAstar (BMG LABTECH) plate reader installed with a FP filter that sets excitation and emission wavelengths at 485 nm and 520 nM separately. Each well of 384-well plate (Coring 3575) contains 10 nM VCB protein, 5 nM FAM-lableled HIF-1α peptide (FAM-DEALAHypYIP-MDDDFQLRSF, "JC9"), and decreasing concentrations of testing compounds (14 concentrations with 2-fold serial dilution starting from 250 µM) in FP assay buffer (100 mM Bis-Tris propane, 100 mM NaCl, 1 mM TCEP, pH 7) with a final DMSO concentration of 5%. The control wells containing the VCB and JC9 with no compound are set as the maximum signals (zero displacement). And the control wells containing JC9 in the absence of protein are set as the minimum signals. Control values were used to obtain the percentage of displacement which was ploted against Log [Compound]. Average IC50 values were determined for each titration using nonlinear regression analysis with GraphPad Prism (v.9.3.1). The Ki values were back-calculated from the Kd of JC9 (1.5 nM -3.4 nM) and the fitted IC50 values, as described previously 87,88 . NanoBRET target engagement assay For VHL and LRRK2 target engagement experiments in live and permeabilized cells, the HEK293 cells were transfected with VHL-NanoLuc fusion vector (Promega, N275A) or LRRK2-NanoLuc fusion vector (Promega, NV3401) following Promega's protocol and seeded into white 384-well plate (Corning3570) at a density of 6000 cells/well. To measure NanoBRET in permeabilized cells, the cells were treated with 50 µg/ ml digitonin (Sigma, D141), 125 nM VHL tracer/125 nM LRRK2 tracer, testing compounds at decreasing concentrations (12 concentrations with 2-fold serial dilution starting from 33 µM), and NanoBRET NanoGlo Substrate (Promega) at concentration recommended by the manufacturer's protocol. In the maximum signal control samples (DMSO control), DMSO was added instead of testing compounds. In the minimum signal control samples (no tracer control), DMSO and tracer dilution buffer were used to replace testing compounds and tracer separately. The filtered luminescence was measured within 10 min following addition of the substrate on a GloMax Discover microplate reader (Promega) or a PHERAstar (BMG LABTECH) plate reader equipped with a 450-nm bandpass filter (donor) and a 600-nm long pass filter (acceptor). To measure NanoBRET in live cells, the cells were treated with 250 nM VHL tracer/500 nM LRRK2 tracer, testing compounds at testing compounds at decreasing concentrations (12 concentrations with 2-fold serial dilution starting from 33 µM) and incubated at 37 °C in an incubator for 2 h. The plates were then cooled down and added with NanoBRET NanoGlo Substrate and Extracellular NanoLuc Inhibitor (Promega, N2160) before performing the same NanoBRET reading as the permeabilized mode on plate readers. NanoBRET ratio of each well was expressed in milliBRET according to the equation: mBRET = [(signal at 610 nM/signal at 450 nM) -(signal at 610 nMno tracer control/signal at 450 nMno tracer control)]×1000. The fractional occupancy was calculated according to the equation: fractional occupancy = (mBRETtesting compound -mBRETno tracer control)/(mBRETDMSO control -mBRETno tracer control). ## NanoBRET-based ternary binding and cooperativity assay The HEK293 cells were transfected with LRRK2-NanoLuc fusion vector (Promega, NV3401) following Promega's protocol and seeded into white 384-well plate (Corning3570) at a density of 6000 cells/well. The cells were then treated with 50 µg/ml of digitonin, 125 nM of LRRK2 tracer, testing compounds at decreasing concentrations (11 concentrations with 2-fold serial dilution starting from 10 µM) or testing compounds and VCB mix (11 concentrations with 2-fold serial dilution starting from 10 µM compound for the compound. The first 6 concentrations of VCB start from 32 µM with 2-fold dilution, the last 5 concentrations of VCB keep at 1 µM), and NanoBRET NanoGlo Substrate (Promega) at concentration recommended by the manufacturer's protocol. In the maximum signal control samples (DMSO control), DMSO was added instead of testing compounds. In the minimum signal control samples (no tracer control), DMSO and LRRK2 tracer dilution buffer were used to replace testing compounds and LRRK2 tracer separately. The filtered luminescence was measured within 10 min following addition of the substrate on a PHERAstar (BMG LABTECH) plate reader equipped with a 450-nm bandpass filter (donor) and a 600nm long pass filter (acceptor). The fractional occupancy was calculated according to the equation: fractional occupancy = (mBRETtesting compound -mBRETno tracer control)/(mBRETDMSO control -mBRETno tracer control). Bodipy 576/589 labeling of VCB VCB was labeled with Bodipy 576/589 following protocol reported previously 89 . Briefly, the VCB complex was mixed with Bodipy 576/589 NHS ester in a 20:1 molar ration and incubated at room temperature (protect from light) for 2h in reaction buffer (0.1 M sodium phosphate, 75 mM KOAc, 2 mM DTT, pH 7.4). The reaction was quenched by diluting 10 times with reaction buffer and the unreacted dye was removed with a PD-10 MiniTrap desalting column (GE Healthcare) equilibrated with 100 mM Bis-Tris pH 7.0, 100 mM NaCl, 1 mM DTT, pH 7. The eluted labeled protein solution was collected and concentrated with Pierce Concentrator, 3K MWCO (Thermo scientific). NanoBRET Ternary complex formation assay HEK 293 cells were transfected with LRRK2-NanoLuc vector (Promega, NV3401) for 24h, harvested, and resuspended into OptiMEM media without phenol red (Life Technologies). The cells were then seeded into white 384-well plate (Corning3570) at a density of 6000 cells/well. Digitonin solution (final concentration 50 µg/ml), testing PROTACs at decreasing concentrations (11 concentrations with 2-fold serial dilution starting from 33 µM) or DMSO, and VCB protein labeled with Bodipy 576/589 (final concentration 0.5 µM) were added separately. Each well was added with NanoBRET NanoGlo Substrate (Promega, N2160) before performing the NanoBRET reading on PHERAstar (BMG LABTECH) plate reader equipped with a 450-nm bandpass filter (donor) and a 600-nm long pass filter (acceptor). NanoBRET ratio of each well was expressed in milliBRET according to the equation: mBRET = [(signal at 610 nM/signal at 450 nM) -(signal at 610 nMno tracer control/signal at 450 nMno tracer control)]×1000. The background signal as shown in the DMSO control samples was subtracted from each sample. Evaluation of mitophagy in immortalised mito-QC MEFs Immortalised mito-QC MEFs 90,91 were maintained in DMEM (Gibco, 11960-044) supplemented with 10% FBS, 2 mM L-Glutamine (Gibco, 2503-081), 1% Na-Pyruvate (Gibco, 11360-070), 1% Non-essential amino acids (Gibco, 11140-035), 1% Antibiotics (Penicillin/Streptomycin 100 U/ml penicillin and 100 μg/ml streptomycin; Gibco), at 37°C under a humidified 5% CO2 atmosphere. To assess mitophagy, MEFs were plated on #1.5 glass coverslips (Epredia, CB00130RAC20MNZ0) and treated for 24 h with XL01126 cis-XL01126, MLi-2 (positive control), or DMSO (vehicle) 91,92 . All treatments were in DMEM (Gibco, 11960-044) supplemented with 10% FBS, 2 mM L-Glutamine (Gibco, 2503-081), 1% non-essential amino acids (Gibco, 11140-035), 1% Antibiotics (Penicillin/Streptomycin 100 U/ml penicillin and 100 μg/ml streptomycin; Gibco) at 37°C under a humidified 5% CO2 atmosphere. MLi-2 was synthesized by Natalia Shpiro (University of Dundee) as previously described 92 . At the end of the treatment, cells were washed twice with DPBS (Gibco, 14190-094), and fixed with 3.7% Paraformaldehyde (Sigma, P6148), 200 mM HEPES, pH=7.00 for 20 min. Cells were washed twice with DPBS, and then incubated for 10 min with DMEM, 10 mM HEPES. After a wash with DPBS, coverslips were mounted on a slide (VWR, Superfrost, 631-0909) with Prolong Glass (Thermo Fisher Scientific, P36984). Images were acquired using a Zeiss LSM880 with Airyscan laser scanning confocal microscope (Plan-Apochromat 63x/1.4 Oil DIC M27) using the optimal parameters for acquisition (Nyquist). 3-5 biological replicates were performed for each experiment with 10 images acquired per condition (124-260 cells per condition). Quantification of red-only dots was semi-automatised using the mito-QC counter plugin on FIJI as previously described 91,93 . One-way ANOVA with Dunnett's multiple comparisons were performed using GraphPad Prism version 9.3.1. p-values are represented as *p < 0.05. Error bars denote SEM. Caco-2 cell permeability Caco-2 cells with Transepithelial electrical resistance (TEER) (TEER= (Resistance sample -Resistance blank)×Effective Membrane Area) = 450 ± 19 Ω•cm2 Ω•cm 2 were used for the experiment. Compounds were dissolved in appropriate buffer (10 mM DMSO stock solutions were diluted with HBSS buffer to a final concentration of 10 µM testing compound and 0.4% DMSO, Lucifer Yellow was introduced in the apical side buffer to test the intactness of the mono-cell layer) and was applied to the apical or basolateral donor side for measuring A-B or B-A permeability (two replicates), respectively. The apical and basolateral plates were prewarmed to 37 °C before placing the apical plate onto basolateral plate. After incubating at 37 °C for 90 min, the apical plate and basolateral plate were separated, and the donor or receiver samples were analyzed with UPLC-MS/MS. Plasma stability assay Frozen plasma was thawed at 37 °C and centrifuged at 3000 rpm for 8 min to remove clots and the supernatant was used in the experiment. The pH of the plasma was recorded and only pH range between 7.4 and 8 was used. The plasma and compounds solution was pre-warmed to 37 °C. 10 µl prewarmed testing compound or reference compound (procaine) solution (20 µM in 0.05 mM sodium phosphate buffer (pH7.4) with 0.5% BSA) was mixed with 90 µl of plasma at different time points to allow for 5, 15, 30, 45, and 60 min of incubation time. For 0 min, the plasma was mixed with vehicle only. Acetonitrile was added to the compound and plasma mixture to quench the reaction and the resulting mixture was centrifuged (5594 g for 15 min). The supernatant was taken and diluted before LC-MS analysis. Solubility in Phosphate buffer and Fed State Simulated Intestinal Fluid (FeSSIF) 8 µl of reference or test compound stock solution (10 mM in DMSO) was added into 792 µl of 100 mM phosphate buffer (pH 7.4) or FeSSIF (pH 5.8). The resulting mixture was shaken for 1h (1000 rpm) at room temperature, then centrifuged for 10 min (12000 rpm) to remove un-dissolved particles. The supernatant was collected and diluted 10 times and 100 times separately with 100 mM phosphate buffer or FeSSIF. 5 µl of the supernatant samples (no diluted, 10 times diluted, 100 times diluted) were mixed with 95 µl of acetonitrile (containing internal standard) separately before injecting into LC-MS/MS for analysis. Mouse liver microsome stability 1.5 µl testing compound or reference compound (500 µM in 5% DMSO and 95% acetonitrile) was mixed with 18.75 µl of 20 mg/mL liver microsome (Corning) and 479.75 µl potassium phosphate buffer (0.1 M potassium phosphate buffer, 1 mM EDTA, pH 7.4). The reaction was started by mixing 30 µL of the above mixture (pre-warmed to 37 °C) with 15 µl of 6 mM NADPH stock solution (pre-warmed to 37 °C). After incubating for 5, 15, 30, or 45 min, 135 µl of acetonitrile containing internal standard was added to stop the reaction. For 0 min, the compound and microsome mixture was mixed with acetonitrile first before adding NADPH. After quench, the reaction mixture was centrifuged, and the supernatant was taken and diluted for LC-MS analysis. Mouse Hepatocyte stability 50 µl of pre-warmed hepatocytes (2 × 10 6 cells/ml) in suspension media (Krebs-Henseleit buffer (Sigma) containing 5.6 g/l HEPES) was mixed with 50 µl pre-warmed compound dosing solution (2 µM in Krebs-Henseleit buffer with 1% DMSO). After incubating at 37 °C for 15, 30, 60 or 120 min, 100 µl of acetonitrile containing internal standard was added to quench the reaction. For 0 min incubation, acetonitrile was mixed with hepatocytes first before adding compound solution. After quenching, the mixture was shaken at the vibrator for 10 min (600 rpm/min) and then sonicated for 2 min before centrifugation (5594 g for 15 min). The supernatant was taken and diluted for LC-MS analysis. Pharmacokinetic (PK) study PK profiling was outsourced and undertaken by Shanghai ChemPartner Co., Ltd. All animal experiments performed were conducted in compliance with the Institutional Animal Care and Use Committee (IACUC) and the Office of Laboratory Animal Welfare (OLAW) guidelines. Six-to eight-weekold C57BL/6 male mice purchased from Jihui Laboratory Animal Co. LTD were used in the study. XL01126 was formulated in 10%HP-β-CD in 50mM Citrate buffer pH=3.0 at 1 mg/ml for IV injection and at 3 mg/ml for IP and PO injection. For IV injections, 5 mg/kg of XL01126 was administered into the tail vein. For IP and PO injections, 30 mg/kg of XL01126 was administered via intraperitoneal injection or oral gavage, respectively. The animals were restrained manually at the designated time points (0.083, 0.25, 0.5, 1, 2, 4, and 8 h); approximately, 110 μl of blood sample was collected via facial vein into K2EDTA tubes. Three mice per time point were used, resulting in a total of 21 mice for each administration route. The blood sample was put on ice and centrifuged at 2000g for 5 min to obtain the plasma sample within 15 min. The plasma, brain, CSF samples were stored at approximately −70 °C until analysis. A 30 μl aliquot of plasma was added with 200 μl of internal standard (Glipozode, 40 ng/mL) in MeCN with 5% citri. The mixture was then vortexed for 1 min and then centrifuged for 10 min at 5800 rpm. The supernatant (100 μl) was transferred to a new plate. The solution (1 μl) was injected to LC-MS/MS. LC-MS/MS instrument used: SCIEX LC-MS/MS-49 (Triple Quad 6500+). Data was analyzed by WinNonLin and Microsoft Excel. ## Supporting Information. The Supporting Information is available free of charge at xxx. First generation compound structures, Synthetic procedures for the first and the second generation PROTACs, abbreviations used, figures, compounds characterizations ## AUTHOR INFORMATION Corresponding Authors

chemsum

{"title": "Discovery of XL01126: A Potent, Fast, Cooperative, Selective, Orally Bioavailable and Blood Brain Barrier Penetrant PROTAC Degrader of Leucine Rich Repeat Kinase 2 (LRRK2)", "journal": "ChemRxiv"}

elucidating_dissociation_activation_energies_in_host–guest_assemblies_featuring_fast_exchange_dynami

## Abstract: The ability to mediate the kinetic properties and dissociation activation energies (E a ) of bound guests by controlling the characteristics of "supramolecular lids" in host-guest molecular systems is essential for both their design and performance. While the synthesis of such systems is well advanced, the experimental quantification of their kinetic parameters, particularly in systems experiencing fast association and dissociation dynamics, has been very difficult or impossible with the established methods at hand. Here, we demonstrate the utility of the NMR-based guest exchange saturation transfer (GEST) approach for quantifying the dissociation exchange rates (k out ) and activation energy (E a,out ) in hostguest systems featuring fast dissociation dynamics. Our assessment of the effect of different monovalent cations on the extracted E a,out in cucurbit[7]uril:guest systems with very fast k out highlights their role as "supramolecular lids" in mediating a guest's dissociation E a . We envision that GEST could be further extended to study kinetic parameters in other supramolecular systems characterized by fast kinetic properties and to design novel switchable host-guest assemblies. ## Introduction For many supramolecular host-guest systems, elucidating their kinetic characteristics is critical for thoroughly understanding their performance and further improving their design as synthetic channels, 1 receptors, 2 transporters, 3 drug carriers, 4 catalysts, 5 stimuli-responsive materials 6,7 and more. Controlling the kinetic properties in such systems can be obtained through an external stimulus that changes the system's activation energy (E a ) so as to yield an "open" or "closed" state of the host and, thus, govern the exchange dynamics of the bound guest. Considerable advances have been made in the design of such switchable open/closed molecular hosts and their response to a variety of external stimuli, such as pH, 8 light, 9 heat, 10 redox 11 and more. However, a robust and accessible tool for studying their effect on the dissociation activation energy E a (E a,out ) in a quantitative manner, which is crucial for the further development and improved performance of such systems, has yet to be offered. Indeed, well-established classical methods, such as stopped-flow experiments, 12 UV-Vis measurements 13 and exchange spectroscopy (EXSY)-NMR, 14 are useful for characterizing slow dynamic processes. Nevertheless, these analytical tools are less favourable when it comes to supramolecular complexes with fast exchange dynamics, including the evaluation of such systems' E a values. Applying the chemical exchange saturation transfer (CEST) NMR method to the study of host-guest systems using a hyperpolarized 129 Xe gas guest 15 has opened new opportunities for quantifying exchange dynamics in supramolecular assemblies. Indeed, hyperCEST was applied to a wide array of molecular hosts, such as cryptophanes, 16,17 cucurbit[n]urils (CBn), pillar[n]arenes 21 and paramagnetic-capsules, 22 demonstrating its applicability to a range of exchange regimes that are dependent on the host properties. 23 The combination of CEST and 19 F-NMR 24,25 and its extension to host-guest systems have expanded the arsenal of molecular guests that are suitable for CEST-based studies beyond that of 129 Xe. This approach, termed guest exchange saturation transfer (GEST), 29 allowsas we introduce hereinnow also the use of conventional NMR-setups to quantitatively study host-guest dissociation rates (k out ) and E a,out . In fact, most host-guest systems that we are aware of display such low activation energies (and thus very fast complex formation and dissociation rate constants) that their study is simply infeasible by the available methods. In several cases, direct binding assays can be used in combination with stopped-flow experiments, particularly, when a chromophoric or emissive guest considerably alters their spectroscopic properties upon binding to the host. However, often in these cases a too fast complex formation is observed which is completed within the mixing time ("dead time") of the technique. Thus, alternative methods for quantifying kinetic parameters applicable to the study of fast equilibrating hostguest systems are still in need. We demonstrate here how the GEST-NMR method can be used to quantify relatively fast k out values, which makes it complementary to other tools, 35,36 including those based on NMR (i.e., EXSY), 37 which are better suited for studying slow exchange rates. By exploring the relationship between k out rates and the applied temperature, we demonstrate GEST's use as an analytical method for the study of E a,out in host-guest systems. Specifcally, we show that GEST-NMR can be used to quantitatively elucidate E a,out values of fluorinated guests (G) from cucurbit uril (CB7). The CB7 molecular host has a broad range of applications through host-guest inclusion complex formation, but also shows an unprecedented affinity to cations through ion-dipole interactions forming "supramolecular-lids" 47-51 that mediate both thermodynamic 52 and kinetic properties 31,53,54 of CB7:G systems. Herein, we demonstrate the capability of GEST-NMR to quantify the effect of cationic-CB7 "lids" on the E a,out values of fast-exchanging guests, thus establishing it as an accessible analytical tool for future kinetic studies in supramolecular systems. ## Results and discussion In the here presented study, we characterized the following three host-guest systems regarding their k out values by GEST-NMR: CB7 as host with halothane (G1), 5-fluorotryptophan (G2), and fluroxene (G3) as guests (Fig. 1). As a frst step, we acquired 19 F-NMR spectra for each system to classify them roughly as either slow or fast exchanging on the NMR timescale (Fig. 2a and S1 †). While CB7:G1 clearly exhibited the typical additional peak of a bound guest (upfeld shifted to that of free G1), the spectra of CB7:G2 and CB7:G3 featured only a single peak, assigned to the non-bound guest. The clear, sharp, distinct peaks in the CB7:G1 spectrum are typical for a relatively slow exchange regime on the NMR timescale. However, faster exchange processes, as in the cases of CB7:G2 and CB7:G3 (Fig. 2a), lead to NMR-line broadening and peak coalescence, which prevent one from distinguishing between free and bound guests in the 19 F-NMR spectra. To further elaborate on this observation, GEST experiments were carried out on solutions containing the studied CB7:G systems. When performing GEST-NMR experiments of a CB7:G1 complex at room temperature (298 K), a well-defned saturation transfer effect was observed at the frequency of the bound peak (Fig. 2a, left). Interestingly, in the equivalent GEST-NMR experiment of CB7:G2, we found a clear GEST effect (Fig. 2b, middle), marked by the lack of the characteristic CB7:G2 peak in the 19 F-NMR spectrum (Fig. 2a, middle). In the CB7:G3 solution at 298 K, there was no observable asymmetry in the GEST-spectrum (Fig. 2b, right). Nonetheless, the isothermal titration calorimetry (ITC) experiments indicated the formation of CB7:G3 with an association constant K a of 7 10 4 M 1 (Fig. S2 †), which indicates a system with fast exchange kinetics that is accompanied by symmetric GEST spectrum (Fig. 2b, right). In order to quantify the guest dissociation rates of each of the CB7:G systems, we set up a series of GEST experiments with varied pre-saturation pulses (B 1 , Fig. 2c). Fitting the experimental data to the Bloch-McConnell equations 55 allowed us to quantitatively evaluate the k out of the studied CB7:G systems. As expected, CB7:G1 (Fig. 2, left) exhibited relatively slow dissociation kinetics, with k out ¼ 15 AE 1 s 1 ; this k out value fts in the slow exchange rate regime on the NMR timescale, with a k out ( Du for a Du of 1.3 ppm (equal to 490 Hz at 9.4 T NMR) offset between free and bound G1 in the 19 F-NMR spectrum. Note that such a slow k out value could also be quantifed with the established EXSY-NMR method, 56 which, as noted above, is not applicable to host-guest systems with faster dissociation rates where two distinct NMR peaks are not detected (as shown for G2 and G3 in Fig. 2). Ideally suited for the study of faster exchange regimes, 57 GEST-NMR was used to quantify the k out of CB7:G2, found to be 2000 AE 100 s 1 (for Du of $1200 Hz; Fig. 2c, middle). Nevertheless, we were unable to determine the exchange rate by which G3 is excluded from its CB7:G3 complex (k out >4000 s 1 , 298 K, Fig. 2c, right), as a very broad z-spectrum was obtained (Fig. 2b, right). Thus, we can use GEST-NMR to differentiate between the kinetic regimes of each of the abovementioned host-guest systems -CB7:G1, CB7:G2 and CB7:G3, representing slow-, intermediate-, and fast-exchange processes on the NMR timescale at room temperature (298 K), respectively. For the elucidation of the binding mechanism, and thus for obtaining deeper insights into non-covalent interactions and supramolecular principles, the knowledge of the activation energies is of utmost beneft. Having identifed two CB7:G systems that experience intermediate-to-fast k out rates (>2000 s 1 ), we turned to evaluate the capability of GEST-NMR to determine the E a,out . To this end, the k out values for both CB7:G2 and CB7:G3 were determined at a series of temperatures and then correlated to the inverse temperature using the Arrhenius equation (eqn (1)): The linear relationship between ln(k out ) and T 1 in eqn (1) can be used to evaluate E a,out values even for host-guest systems that experience fast exchange (k out [ Du) at a given temperature (e.g., for CB7:G3 at 298 K, Fig. 2b, right). This can be achieved by simply performing a series of GEST experiments at lower temperatures where the condition k out # Du is fulflled. Therefore, we conducted GEST-NMR experiments of CB7 with either G2 or G3 in a phosphate buffer solution (5 mM sodium phosphate, pH ¼ 7) at different temperatures (Fig. 3a, b, S6 and S7 †), from which different k out values were extracted and plotted as a function of T 1 (Fig. 3c). The obtained linear relationships allowed the estimation of the dissociation E a values (eqn (1)) from the slope of these plots. Our fndings clearly demonstrate that the E a,out value of CB7:G2 (53 AE 1 kJ mol 1 ) is much higher than that of CB7:G3 (E a,out ¼ 32 AE 1 kJ mol 1 ), which is in good correlation with the observed differences in the extracted k out values. In comparison, dissociation activation energies of the fluorescent guest berberine and other fluorescent alkaloids are much larger, i.e. E a,out > 65 AE 1 kJ mol 1 , 58 as was determined by direct binding assays. Neither faster equilibrating guests for CB7 with lower E a,out barriers nor non-chromophoric guests can be investigated by established stopped-flow-coupled direct binding assays. Indeed, we thoroughly attempted to obtain the binding kinetics and activation energies for the CB7:G3 complex by fluorescentbased stopped-flow measurements at a range of different temperatures, pH and salt concentrations. However, in all cases the low emission signal change upon binding and the very fast guest inclusion kinetics prevented the extraction of any meaningful data with this established protocol. Likewise, even with the newly introduced kinetic versions of the indicator-and guest-displacement assays (kinGDA and kinIDA) that are applicable also to non-chromophoric guest, 59 we did not succeed in ftting reliable rate constants for these fast equilibrating guests (Fig. S3-S5 †). These results show that by applying GEST-NMR on supramolecular systems with fast exchange dynamic, one can directly quantify E a,out values, providing for the frst time access to activation energies of host-guest systems with fast formation and dissociation kinetics. It is important to mention, that CESTbased approaches are less suited for systems with a slow exchange dynamic. In this regard, the slow k out of G1 at 298 K and its relatively low boiling temperature prevented us from performing GEST at higher temperatures to obtain faster k out and thus did not allow to accurately evaluating the E a,out value of CB7:G1. To investigate GEST-NMR's applicability to systems where the E a,out is mediated also by external factors, we utilized the "supramolecular-lidding" capabilities of monovalent cations known to increase the binding affinities of guests to CBn in systems with very slow k out characteristics. 31,53,60,61 As a frst step, we used GEST-NMR to determine and quantify the effect of different monovalent cations on the k out of the studied CB7:G3 system (Fig. 4). In contrast to the very fast k out of G3 from its CB7:G3 complex and no asymmetry in the z-spectrum in phosphate buffer solution (Fig. 2b right, 298 K), an increase in the salt concentration resulted in a signifcant observable asymmetry of the z-spectrum plot. Altering the added cation (i.e., 140 mM of Li + , Na + , K + , Rb + , Cs + or NH 4 + ) resulted in different z-spectrum profles (Fig. 4a-c and S8 †), indicating different k out values (Fig. 4d). Specifcally, ftting of the experimental GEST data revealed that the fastest dissociation rate constant occurred in the presence of Li + (k out ¼ 2800 AE 300 s 1 , at 298 K), and the slowest one, in the presence of Na + (k out ¼ 1300 AE 100 s 1 , at 298 K), with an intermediate k out value in the presence of Rb + (k out ¼ 2000 AE 150 s 1 , at 298 K). The reproducibility of this observation was examined and the difference between the evaluated k out values was found to be statistically signifcant (Fig. S9 †). Note here, that this effect and the obtained dissociation rates were not affected by changes in the pH (Fig. S10 †), with similar k out values extracted for the same Na + containing solution but with a variety of pH values, i.e., pH ¼ 3 (1100 AE 80 s 1 ), pH ¼ 5 (1300 AE 80 s 1 ), and pH ¼ 7.2 (1300 AE 100 s 1 ). This observation indicates that the dissociation process of the guest from the CB7 cavity is governed primarily by the cation content in the system. To validate that the obtained exchange process is indeed between bound (CB7:G3) and free G3 in solution, a guest that strongly binds to CB7 (i.e., 1aminoadamantane) was used as a competitor (Fig. S11 †). The preferable binding of 1-aminoadamantane to CB7 completely eliminates the GEST effect, confrming that the observed exchange dynamics depend on the availability of the CB7 cavity to accommodate G3. After determining various cation effects on the k out rates in the fast exchanging system CB7:G3 (Fig. 4), we turned to study their effect on the E a,out values. To this end, we performed GEST-NMR experiments at different temperatures (Fig. 5a and S12 †) on CB7:G3 in 5 mM phosphate buffer solution to which LiCl (fast exchange, Fig. 4a), RbCl (intermediate exchange, Fig. 4c) or NaCl (slow exchange, Fig. 4b) was added. The obtained E a,out values (evaluated from the slopes of the linear plots in Fig. 5b) are shown in Fig. 5c. We found that the E a,out value, which was found to be 32 AE 1 kJ mol 1 in the absence of cations (Fig. 3d), increased in the presence of Li + , Rb + and Na + to 34 AE 2, 37 AE 5 and 42 AE 3 kJ mol 1 , respectively (Fig. 5c). This observed dependency of the dissociation activation energy for guest exclusion on the cationic content manifests its role in mediating the dissociation process even in systems with fast exchange kinetics. By combining E a,out values with the enthalpy change of the reaction (extracted from the ITC data for each system, Fig. S13 †), the association activation energy (E a,in ) values were accessible. Furthermore, assuming a one-step reaction, we used the correlation of the Eyring equation to calculate the dissociation activation free energy (DG # out ) as summarized in Table 1. To confrm the formation of supramolecular CB7 "capsules" with M + $CB7:G3$M + and to assure that M + -CB7 capping indeed occurred and mediated the obtained E a,out values, 1D-NMR ( 7 Li-NMR, 23 Na-NMR and 87 Rb-NMR) and 7 Li-and 23 Na-diffusion NMR experiments were performed. 62,63 This entailed the direct measurement of the NMR-characteristics of the cations (Li + , Na + or Rb + ) in aqueous solutions of LiCl, NaCl and RbCl with and without CB7. From the obtained 1D 7 Li-NMR, 23 Na-NMR and 87 Rb-NMR spectra, it is evident that CB7 has no effect on the chemical shift of 7 Li + (Fig. 6a, left), which is in contrast to the pronounced effect on the chemical shift of 23 Na + (Fig. 6a, middle) and 87 Rb + (Fig. 6a, right). These observations indicate a stronger interaction of CB7 with the Rb + and Na + cations as compared to that with the Li + cations and correlate with previous studies showing that different cations have different affinities to the portals of CBns. 47,52,60,61, Our 7 Liand 23 Na-diffusion NMR experiments are in agreement with previous reports 68 and further corroborate these observation (Fig. 6b). Notably, we found no signifcant change in the diffusion coefficient of Li + upon the addition of CB7, further corroborating that this cation does not bind to the portals of the host (Fig. 6b, left). In contrast, we noted a signifcant reduction in the diffusion coefficient of Na + upon the addition of the CB7 host, either with or without G3 (Fig. 6b, right). The shift in the 133 Cs-NMR spectrum (Fig. S14a †) and the decrease in the diffusion coefficient of 133 Cs + in the presence of CB7 (Fig. S14b †) were similar to those obtained for Na + . This correlates with the slowest and similar k out values calculated for CB7:G3 in the presence of either Na + or Cs + (Fig. 4d). Such a reduction in Na + (or Cs + , Fig. S14b †) diffusivity confrms that these cations strongly bind to the CB7 portals and serve as an active "lid", in comparison to the in size smaller Li + cation. These observed different affinities of various cations to the host portals govern the changes in the transition energy barrier of the host-guest complex and, therefore, mediate guest egression kinetics. 61 ## Conclusions In summary, GEST-NMR was used to elucidate dissociation activation energies in host-guest assemblies featuring fast exchange dynamics highlighting the role of "supramolecular lids" in mediating guests' dissociation E a . Our results emphasize GEST's ability to quantify exchange rates that cannot be measured by other approaches used for the study of kinetics in host-guest systems, in general, and in particular in CBn-guest systems. 31,53,69,70 Performing GEST-NMR at a range of temperatures and plotting the quantifed k out values as a function of the (inverse) experimental temperature allowed the evaluation of different dissociation activation energies with various kinetic profles. Finally, we demonstrated the role of monovalent cations in mediating k out and the energetic barrier of guest dissociation by their supramolecular capping features for a fastexchanging system. Thus, GEST can serve as an important analytical tool in designing supramolecular systems where controlling the E a is crucial, such as switchable molecular host systems. The fact that GEST can be applied with a conventional NMR setup, which is available at any research institute, offers new opportunities to explore dissociation dynamics and E a in a variety of supramolecular systems and should provide insights into less-studied mechanisms. The extension of CEST-NMR experiments to 15 N-and 13 C-and its implementation in other dynamic molecular systems, such as proteins, 71,72 emphasizes the potential of the proposed approach to be further developed. Because state-of-the-art experimental methods for determining the rate constants of host-guest complexes were so far limited to comparably slow binding systems (<10 s 1 ), having now a method at avail that can also be applied to rapidly unbinding guests (>1000 s 1 ) may open new possibilities for shedding light on fundamental questions in host-guest complexation kinetics. GEST can thus provide an additional insight into binding mechanisms in host-guest systems, as for example in the influence of guest and host desolvation that remains hidden so far. Therefore, we envision that using GEST-NMR for the study of binding kinetics and dissociation E a will advance the knowledge of supramolecular systems toward better understanding their kinetic properties and allow their further development as functional materials. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Elucidating dissociation activation energies in host\u2013guest assemblies featuring fast exchange dynamics", "journal": "Royal Society of Chemistry (RSC)"}

local_environment_and_dynamic_behavior_of_fluoride_anions_in_silicogermanate_zeolites:_a_computation

## Abstract: In silicogermanate zeolites containing double four-ring (d4r) building units, the germanium atoms preferentially occupy the corners of these cube-like units, but the absence of long-range order precludes a determination of the preferred arrangements of Si and Ge atoms by means of crystallographic methods. If fluoride anions are present during the synthesis, they are incorporated into these cages. Due to the sensitivity of the 19 F chemical shift to the local environment, NMR experiments can provide indirect insights into the predominant (Si,Ge) arrangements. However, conflicting interpretations have been reported, both with regard to the preference for, or avoidance of, Ge-O-Ge linkages, and concerning the equilibrium position of fluoride inside the cage, where fluoride might either occupy the cage center or participate in a partly covalent Ge-F bond. In order to shed light on the energetically preferred local arrangements, periodic dispersion-corrected density functional theory (DFT) calculations were performed for the AST framework, which is synthetically accessible across the range of (Si 1-n ,Ge n )O 2 compositions (0 ≤ n ≤ 1). DFT structure optimizations for (Si,Ge)-AST systems containing fluoride anions and organic cations revealed that arrangements of Si and Ge which maximize the number of Ge-O-Ge linkages are energetically preferred, and that fluoride tends to form relatively short (~2.2 to 2.4 Å) bonds to Ge atoms surrounded by Ge-O-Ge linkages. The preference for Ge-O-Ge linkages disappears in the absence of fluoride. DFT-based Molecular Dynamics calculations were performed for selected AST models to analyze the dynamics of fluoride anions confined to d4r cages. These calculations showed that the freedom of movement of fluoride varies depending on the local environment, and that it correlates with the average Ge-F distance. An analysis of the Ge-F radial distribution functions provided no evidence for a coexistence of separate local energy minima at the cage center and in the proximity of a germanium atom. The computational approach pursued in this work provides important new insights into the local structure of silicogermanate zeolites with d4r units, enhancing the atomic-level understanding of these materials. In particular, the findings presented here constitute valuable complementary information that can aid the interpretation of experimental data. ## Introduction Sparked by the increased interest in novel porous materials for various applications (catalysis, separation, energy storage etc.), the last 20 years have seen major advances in the field of silicogermanate zeolites, neutral-framework zeolites in which the tetrahedral sites (T sites) are occupied by silicon and germanium. While a few zeolite frameworks, like AST, are accessible both in pure-SiO 2 form and as silicogermanates with various Ge contents, there are also many instances where the incorporation of Ge stabilizes frameworks that are not accessible in all-silica or aluminosilicate composition. 1 Some of these frameworks have very large pore openings, two of the most spectacular examples being the mesoporous zeolite ITQ-37, with pore apertures surrounded by 30 T atoms, 2 and ITQ-44, which possesses 18-ring pore apertures and has an exceptionally low framework density. 3 Besides efforts in the direct synthesis, it has been shown that frameworks which are composed of Si-rich sheets connected by Ge-rich building units can be disassembled through selective Ge removal, and the layers reassembled to form new materials. 4,5 This ADOR (assembly-disassembly-organisation-reassembly) approach has led to the discovery of several new zeolites that are not accessible through conventional synthesis routes. 6 The equilibrium Ge-O-Ge angle (~130 deg) is smaller than the corresponding Si-O-Si angle (~145 deg). 7 As the formation of small (three-or four-membered) rings requires rather low T-O-T angles, the incorporation of germanium stabilizes such small rings. In particular, the presence of germanium may lead to the preferred formation of double four-ring (d4r) units, which are stabilized by small T-O-T angles (typically <140 degrees). 8,9 Different experimental methods have provided evidence that Ge preferentially occupies the vertices of the d4r cages: In favorable cases, a direct refinement of the occupancies of different T sites by Si and Ge from X-ray diffraction data is possible. Indirect information on the Ge siting can be obtained from 19 F and 29 Si solid-state NMR experiments, and such studies have been performed -among other systems -for ITQ-7 (ISV framework), 14 ITQ-17 (BEC), 11 ITQ-13 (ITH), 15,16 and (Si,Ge) analogues of octadecasil (AST framework). 17 Besides the experimental work, a preference of Ge for the d4r unit has also been found in computational modelling studies using a variety of techniques, comprising force field calculations, 9,11,12 Hartree-Fock calculations for small cluster models, 14 and periodic density functional theory (DFT) calculations. In systems with a mixed occupation of the vertices of a d4r unit by Si and Ge, the distribution of the two elements exhibits no long-range ordering, precluding a determination of preferred arrangements by means of crystallographic methods. Some information on the local ordering can be inferred from 29 Si-NMR spectroscopy. However, the influence of nearest-neighbor Ge atoms on the 29 Si chemical shift is relatively small, and contributions of several Ge atoms do not show a strictly additive behavior. 21 In zeolites synthesized via the fluoride route, fluoride anions tend to be incorporated in the smallest cage available in the structure, and they are therefore typically located in d4r cages if such building units are present. 22 Because the 19 F chemical shift depends on the composition of the d4r cage, 19 F-NMR experiments can provide insights into the local environment. It is well established that (8Si,0Ge) d4r cages give rise to a signal at -38 ppm (relative to CFCl 3 ), 17,23 , whereas (0Si,8Ge) d4r cages result in a signal at ~-15 ppm 17,22 (throughout this work, we use ((8-x)Si, xGe) as a shorthand notation to represent the occupation of the eight vertices of the d4r unit by 8-x Si and x Ge atoms). Other signals appearing at -20 ppm and -8 ppm in mixed (Si,Ge) systems were mostly interpreted as being due to (7Si,1Ge)/(6Si,2Ge) and (5Si,3Ge)/(4Si,4Ge) d4r units, respectively. 11,12,14,15,17, As an interpretation on the basis of experimental data alone may lead to ambiguities, complementary computational investigations were carried out by Sastre and co-workers: Using a combination of force field based calculations (to optimize the structures) and DFT calculations (to predict NMR shifts), they found that the computed NMR shifts for mixed (Si,Ge) systems agree much better with experiment when the F atom is not located at the center of the cage, but displaced towards a Ge atom. 26,27 For several systems, the calculations delivered pentacoordinated germanium with Ge-F bond distances in the range of 1.9 to 2.0 . The local environment of the pentacoordinated Ge atom, specifically the number of adjacent Ge-O-Ge linkages along edges of the d4r cage, was found to be a dominant factor in determining the chemical shift. This was corroborated in recent work on STW zeolite by Rigo et al., who found three distinguishable Ge environments on the basis of a combination of NMR experiments and DFT calculations: 13 (1) "Isolated" Ge atoms surrounded only by Ge-O-Si linkages: Ge(Ge) 0 , (2) Ge atoms surrounded by one or two Ge-O-Ge linkages along the cage edges: Ge(Ge) 1 and (Ge(Ge) 2 ), (3) Ge atoms surrounded by three Ge-O-Ge linkages: Ge(Ge) 3 (the fourth T-O-T linkage, which has no documented impact on the NMR shift, points away from the d4r cage, so three is the highest possible number of Ge-O-Ge linkages along cage edges). The AST zeolite framework was synthesized in pure-silica form by Caullet et al., who labelled it "octadecasil" to emphasize its clathrasil nature. 23 A pure-GeO 2 AST framework dubbed ASU-9 was later reported by Yaghi and Li, 28 and mixed (Si,Ge)-AST systems were prepared by Wang et al. and by Tang. 17,29 AST-type zeolites were also used as model systems in the aforementioned computational studies by Sastre and co-workers. 26,27 According to a Rietveld refinement of the GeO 2 -AST end member by Wang et al., the F atoms are located at the center of the d4r cage, however, a large isotropic displacement factor indicates rather large freedom of motion. 17 A slight splitting of the 19 F-NMR signal in this system was interpreted as being due to a displacement of fluoride from the cage center, pointing to the possibility of coexisting off-center local minima, as proposed earlier by Villaescusa et al. on the basis of semiempirical calculations. 30 29 Si-NMR results for mixed (Si,Ge) systems led to the conclusion that there is a preference for an alternating arrangement of Si and Ge, i.e. an avoidance of Ge-O-Ge linkages. 17 To clarify the rationale behind the present work, it is useful to point out some conflicting observations in the existing literature: (1) As discussed above, some NMR results have been interpreted as being indicative of an avoidance of Ge-O-Ge linkages, e.g. for (Si,Ge)-AST. 17 Force field based simulations on ITQ-21 and AST-type frameworks provided evidence for an energetic "penalty" for the formation of Ge-O-Ge linkages, with the relative energy increasing roughly linearly with increasing number of such linkages. 12,26 Conversely, a DFT study by Kamakoti and Barckholtz of the BEC framework delivered an arrangement in which two Ge atoms occupy adjacent cage vertices, i.e. in which a Ge-O-Ge link is present, to be the most favorable scenario for a (6Si,2Ge) d4r cage. 18 A configuration with two Ge-O-Ge linkages was found to be the most stable distribution for a (5Si,3Ge) d4r cage. Other computational studies using electronic structure methods did not predict an energetic penalty for the formation of Ge-O-Ge linkages. 14,19 In fact, recent NMR studies on ITQ-13 and STW-type silicogermanates delivered evidence for a presence of Ge-O-Ge linkages at relatively low overall Ge contents. 13,16 (2) Crystallographic investigations on SiO 2 -and GeO 2 -zeolites and mixed (Si,Ge) systems containing d4r units always located fluoride at the center of the cage, 12,17,23, whereas some NMR studies have pointed to an off-center displacement in Ge-containing systems. 17,30 While there is no direct experimental evidence for the presence of Ge-F bonds, the computations performed by Sastre and co-workers, which were validated against experimental NMR data, predicted the existence of pentacoordinated germanium atoms in both GeO 2 -AST (d(Ge-F) ≈ 2.2 ) and in mixed (Si,Ge) systems (d(Ge-F) ≈ 1.9 to 2.0 ). 26,27,34 To address the first point, structure optimizations using dispersion-corrected DFT were performed for models of (Si,Ge)-AST across the full compositional range. AST was chosen as a convenient model due to its relative simplicity (two non-equivalent T sites, 20 T atoms in the conventional unit cell, Figure 1) and due to the fact that it is synthetically accessible for essentially any Ge content. Calculations were performed for structure models including fluoride anions and tetramethylammonium cations, dubbed (TMA,F)-AST, and for models of the bare AST framework. On the basis of the calculations, the energetically preferred arrangements of Si and Ge were identified, and it was attempted to establish trends regarding the preference for, or avoidance of, certain arrangements. With regard to the second point, the structure optimizations for (TMA,F)-AST models also deliver insights into the preferred fluoride positions and, thus, the presence or absence of pentacoordinated Ge atoms. Another interesting aspect is the dynamic behavior of fluoride in the d4r cage: For example, it is conceivable that several local minima exist within one cage -e.g. one at the center and another one close to a Ge atom -and that fluoride atoms "hop" between minima over time. To elucidate the dynamic behavior, DFT-based Molecular Dynamics (MD) calculations were performed for selected models of (TMA,F)-AST. ## Figure 1. Top: Natural tiling representation of the AST framework. Middle: Atomistic representation of fundamental building units: d4r cages with fluoride anions and ast cages with TMA cations. Bottom: Visualization of (TMA,F)-SiO 2 -AST in tetragonal space group 4 (see text). ## Models of AST structure The AST framework is a relatively simple one, as it can be assembled from two types of natural tiles, namely cube-like d4r cages (t-cub tile, face symbol ) and larger octadecahedral ast cages (t-trd tile, face symbol [4 6 •6 12 ]). 35,36 Crystallographic investigations have established that fluorine atoms occupy the center of the d4r cages, whereas the organic structure directing agents (OSDAs), also called "templates", reside in the ast cages. 23,31 There are two distinct T sites in the structure: The T1 site corresponds to the vertices of the d4r cage, with the three surrounding T1-O-T1 linkages forming edges of the cage. The fourth linkage from each T1 corner forms a connection to the T2 site, which is connected to T1 sites belonging to four different d4r cages (Figure 1). In the cubic aristotype of AST, the T1-O-T2 linkage is linear. However, the actual symmetry of both as-made and calcined all-silica AST (octadecasil) is tetragonal (space group 4/ ), thereby avoiding linear T-O-T linkages. 23,31,37 Tetragonal symmetry was also found for GeO 2 -AST and for most mixed (Si,Ge) systems. 17,29 The starting models used in the calculations for OSDA-containing AST included tetramethylammonium (TMA) molecules, as this is a fairly simple OSDA that has been successfully used to direct the synthesis of SiO 2 -AST (octadecasil). 23,31 One complication arises, as the OSDA is disordered in the experimentally determined structure, and an ordering requires a lowering of the symmetry. Different relative orientations of the OSDA molecules in adjacent cages would lead to different resulting space groups. In the present work, an orientation was chosen that retains the body-centering of the lattice, leading to a structure in space group 4 (Figure 1). While (Si,Ge)-AST and GeO 2 -AST systems have not been synthesized with TMA, the reported syntheses used other alkylammonium OSDAs. 17,29 In addition to the pure-SiO 2 and pure-GeO 2 end members, models with Ge contents n(Ge) ranging from 0.1 to 0.9 were prepared. These models do no longer possess tetragonal symmetry (unless in very few special cases), but it was assumed that the body centering is preserved, i.e. the d4r units at the center and at the corners of the pseudotetragonal unit cell have an identical arrangement of Si and Ge. The arrangements considered, which were generated for a model of a single d4r cage using the Supercell code, 38 are visualized in Figure 2. The labelling scheme to designate these systems makes use of a) the number of Si and Ge atoms occupying the vertices of the cage, b) the point group symmetry of the (Si,Ge) arrangement, and c) the number of Ge-O-Ge linkages (this is dropped from the label if there is only one (Si,Ge) distribution). For example, AST_(4Si,4Ge)_C 4v _4GeGe has 4 Si and 4 Ge atoms at the vertices of the d4r cages, which are arranged in a way that the point group symmetry of the (Si,Ge) distribution is C 4v -this corresponds to an occupation of all corners of one face by Ge, and therefore 4 Ge-O-Ge linkages (the edges of that face). According to Kamakoti and Barckholtz's study of BEC, an occupation of up to 4 T sites per d4r unit by Ge is energetically favorable. Beyond that, other T sites are occupied. To account for a possible occupation of the T2 site by germanium, comparisons between models in which T2 is occupied either by Si or by Ge were made for compositions ranging from n(Ge) = 0.5 (10 Si and 10 Ge per unit cell) to n(Ge) = 0.8 (4 Si and 16 Ge per unit cell). T2 = Si was assumed for n(Ge) up to 0.4 and T2 = Ge for n(Ge) = 0.9. A mixed occupancy of the T2 site by Ge and Si was not considered. If the T2 site is occupied by Si, it is omitted from the label (except in the case of AST_(0Si,8Ge)_O h _T2Si), but if it is occupied by Ge, the designator "T2Ge" is appended to the label. We have to note that the symmetry of (TMA,F)-AST, which is tetragonal, is reduced with respect to the cubic aristotype. Therefore, the d4r cages possess two different types of edges (T-O-T linkages), along the tetragonal c-axis and perpendicular to it. If this was accounted for in the setup of the model systems, a larger number of configurations than those shown in Figure 2 would arise. For example, two neighboring Ge atoms in the AST_6Si_2Ge_C 2v _1GeGe model could be linked along c or perpendicular to c. Test calculations for a few models indicated that the energy differences between such different configurations are small, and therefore only one arrangement of Si and Ge at the cage vertices was considered for each case. It has to be noted that different observations were made in a DFT study of BEC by Kamakoti and Barckholtz. 18 However, in that system, the T-O-T angles at the edges of the d4r unit vary considerably, and the occupancy of neighboring T sites by Ge is energetically favored when they are connected via a ## Computational details DFT structure optimizations and DFT-based MD calculations were performed using the CP2K code (version 2.6.2, installed on the HLRN supercomputer "Konrad"), which uses a hybrid Gaussian and plane wave scheme. 39,40 All calculations used the PBE exchange-correlation functional in conjunction with the "Grimme-type" D3 dispersion correction, 41,42 a plane wave energy cutoff of 600 Ry, and Goedecker-Teter-Hutter pseudopotentials devised by Krack. 43 Only the gamma point was used to sample the first Brillouin zone. All calculations used Gaussian "MOLOPT" basis sets that are included in the current distribution of CP2K: 44 The structure optimizations used triple-zeta (TZVP) basis sets, as it was found that these give a much more accurate difference in lattice energy between SiO 2 -AST and quartz than double-zeta (DZVP) basis sets (TZVP: 11.6 kJ mol -1 ; DZVP: 16.1 kJ mol -1 , experiment: 10.9 kJ mol -1 ). However, both basis sets were found to deliver very similar equilibrium structures, justifying the use of the less demanding DZVP basis in the MD simulations. The structure optimizations were performed for the conventional unit cell of tetragonal AST. All atomic coordinates and the lattice parameters were optimized, fixing the symmetry of the lattice to tetragonal (a = b, all angles = 90 degrees). The optimizations used the following convergence criteria: Maximal geometry change = 2⋅10 -5 bohr, maximal residual force = 1⋅10 -6 Ha bohr -1 , maximal pressure deviation = 0.01 GPa. The MD simulations took the optimized structures as starting point and used a 2×2×1 supercell. MD simulations were performed in the canonical (NVT) ensemble for a temperature of 298 K, using a Nosé-Hoover thermostat with a timestep of 0.5 fs and a time constant of 50 fs. To improve the statistics, three independent trajectories were run for each system, with each trajectory consisting of an equilibration stage of 2.5 ps (5000 steps) and a production stage of 7.5 ps (15000 steps). All results presented below correspond to averages over three trajectories. For the analysis of the production part of each MD trajectory, the root mean square displacements (RMSD) of all elements and radial distribution functions (RDF) of selected pairs of elements were computed using the VMD software, version 1.9.3. 45 ## Equilibrium structures of the end members: SiO 2 -AST and GeO 2 -AST Initial DFT structure optimizations were performed for the end members SiO 2 -AST and GeO 2 -AST, both with and without TMA template and F anions. The resulting lattice parameters are shown in Table 1, together with experimental data, where available. 17,31,37 For SiO 2 -AST, the computed lattice parameters agree reasonably well with experiment, although there is a systematic tendency to overestimate the length of the c-axis, while underestimating a. Similar findings have been discussed in more detail in previous benchmarking work. 46,47 For (TMA,F)-GeO 2 -AST, the c-axis is overestimated even more markedly. However, it has to be considered that the experimental sample contained a different OSDA than the model used in the calculations (dimethyldiethylammonium instead of TMA), and that the dimensions of the OSDA will affect the equilibrium lattice parameters. In addition to the lattice parameters, we also evaluated the difference in DFT energy per TO 2 unit between template-free SiO 2 -AST/GeO 2 -AST and α-quartz/quartz-type GeO 2 (∆E DFT ). This quantity can be taken as a realistic approximation to the enthalpy of transition ∆H trans . The enthalpy of transition of SiO 2 -AST has been determined as 10.9±1.2 kJ mol -1 per formula unit using solution calorimetry experiments. 49 The DFT-calculated energy difference is within the experimental error limits, amounting to 11.6 kJ mol -1 . Calculations for several other all-silica zeolites and α-cristobalite, summarized in the Supporting Information (SI, Table S1), provided similarly good agreement, giving confidence in the predictions of the relative stability of AST systems, presented below. We may note that calculations using the plane-wave code CASTEP and employing similar dispersion-corrected DFT methods gave ∆E DFT values in the same range: A (so far unpublished) energy difference of 10.9 kJ mol -1 was obtained with the PBE-D2 functional, 46 and the PBEsol-D2 functional delivered 11.4 kJ mol -1 . 47 The calculated ∆E DFT for GeO 2 -AST with respect to quartz-type GeO 2 is approximately 1.5 times as large as for the silica systems, amounting to 17.7 kJ mol -1 . This is in line with experimental findings, as an increase in metastability (corresponding to larger values of ∆H trans ) with increasing germanium content has been observed in calorimetric experiments on silicogermanate zeolites. 50 In the (TMA,F)-systems, the fluoride anions occupy the center of the d4r cages, as shown in Figure 3 containing d4r units, and in molecular, anionic d4r-like [T 8 O 12 (OH) 8 F]units. 30,51,52 High-level wavefunction based calculations for these molecular systems pointed to a moderately strong non-covalent interaction ("tetrel bonding"), but no off-center displacement of fluoride. 53 On the contrary, both force field and DFT-PBE calculations by Sastre and Gale for as-synthesized GeO 2 -AST predicted a displacement of F towards one Ge atom, with a Ge-F distance of ~2.17 , indicating a significant covalent bonding component. 34 These contrasting findings could indicate the presence of different local minima within one cage, a point to which we return in the context of the MD calculations. ## Equilibrium structures of (TMA,F)-AST across the compositional range With regard to (TMA,F)-AST models across the range of compositions, we first take a brief look at the evolution of the DFT-optimized lattice parameters a and c as a function of the Ge content, shown in Figure 4. As expected, both lattice parameters increase with n(Ge). In the composition range from n(Ge) = 0.5 to 0.8, there is a marked dependence on the occupation of the T2 site, with models with T2 = Ge having a longer c-axis and shorter a-axis than models with T2 = Si. Experimental values are also included in Figure 4. 17,29,31 The quantitative deviations between DFT and experiment can, at least in part, be explained with the presence of a different OSDA in all experimental samples except SiO 2 -AST. It is, however, interesting to observe that the GeO 2 -AST sample has a shorter a-axis than most (Si,Ge) systems, whereas its c-axis is about 0.3 longer than that of the most Ge-rich (Si,Ge) sample (where n(Ge) = 0.833). In the light of the relationships discussed above, this might indicate that the T2 site is preferentially occupied by Si even in Ge-rich samples. As we will see below, the calculations predict T2 = Si to be preferred up to n(Ge) = 0.7. 17,29,31 In the following, we concentrate on the relative stabilities of different (Si,Ge) distributions for a given Ge content, and on the environment of the fluoride anions in the most stable structures. To analyze this in a systematic fashion, we present the most significant findings for each composition from n(Ge) = 0.1 to 0.9. The relative energies and shortest Ge-F distances for all models are compiled in Table 2, and fluoride environments of selected models are shown in Figures 3, 5, and 6. In these figures, Ge-F distances of less than 2.4 are indicated using thick bicolor lines, like covalent bonds, whereas distances between 2.4 and 2.7 are represented using thin grey lines (Ge-F distances above 2.7 are not shown). Of course, the choice of a distance of 2.4 as a threshold is entirely arbitrary. In reality, there will be a more or less smooth transition between short, partly covalent Ge-F bonds and long non-covalent Ge-F contacts. n(Ge) = 0.3: For this composition, the lowest-energy model is AST_(5Si,3Ge)_C s _2GeGe, where one germanium atom participates in two Ge-O-Ge links. As is visible in Figure 6 a, fluoride forms a rather short Ge-F bond to this Ge(Ge) 2 atom, whereas the distances to the other two Ge atoms are significantly longer. Unlike in the n(Ge) = 0.2 case, where the energetically preferred model is the only system having a short Ge-F bond, such a bond is also found in the second best system, AST_(5Si,3Ge)_C s _1GeGe. n(Ge) = 0.4: The complexity increases further for this composition, with a total of six distinct (Si,Ge) distributions. All in all, the tendency observed above is corroborated: The energetically most favorable model AST_(4Si,4Ge)_C 3v _3GeGe contains one Ge(Ge) 3 atom, and fluoride is bonded to this atom, with a short Ge-F distance of 2.17 (Figure 6 b). There is, however, a second configuration with longer Ge-F distances that is almost identical in energy. In this AST_(4Si,4Ge)_C 4v _4GeGe model, shown in Figure 6 c, the four germanium atoms surround one face of the d4r unit. The fluoride anion is displaced from the cage center towards this face, with four rather similar Ge-F distances ranging from 2.42 to 2.54 . We need to add a few words of caution to the discussion of the relative stability of different arrangements, pointing out the simplifications that -necessarily -have to be made in such a study: First of all, it has to be kept in mind that the arrangement in a real crystal structure is a result of the processes happening during hydrothermal zeolite synthesis (assembly of isolated TO 4 tetrahedra to form precursor building units followed by the connection of these building units to form the extended structure). Such a complex assembly process, which is governed by the interplay of thermodynamics and kinetics, cannot, at present, be captured with DFT methods. While we see no reason to expect that fundamentally different preferences, especially an avoidance of Ge-O-Ge linkages within the d4r units, would be observed if isolated building units were studied, there is clearly scope for further computational studies, especially with regard to the effect of the solvent. Secondly, it has to be re-emphasized that there are many cases where the energy differences between several arrangements having a given composition are fairly small (often below 5 kJ mol -1 ). Therefore, different local environments are likely to coexist within a crystal, and one should not be misled to expect that any of the building units shown in Figures 5 and 6 would correspond to the exclusively occurring arrangement within a real zeolite. A coexistence of different arrangements is indeed found experimentally, where up to three distinct 19 F NMR signals have been observed in (Si,Ge)-STW with intermediate Ge contents, with two of these signals corresponding to several possible arrangements. 13 As a final remark to this section, we compare the present results to the earlier force field based calculations by Sastre and co-workers. 26 Their calculations for AST models containing 2, 3, and 4 Ge atoms in one d4r unit delivered a completely different energetic ordering: For each case, systems having no Ge-O-Ge link at all were found to be the energetically favored configurations (AST_(6Si,2Ge)_C 2v _0GeGe, AST_(5Si,3Ge)_C 3v _0GeGe, and AST_(4Si,4Ge)_T d _0GeGe in the nomenclature of the present work). These models also exhibited the shortest Ge-F distances (with 1.9 to 2.0 ). On the other hand, DFT-based predictions of the 19 F-NMR shifts agreed best with experimental observations for the AST_(6Si,2Ge)_C 2v _1GeGe, AST_(5Si,3Ge)_C s _2GeGe, and AST_(4Si,4Ge)_D 2h _2GeGe models. Interestingly, the former two are the lowest-energy configurations for the respective compositions in the present study. Because these arrangements appeared thermodynamically unlikely on the basis of the force field calculations, Sastre and co-workers attributed the apparent preference to form d4r cages with Ge-O-Ge linkages to kinetic effects. However, the present DFT results indicate that the energetic ordering obtained in the force field calculations may be unreliable. ## Equilibrium structures of template-free AST across the compositional range Many previous computational studies of the relative stability of different Ge arrangements in silicogermanates have exclusively investigated models of the bare zeolite frameworks, without considering the influence of organic templates and fluoride anions on the energetic ordering. This is a significant simplification, since the more likely configurations in calcined samples will correspond to those that are energetically favored in the as-synthesized material, unless a rearrangement of the T atoms occurs upon template removal (which appears unlikely). As a consequence, it is interesting to assess whether the energetic ordering changes when then non-framework species are removed from the structure models. The energy differences ∆E DFT obtained for models of the bare AST framework are included in Table 2. First of all, it is worth noting that the range of ∆E DFT values for a given composition is considerably smaller for template-free framework models: The difference between the most and least favorable arrangements never exceeds 9 kJ mol -1 (per d4r unit), whereas it amounts to more than 20 kJ mol -1 for some Ge contents in the case of (TMA,F)-AST. Secondly, the energetic ordering changes for all compositions except n(Ge) = 0.3. In particular, models with T2 = Ge are now favored from n(Ge) = 0.5 onwards. An inspection of the unit cell volumes (SI, Figure S1) shows that these models have systematically smaller volumes than those with T2 = Si. In contrast, the volume of the (TMA,F)-AST systems increases with Ge content, but does not vary appreciably for a given composition. Unlike for the template-containing systems, where (Si,Ge) arrangements having a larger number of Ge-O-Ge linkages are preferred, there is no such clear trend for the template-free models. Taken together, it can be concluded that the presence of fluoride anions and organic templates has a significant impact on the relative stability of different (Si,Ge) arrangements. Therefore, some caution should be exercised in the interpretation of computational results obtained for template-free models. ## Dynamics of fluoride anions confined to d4r cages As shown in the previous parts, the presence of localized Ge-F bonds depends on the heterogeneity of the environment. One might now wonder whether there could be several coexisting local minima for fluoride inside the cage, e.g. one at the center and another one in the proximity of a Ge atom, only one of which would be found in a static DFT optimization. In order to sample a significant part of the potential energy surface, constrained optimizations were performed in which the position of fluoride was varied along the body diagonal of the d4r cage. These optimizations were performed for SiO 2 -AST, GeO 2 -AST and AST_(7Si,1Ge)_C 3v , starting from the DFT-optimized structures. The coordinates of the atoms forming vertices and edges of the cage were optimized, whereas those of fluoride and of all other framework and non-framework atoms were held fixed. The resulting potential energy curves are shown in Figure 7 a. For both SiO 2 -AST and GeO 2 -AST, the energy minimum is located at the center of the cage, and the energy increases smoothly when moving towards a corner. There is a good correspondence between the potential energy curve obtained for SiO 2 -AST and that calculated by Goesten et al. for fluoride at the center of a silsesquioxane model of a d4r unit. 54 The potential well is somewhat wider for GeO 2 -AST than for SiO 2 -AST due to the larger dimensions of the d4r cage. For AST_(7Si,1Ge)_C 3v , an asymmetric potential energy curve is found, with an energy minimum that is located at a distance of ~2.2 from the germanium atom, in accordance with the Ge-F distance obtained from the optimization. If fluoride is placed at the center of the cage, the energy is about 8 kJ mol -1 higher, and there are no indications for a secondary local minimum at the center (or elsewhere along the body diagonal). In order to sample the potential energy surface more comprehensively, ab-initio MD simulations were performed for the systems visualized in Figures 3, 5, and 6, thus, at least one model was considered for each composition of the d4r cage from (8Si,0Ge) to (0Si,8Ge). To start with, the root mean square displacements were evaluated for all elements (note that the RMSDs were calculated using the average coordinates from the 7.5 ps trajectories as reference). The RMSDs obtained for SiO 2 -AST, GeO 2 -AST, and AST_(7Si,1Ge)_C 3v are compiled in Table 3. For the T atoms, a slightly larger RMSD is found for Ge in GeO 2 -AST (0.17 ) compared to Si in SiO 2 -AST (0.14 ), despite the larger mass of germanium. This can be explained with the lower rigidity of Ge-O bonds compared to Si-O bonds, leading to an increased freedom of motion of the atoms at the center of the TO 4 tetrahedra. For oxygen, the RMSDs are essentially identical for all three models (0.22 ), as one would expect for isostructural systems. Very pronounced differences among the three systems are found for fluoride: For SiO 2 -AST, the RMSD of 0.28 reflects the larger freedom of motion of the tetrel-bonded fluoride anions as compared to the (slightly lighter) oxygen atoms, which are held in place through directional Si-O bonds. The increased RMSD of fluoride in GeO 2 -AST of 0.36 can be attributed to the larger dimensions of the d4r cage, permitting larger displacements from the cage center, in line with the wider potential well found above. For AST_(7Si,1Ge)_C 3v , where fluoride participates in a Ge-F bond, the RMSD is decreased to 0.22 , thus being virtually the same as for the oxygen atoms. This clearly shows that the fluoride anions lose a significant portion of their freedom of motion upon formation of Ge-F bonds, corroborating the localized (partly covalent) nature of these bonds. With regard to the TMA molecules, the RMSD of the nitrogen atoms is only moderately larger than that of fluoride confined to d4r cages, whereas the RMSDs of C and H atoms are much larger. As the nitrogen atom constitutes the center of mass of the TMA molecule, its small RMSD indicates that the overall displacement of the TMA molecules within the ast cages is only modest. The increased RMSDs of the atoms belonging to the methyl "arms" can be attributed to rotations of the TMA molecule about its center of mass. The larger dimensions of the ast cage in GeO 2 -AST as compared to SiO 2 -AST lead to an increased motion of the TMA molecules, reflected by systematically larger RMSDs. The RMSDs for other mixed (Si,Ge) systems are compiled in the SI (Table S3). As the analysis of the RMSDs of the other elements reveals no trends apart from those already mentioned, the following discussion will focus on the RMSD of fluoride anions, shown in Table 4. For the three systems with n(Ge) = 0.2, the RMSD varies from 0.23 for AST_(6Si,2Ge)_C 2v _1GeGe, the system having a localized Ge-F bond with d(Ge-F) = 2.26 , to 0.32 for AST_(6Si,2Ge)_D 3d _0GeGe, where fluoride resides almost at the cage center (Figure 5 c). This evolution agrees with the trend identified in the previous paragraph, and one might now expect that the reduction of the freedom of motion of fluoride is an entirely "local" phenomenon that depends solely on the interaction with the closest germanium atom. However, when plotting the values for all systems compiled in Table 4 against the shortest Ge-F distance (SI -Figure S2), it is apparent that there is no perfect correlation, as there are some systems where the RMSDs are much larger than what would be expected from the shortest Ge-F distance. In contrast, an alternative plot that uses the average Ge-F distance, calculated over all x Ge-F contacts within the ((8-x)Si,xGe) d4r unit, shows a near-perfect correlation for all systems (Figure S2). Thus, the freedom of motion of fluoride depends not only on the nearest neighboring Ge atom in the equilibrium structure, but also on the presence of, and distance to, other Ge atoms in the d4r cage: When there is only one short Ge-F contact, or two Ge-F contacts that are similarly short, as in AST_(6Si,2Ge)_C 2v _1GeGe, fluoride remains rather confined to its equilibrium position at the temperature considered, with an RMSD that hardly exceeds that of strongly bonded framework oxygen atoms. However, when there are additional germanium atoms that are further away, attractive secondary interactions with these atoms cause a more dynamic behavior. The first peaks in the T-F radial distribution functions g(r), which correspond to the distances between the fluoride anion in a d4r cage and the T atoms at the cage vertices, are shown in Figures 7 b, 8, and 9 (the RDFs were always normalized in a way that the cumulative g(r) for the first maximum corresponds to the number of Si or Ge atoms at the vertices of the cage). Furthermore, the median Ge-F distances, which mark the separation between the lower and upper 50% of the Ge-F distances within the first peak, are included in Table 4. First of all, it is worth pointing out that the median Ge-F distances obtained from the MD calculations are always within 0.04 of the average d(Ge-F) value measured in the DFT-optimized structures. If there were some cases in which fluoride anions moved from one local minimum to another one during the MD run, and remained confined to that minimum for a longer period of time, it could be expected that there were more pronounced deviations between the two values. The RDFs of SiO 2 -AST and GeO 2 -AST, shown in For the case of AST_(7Si,1Ge)_C 3v , also shown in Figure 7 b, there are two well-separated peaks for the Ge-F and Si-F distances in the RDF, with the position of the maxima corresponding to the (average) distances in the DFT-optimized structure. The Ge-F RDF has a rather sharp maximum, as one can expect from the presence of a localized bond that limits the freedom of motion. If there was a possibility for fluoride to relocate to a secondary local minimum at or near the cage center, one would expect a significant amount of Ge-F distances between of 2.6 to 2.7 , however, the value of g(r) in this distance range is close to zero. Moreover, the curve falls to zero for Ge-F distances below 2.0 . Thus, the MD calculations provide no evidence for the existence of very short Ge-F bonds with a length of ~1.9 , which have been proposed in previous computational studies. 26,27 A plot of the T-F distances over time (Figure S5) corroborates both the absence of a local minimum at the cage center and the instability of short Ge-F bonds <2.0 . Altogether, the results for the simplest systems permit us to conclude that there is only one local minimum for fluoride in each case, located at the cage center for the pure end members, and at a distance of about 2.2 from the Ge atom for AST_(7Si,1Ge)_C 3v . The situation becomes inevitably more complex -and less straightforward to analyze -when several corners of the d4r cage are occupied by germanium, where a broadening of the RDF peaks will be caused by a combination of dynamic effects and the presence of different Ge-F distances. Nevertheless, the inspection of the RDFs can deliver useful insights. Figure 8 visualizes the RDFs for all three systems having 6 Si and 2 Ge atoms at the cage vertices. For the energetically preferred AST_(6Si,2Ge)_C 2v _1GeGe model, the T-F RDFs show no unusual features, with smooth distributions around the equilibrium T-F distances. As there are two similarly short Ge-F bonds in this system, a motion of fluoride towards either of the two Ge atoms will cause only a minor broadening of the g(r) curve. In AST_(6Si,2Ge)_C 2v _0GeGe, the maximum is shifted towards shorter distances when compared to the equilibrium distances in the DFT-optimized structure, both of which are close to 2.5 . The rather large values of g(r) between 2.1 and 2.3 indicate that there is a significant probability for fluoride to be displaced from its equilibrium position towards either of the two Ge atoms. The most interesting observations can be made for AST_(6Si,2Ge)_D 3d _0GeGe, where the Ge-F RDF exhibits one very broad maximum with similar g(r) values from 2.2 to 2.9 (Figure 8 c), whereas the Si-F RDF has a narrower distribution that resembles that of SiO 2 -AST. Apparently, the fluoride anions move primarily along the body diagonal connecting the two Ge atoms, while maintaining rather similar distances to the surrounding Si atoms. A plot of the Ge-F distances over time (Figure S6) shows a rapid exchange of the fluoride anions among the two Ge atoms at opposite corners of the cage. For this particular case, additional calculations of potential energy curves analogous to those presented in Figure 7 a were performed, considering displacements of the fluoride anion along the Ge-Ge body diagonal and along one of the Si-Si body diagonals. The resulting curves, shown in Figure S8, show a very shallow potential well along the Ge-Ge diagonal: A displacement from the center towards either of the Ge atoms by 0.3 incurs an energy increase of only ~2 kJ mol -1 , whereas a displacement of the same magnitude towards an Si atom leads to an increase of about 6 kJ mol -1 . This anisotropy of the potential energy surface explains the rather peculiar shape of the Ge-F RDF observed above. While the environment of the Ge atoms is identical to that in the system with only one Ge atom, AST_(7Si,1Ge)_C 3v , there are no local minima at Ge-F distances of ~2.2 . This indicates that the overall (Si,Ge) arrangement at the vertices of the d4r cage, rather than the local environment of the Ge atoms, governs the relative stability of short Ge-F bonds compared to the center-of-cage position of fluoride. The radial distribution functions for systems with higher Ge contents are shown in Figure 9. Here, symmetric maxima in the Ge-F RDF appear for those systems where the distances between fluoride and germanium in the equilibrium structure are relatively similar, whereas the maxima for models having one or two short Ge-F contacts are asymmetric. Altogether, the radial distribution functions can be well described by assuming a statistic displacement of fluoride around its equilibrium position. Interestingly, even in the Ge-rich d4r cages, the average location of fluoride lies much closer to the Ge atoms than to the Si atoms: For example, the median Si-F distance in AST_(1Si,7Ge)_C 3v _T2Ge amounts to 2.98 , being almost 0.25 longer than the median Ge-F distance. In other words, the presence of a single Si atom at one of the vertices produces a significant anisotropy in the potential energy surface, which is pronounced enough to be clearly detectable even when the thermal motion of fluoride at 298 K is accounted for. ## Concluding remarks On the basis of the calculations presented above, the following conclusions can be drawn with regard to the preferred arrangements of Si and Ge at the d4r cages of AST frameworks, the equilibrium locations of fluoride, and its dynamic behavior: In fluoride-containing mixed (Si,Ge) systems, arrangements that maximize the number of Ge-O-Ge linkages are energetically favored. In the absence of fluoride, the computations predict a different energetic ordering without a clear trend. These findings strongly indicate that the presence of fluoride has a significant impact on the thermodynamically most stable (Si,Ge) arrangement, i.e. that there is a "templating effect". This may be particularly relevant for frameworks that can be synthesized both in the presence and in the absence of fluoride anions, where the most probable local structure may differ depending on the synthesis route. There are, however, a few caveats in the interpretation of the DFT results: On the one hand, the complexity of the assembly process cannot be captured with DFT calculations for the periodic structure, and further work using a "bottom-up" approach would be needed to elucidate how thermodynamics and kinetics of the assembly affect the final structure. On the other hand, the energetic ordering of different distributions depends on the "deformability" of the T-O-T linkages in the structure, thus, on the framework type. 18,55 Therefore, the trends discussed above for AST are not necessarily valid for other frameworks with d4r building units. In this context, it is worth noting that EXAFS results indicated that Ge atoms preferentially locate at a single face of a (4Si,4Ge) d4r cage in IM-12 (UTL), 4 whereas a detailed NMR study showed that other arrangements are predominant in ITQ-13 and ITQ-22. 55 With regard to the equilibrium position of fluoride, the calculations provide no evidence for the formation of pentacoordinated Ge atoms in GeO 2 -AST, where fluoride resides at the center of the cage, as in SiO 2 -AST. Formation of a relatively short Ge-F bond of ~2.2 occurs in (7Si,1Ge) d4r cages. Such short bonds are also prominent in d4r cages containing 2, 3, or 4 Ge atoms if the Ge atoms are located at neighboring vertices (many Ge-O-Ge linkages). At higher Ge contents, fluoride still tends to maintain shorter Ge-F than Si-F contacts, but there is no longer a formation of pentacoordinated germanium atoms. Neither DFT-based MD simulations nor calculations of the potential energy curve along the body diagonal of the d4r cage provide any indications for a coexistence of distinct local minima at the cage center and in the proximity of a Ge atom. By and large, the fluoride anions oscillate about the equilibrium positions obtained from the DFT optimizations at 298 K. Their freedom of motion, as measured through the RMSD, is correlated with the average Ge-F distance: If there is only a single Ge-F bond, the motion of fluoride is quite restricted, with an RMSD similar to that of the framework oxygen atoms, but if there are several Ge atoms in the vicinity, pairwise interactions with all these atoms lead to increased oscillations. Altogether, the present study provides significant new insights into the local structure of fluoridecontaining silicogermanate d4r units, insights that are not accessible through crystallographic methods, and may only be obtained indirectly via NMR spectroscopy or other spectroscopic methods. To this end, the results presented here should be of considerable value for future experimental studies of d4rcontaining silicogermanate zeolites, e.g. in the context of an in-depth characterization of new materials obtained via hydrothermal or ADOR-based synthesis routes. In order to predict quantities that are directly measurable experimentally, we are currently working on a DFT-based prediction of the 19 F-NMR shifts for different (Si,Ge)-AST models. Going beyond the AST framework, future computational work should consider a range of framework types in order to assess the differences and common features of various silicogermanate zeolites. With regard to the dynamics of fluoride anions under confinement, MD studies of related materials could deliver further insights. In all-silica zeolites that do not contain d4r units, fluoride is bonded to a single Si atom, forming SiO 4 F trigonal bipyramids, with the Si-F bonds typically pointing into small cages. 56 Thus, it would be interesting to compare the fluoride dynamics in structures where fluoride resides in different building units. Other relevant groups of systems comprise fluoride-containing aluminosilicates 22 as well as alumino-and gallophosphates. 22,57 For gallophosphates containing d4r units, both a central position of fluoride (e.g. in LTA-type GaPO 4 ) 58 and off-center displacements towards two or three Ga atoms (e.g. in cloverite) 59 have been reported in X-ray crystallographic studies. As in the present work, DFT-based MD calculations could enhance our understanding of the dynamics of fluoride in these systems, especially whether a dynamic exchange occurs between different local minima in a cage.

chemsum

{"title": "Local Environment and Dynamic Behavior of Fluoride Anions in Silicogermanate Zeolites: A Computational Study of the AST Framework", "journal": "ChemRxiv"}

reactor_environment_during_the_fukushima_nuclear_accident_inferred_from_radiocaesium-bearing_micropa

## Abstract: Radiocaesium-bearing microparticles (csMps), which are substantially silicate glass, were formed inside the damaged reactor and released to the environment by the fukushima Dai-ichi nuclear power plant accident in March 2011. The present study reports several valuable findings regarding their composition and structure using advanced microanalytical techniques. X-ray absorption near-edge structure of Fe L 3 -absorption indicated that the oxidation state of the iron dissolved in the glass matrix of the csMps was originally nearly divalent, suggesting that the atmosphere in which the csMps were formed during the accident was considerably reductive. Another major finding is that sodium, which has not been recognised as a constituent element of CsMPs thus far, is among the major elements in the glass matrix. The atomic percent of Na is higher than that of other alkali elements such as K and Cs. Furthermore, halite (NaCl) was found as an inclusion inside a CsMP. The existence of Na in CsMPs infers that seawater injected for cooling might reach the inside of the reactor before or during the formation of the CsMPs. these results are valuable to infer the environment inside the reactor during the accident and the debris materials to be removed during the decommissioning processes.A significant amount of radionuclides, including radiocaesium ( 134 Cs and 137 Cs), were released into the environment by the Fukushima Dai-ichi Nuclear Power Plant (FDNPP) accident in March 2011 1 . Radiation contamination around the nuclear plant is now mainly caused by 137 Cs because of its relatively high amount and long half-life (30.2 years). Most of the released radiocaesium was in a gaseous state at the time of the accident, fell onto the ground, and fixed to mineral grains, such as partially vermiculitised biotite, on the ground 2 . However, part of the radiocaesium was incorporated into micron-sized particles inside the reactor and these particles, termed radiocaesium-bearing microparticles (CsMPs), were emitted from the damaged reactor. In addition to the CsMPs which are also termed Type-A particles in some literature, Type-B particles larger than several dozen microns were found around the nuclear plant [3][4][5] . Although Type-B particles generally possess higher radioactivity than Type-A particles, specific radioactivity of Type-A is far higher than that of Type-B. Based on the activity ratios of the radiocaesium isotopes ( 134 Cs/ 137 Cs), Type-A and B particles are considered to have been formed in the Unit 2 or 3 and Unit 1 Reactors, respectively, because the activity ratios are varied by fuel burnup differences among the units 6 . All the CsMPs analysed in this study are classified to Type-A particles because they were spherules of a few microns and caesium could be detected using X-ray composition analysis owing to the high specific radioactivity 4 .CsMPs were first identified in aerosol filters collected in Tsukuba, Japan 7 . Microscopic analyses using synchrotron radiation (SR) X-ray and transmission electron microscopy (TEM) showed that the CsMPs are substantially silicate (SiO 2 ) glass with Cl, K, Fe, Zn, Rb, Sn, and Cs dissolved in the glass as major components [8][9][10] . Some of these elements often show inhomogeneous distributions inside the CsMPs. In many CsMPs, the Cs concentration increases from the centre to the surface of the particles, suggesting that gaseous Cs was present in the reactor and diffused into the molten silicate particles following their formation 10,11 . When Cs is abundant near the surface, other alkali elements (K and Rb) are inclined to possess reversed distributions. Fe and Zn are also enriched near the surface in some CsMPs 11 . Results of quantitative analysis of the glass considering the radial distributions of the elements showed the SiO 2 component reaches 60-70 wt% 10 . Further analysis using electron energy-loss spectroscopy (EELS) indicated that CsMPs do not contain boron 11 , which is easily incorporated into silicate glass if it exists in the surrounding atmosphere 12 . This infers that the B 4 C control rods created a eutectic alloy with stainless steel without vaporization during the accident. In this manner, characterization of CsMPs can provide information regarding the conditions inside the damaged reactors. However, a few issues relating to the composition of CsMPs remain inconclusive. Among them is the valence state of transition metals such as iron in the CsMPs, which may reflect the redox state of the atmosphere in the reactor during the accident. X-ray absorption near-edge structure (XANES) analysis using an SR X-ray microbeam indicated the high oxidation states of Fe, Zn, Mo, and Sn in the CsMPs, i.e., Fe 3+ , Zn 2+ , Mo 6+ , and Sn 4+ , suggesting that the CsMPs were formed in an oxidative atmosphere 8 . However, inclusions of chromian magnetite (Fe 2+ (Cr 3+ , Fe 3+ ) 2 O 4 ) were often found in the CsMPs, which conversely suggests that the atmosphere was rather reductive 11 . This discrepancy should be solved to determine the atmosphere inside the damaged reactor which might have filled with hydrogen and superheated steam during the accident 13 . Another issue is comprehensive determination of the chemical composition of the CsMPs. As previously mentioned, the absence of boron was confirmed by EELS in TEM but the existence of other light elements remains uncertain. Sodium is among the elements to be clarified. Although Cl in CsMPs was suggested to have originated in seawater injected into the reactor during the accident for cooling, the existence of Na has not been reported thus far, especially for type-A particles 9,10,14 . This leads a speculation that the seawater was not the origin of the Cl and did not sufficiently reach the inside of the reactor when the CsMPs were formed 15 . X-ray fluorescence analysis generally has low sensitivity for detection of light elements such as Na. Conventional energy-dispersive X-ray spectrometer (EDS) equipped to electron microscopes also cannot clearly confirm the presence of Na in CsMPs if Zn exists in samples because of the peak overlapping of Na Kα and Zn L (Na Kα: 1.041 keV, Zn Lα: 1.012 keV, Zn Lβ: 1.034 keV). Some previous research reported the Na peak in EDS spectra, however, the overlapping peaks of Na and Zn were blindly assigned to the Na peak in the papers. Although a wavelength-dispersive X-ray spectrometer, commonly equipped to an electron probe microanalyser, is able to resolve these peaks, Na in glass materials rapidly migrates and disappears because of intense electron-beam irradiation 16 . For these reasons, the presence of Na in CsMPs has not been clarified thus far. In this study, we attempted to clarify these two issues using XANES with scanning transmission X-ray microscopy (STXM) and EDS with scanning TEM (STEM) paying careful attention to X-ray and electron-beam irradiation effects. The following new results with respect to the composition and structure of CsMPs leads to valuable insights into the condition of the reactor during the accident. ## Results Oxidation state of iron in the CsMPs. To determine the oxidation state of iron, which is sensitive to the redox state of surrounding atmosphere, XANES analysis for the Fe L 3 edge was conducted using STXM. In the Fe L 3 edge, the peak energy of Fe 2+ (~708.2 eV) is lower than that of Fe 3+ (~709.6 eV) and their ratio can be estimated from the intensity ratio of the separated peaks 17 . For STXM measurement, thin specimens of two CsMPs, termed CsMP-HD and CsMP-Ma, were prepared for penetration by low-energy X-ray using a focused ion beam (FIB) apparatus. Figure 1a shows the Fe oxidation state map of CsMP-HD. In the map, divalent and trivalent iron are expressed by the gradation of green and red colours, respectively, using a singular value decomposition (SVD) method based on the spectra acquired from Fe 2+ -and Fe 3+ -rich regions 18 . In this study, the peak energy at 709 eV (Fe 2+ ) and 710.4 eV (Fe 3+ ) was slightly higher than the reference values because of inaccurate initial energy calibration. Averaged XANES spectra for the Fe L 3 edge from the specific regions are shown in Fig. 1b. The 1a clearly shows that the distribution of the iron oxidation state is inhomogeneous inside the CsMP. In the vicinity of the periphery of ~0.2 μm in thickness, the iron is completely oxidised to Fe 3+ . Inside the CsMPs, Fe 2+ is dominant but Fe 3+ increases toward the upper region. The thin specimens for STXM were prepared using FIB and the sample thinned toward the upper region, as confirmed by the relative thickness measurement using EELS (the detail of the measurement is described in the next section). This implies that the surfaces of the thin specimens were more oxidised than the inside, which is definitely an artefact and probably formed during sample preparation using FIB or XANES measurement using STXM. To evaluate the X-ray irradiation effect, a specific part of CsMP-Ma was repeatedly measured three times using STXM (Fig. 2a,b). The red-coloured area corresponding to Fe 3+ expanded and the Fe 3+ peak at 710.4 eV increased with repetitive measurement, indicating that the specimen was progressively oxidised from its surface because of X-ray irradiation. The STXM analysis for the whole area of CsMP-Ma following the repetitive measurement also showed a comparable result to that shown in Fig. 1 (Fig. 2c), but the XANES spectra acquired from the specific regions inside the CsMPs indicated that the Fe 3+ peaks in CsMP-Ma were higher than those in CsMP-HD (Fig. 2d). This result is because CsMP-Ma was thinner than CsMP-HD as confirmed again by the relative thickness measurement using EELS. Considering the preliminary X-ray irradiation for searching the analysis area, it is likely that the state of iron in the CsMPs before the measurement was nearly divalent except in the vicinity of the rim of the CsMPs. To evaluate the effects of FIB on the oxidation state, the synthetic glass with a similar composition was thinned in the same manner as CsMPs and analysed using STXM (Fig. S1). The means to synthesise the glass is described in the method section and its composition was determined using inductively coupled plasma (ICP) analysis as listed in Table S1. This glass was synthesised by using the previously reported composition of CsMPs as a reference, although some points are different from the actual CsMPs (for instance, Cl is not included in this glass) 10 . According to the XANES spectrum of the synthetic glass, the ratio of Fe: 2+ Fe 3+ is approximately 2:3. On the other hand, Mössbauer spectroscopy for the bulk synthetic glass also showed that the ratio is approximately 2:3 (Fig. S2 and Table S2). Thus, sample preparation by FIB had little effects on the oxidation state of iron in glassy materials. Following STXM analysis, elemental mapping was conducted using STEM-EDS for CsMP-HD and CsMP-Ma (Fig. 3). Their constituent elements were consistent with those of previously reported CsMPs. The Cs concentration increased from the centre to the surface, and in particular, Cs was predominantly concentrated at the rim region in CsMP-Ma. In contrast, K had a distribution opposite to that of Cs, which has also been previously reported. However, transition metals (Fe, Zn, and Sn) are homogeneously distributed in these CsMPs. It is probable the distributions of these elements and that of the iron valence state are not affected by each other. Sodium in the CsMPs. To examine the presence of Na in the CsMPs, Na K edge XANES spectra were acquired using STXM for the same thin specimens of CsMP-HD and CsMP-Ma (Fig. 4). A commercial soda-lime glass containing 13.7 wt.% Na 2 O was also measured as a reference of Na-containing glass and showed two peaks at the Na K edge 19 . The spectra from CsMP-HD and CsMP-Ma clearly showed an edge jump with two peaks at the Na K edge, indicating the presence of Na in the CsMPs, although the Na content is significantly lower than that of the soda-lime glass based on their peak heights. Based on this result, EDS analysis must also detect an Na Kα signal in spite of its peak overlapping with Zn L. Therefore, we carefully compared EDS spectra from the CsMPs to those from the synthetic glass with a similar composition but without Na. Before the comparison of EDS spectra, the sample thickness was estimated using EELS. Thickness t of TEM samples can be evaluated using the equation t/λ = Ln(I T /I 0 ), where λ is the inelastic mean free path of electrons with specific incident energy in the specimen, I T is the total transmitted electron intensity, and I 0 is the zero-loss peak intensity 20 . Assuming the values of λ are comparable between the synthetic glass and the investigated CsMPs, the relative thickness t/λ can be used for thickness evaluation. Figure 5a-c shows the determined relative thickness maps for the synthetic glass, CsMP-HD, and CsMP-Ma. Next, the EDS spectra were acquired from the rectangular areas shown in Fig. 5a-c and normalised by each Zn Kα peak intensity. The overlapping peaks of Zn L and Na Kα in the normalised spectra are shown in Fig. 5d. Because the synthetic glass did not contain Na, their peaks arose only from Zn L. The peak intensity of the synthetic glass decreased with increasing thickness because of X-ray absorption inside the samples. The peaks from CsMP-HD and CsMP-Ma are obviously higher than those from the synthetic glass although their sample thickness was similar. The difference in the peak intensity originated from the Na Kα, indicating the presence of Na in the CsMPs. Furthermore, the Na/Si ratio in the CsMPs was estimated by accurately calculating the intensity of the Na Kα and Si Kα peaks for the EDS spectra acquired from the rectangular areas shown in Fig. 5a-c. Because the energy of the X-ray of Na Kα and Si Kα is rather low, correction for the X-ray absorption inside the specimen is necessary to determine the accurate intensity of their peaks. The detailed calculation for the absorption correction is described in the Supplementary Information. As a result, the centre and periphery of CsMP-HD and CsMP-Ma showed an Na/Si atomic ratio of 0.185, 0.132, and 0.172, respectively (Table 1). The atomic ratio between other alkali elements (K, Rb, and Cs) and Si calculated using the Cliff-Lorimer method is also shown. The data clearly indicates that CsMPs contain much more Na than other alkali elements in the glass matrix, even in the periphery of CsMP-HD where the Cs concentration is higher. Furthermore, the Na/Si ratio in other three CsMPs was also calculated in the same manner (Fig. S4 and Table 1). First is termed CsMP-Fc which is also described in the next section, second is CsMP-Sg which is a new specimen with homogeneous distribution of Cs (Fig. S5), and third is WHT-1 which were a previously reported CsMP with high concentration of Cs at the periphery 11 . Consequently, these CsMPs also contain more Na than other alkali elements although the Na/Si ratio in CsMP-Sg is relatively lower. www.nature.com/scientificreports www.nature.com/scientificreports/ Halite included inside the CsMPs. Another CsMP, CsMP-Fc, was observed using (S)TEM (Fig. 6). Three types of inclusions were identified in the elemental maps of the CsMP (Fig. 6a). The first inclusion indicated by arrow c in Fig. 6b was rich in Cl. Figure 6c shows the electron diffraction (ED) pattern acquired from the inclusion, which corresponded to halite (NaCl) observed along <100>. The second inclusion indicated by arrow d in Fig. 6b was rich in Cr and Fe. Its ED pattern shown in Fig. 6d was explained by a spinel structure observed along <332>. The third inclusion was rich in S and/or Mo and was presumed to be molybdenite (MoS 2 ) which is also present in another CsMP 11 , although we failed to acquire its ED pattern. The EDS spectra of inclusions c and d and the glass matrix are shown in Fig. 6e. Inclusion c contained a large amount of Cl, and furthermore, the overlapping peak of Zn L and Na Kα was clearly higher than that of the glass matrix, indicating that the Na content was also high in the inclusion. This composition rich in Na and Cl also suggests that the inclusion is halite. The composition and ED pattern identified inclusion d as chromian spinel, which is a frequently observed inclusion inside CsMPs 11 . Thus, Na was present not only as a dissolved element in the glass matrix but also as a halite inclusion in CsMP-Fc. ## Discussion The previous XANES analysis using an SR X-ray microbeam reported that Fe, Zn, Mo, and Sn in the CsMPs were present in high oxidation states as Fe 3+ , Zn 2+ , Mo 6+ , and Sn 4+ , respectively 8 . In this study, we obtained a different result based on the STXM-based XANES analysis that iron in the CsMPs is nearly divalent. The result in previous XANES research 8 might reflect the properties of oxidised CsMPs because the iron in the CsMPs was rapidly oxidised resulting from X-ray irradiation during the experiments. Generally, the valence states of redox-sensitive elements including iron in glass are not strictly fixed and can change with X-ray irradiation . In fact, iron oxidation was also observed depending on X-ray irradiation in our STXM measurements in spite of the sample chamber being purged with 0.1 atm of helium. Accordingly, iron in the CsMPs must be immediately oxidised when irradiated by X-ray in an ambient atmosphere. The presence of only divalent iron can constrain the oxygen fugacity of the reactor atmosphere during CsMP formation from a thermodynamics perspective 24 . Nevertheless, it may be possible that the CsMPs analysed in the present study have different properties from those reported in the previous research, because the numbers of CsMPs measured in this and previous studies are limited. STXM analysis of iron also showed that the rim of both CsMPs consists of trivalent iron regardless of the X-ray irradiation. One possible explanation is that iron in the rim regions was oxidised in the environment. Because CsMPs with cassiterite (SnO 2 ) precipitation were previously reported 10,25 , the surface condition of CsMPs can be modified in the environment after release from the reactor. Another possible explanation is that the reactor atmosphere changed from reductive to oxidative during CsMP formation. We previously reported CsMPs with a higher concentration of Fe and Zn near the surface 11 and such an inhomogeneous distribution can also occur via atmospheric change because divalent cations in silicate glass tend to diffuse outward when the glass is exposed to an oxidative atmosphere after glass formation in a reductive atmosphere 26 . To explain the presence of trivalent iron at the rim regions, more CsMPs should be examined to confirm whether it is a general feature of CsMPs. www.nature.com/scientificreports www.nature.com/scientificreports/ Both STXM and EDS analyses confirmed that Na is among the major constituent elements in the CsMPs. The possible origin of the Na is presumed to be seawater injected for cooling during the accident. In this case, it is certain that the seawater was heated and concentrated until sea salt precipitation occurred because halite was found inside the CsMPs. But it is still possible that Na may have originated from other substances such as structural materials. The existence of Na in CsMPs can also explain the reason why Fe and Zn were not precipitated as a separate phase but were dissolved in the glass matrix. It is known that an immiscibility gap exists in binary RO-SiO 2 systems in silicate melt, but a small amount of alkali elements can eliminate the immiscibility 27,28 . Although the glass matrix of CsMPs contains a certain amount of K and Cs as alkali elements, many CsMPs show a lower alkali concentration near the centre 10 . The synthetic glass which simulated the composition of CsMPs was segregated into (Si and alkali)-rich and (Fe and Zn)-rich phases in a previous study 10 but this phase separation can be prevented by adding Na to the starting materials. Furthermore, the iron in the glass synthesised at an ambient atmosphere was mainly trivalent and the immiscibility is larger for trivalent compared to divalent iron 27 . Therefore, the glass synthesis under a controlled redox atmosphere and containing divalent iron also leads to elimination of immiscibility. If the glass which is more similar to CsMPs can be synthesised based on new information obtained in this study, such a synthetic glass can be used to infer the various properties of CsMPs. For instance, thermal and dissolution behaviour of CsMPs, which have been investigated by laboriously collecting actual CsMPs , can be more easily and accurately elucidated. Furthermore, the atmosphere inside the damaged reactors during the accident can be estimated by determining the conditions for synthesising the same glass as CsMPs, providing useful information on the debris materials to be removed during decommissioning processes. ## Methods Collection of the CsMPs. CsMPs were collected from non-woven fabric cloth laid on a vegetable field approximately 25 km west-northwest from the FDNPP, "Nakadori" area in Fukushima Prefecture. The cloth was left outside for approximately 6 months following the accident. The cloth was then cut into fragments 15 × 15 mm in size and exposed to an imaging plate (BAS-2500, Fujifilm) for 10 min to confirm the presence of CsMPs as bright spots in the readout images. The isolation of individual CsMPs was conducted using a previously described method 32 . The CsMPs attached to the cloth were transferred to 10 mL of ion-exchanged water via ultrasonication. After removing the cloth from the water, the water was divided into 0.5-mL aliquots and the aliquots of water containing CsMPs were identified by measuring the radioactivity using an automatic gamma counter (Wizard2480, PerkinElmer). Ten millilitres of ion-exchanged water were added to each selected aliquot of water and these aliquots were again divided into 20 aliquots of 0.5 mL. The aliquots containing CsMPs were identified again. This process was repeated several times to isolate individual CsMPs from other unrelated particles. The water suspending the CsMPs was dropped onto a piece of double-sided Kapton tape (P-224, Nitto) and dried at ambient temperature. The tape was carbon coated and the CsMPs were searched using a scanning electron microscope (SEM; S-4500, Hitachi) equipped with an EDS spectrometer (Sigma, Kevex). SEM images of the CsMPs are shown in Fig. S6 and the radioactivity determined with imaging plates is appended to the images. The identified CsMPs were thinned to an X-ray and electron-transparent thickness using an FIB system with a micro-sampling unit (FB-2100, Hitachi). ## Glass synthesis. Glass with a similar composition to that of the CsMPs was synthesised as a reference material. The synthesis processes were the same as previously reported 10 but we used a platinum instead of an alumina crucible and reduced the amount synthesised at one time for more rapid quenching. The target composition was 0.6K 2 O-0.6Rb 2 O-10.2Cs 2 O-10.9ZnO-7.7Fe 2 O 3 -3.3SnO 2 -66.7SiO 2 (wt%), which is the averaged composition of the CsMPs reported in our previous study 10 but without chlorine. The resulting glass was homogeneous and no phase separation was observed as shown in Fig. S7. A fragment around the centre of the crucible was crushed in an agate mortar and dispersed on carbon tape. The tape was carbon coated and the glass was thinned to an X-ray and electron-transparent thickness in the same manner as that of the CsMPs. After the glass was crushed and dissolved in HF solution, the composition of the synthetic glass was estimated using ICP analysis. The Cs atoms was measured by ICP-mass spectrometry (SCIEX-ELAN 6000, PerkinElmer), and the other atoms were determined by ICP-optical emission spectrometry (SPS3520, Hitachi) 57 .Fe-Mössbauer measurements were also performed for the crushed glass in a conventional transmission mode at room temperature on a Mössbauer spectrometer (Model-222, Topologic System) with a 57 Co(Rh) source (925 MBq). Curve fitting of the Mössbauer spectra was performed by a nonlinear least-squares method using a MossWinn 4.0Pre program, assuming that all spectra were composed of Lorentzian-shaped peaks. The isomer shift and Doppler velocity scale were calibrated with respect to the sextet of α-Fe at room temperature. STXM. STXM analysis was conducted in transmission mode using a compact STXM system installed at BL-13A in the Photon Factory, KEK (Tsukuba, Japan) 33 . The thin samples were placed on a piezo-driven stage in a chamber purged with 0.1 atm of helium. The samples were scanned using an X-ray beam focused with a Fresnel zone plate. Mapping of transmitted X-ray from the pre-to post-edges of the interested elements was recorded using the image stacking method 34 . XANES spectra at each edge were obtained using the aXis2000 software. The Fe oxidation state maps were acquired using a SVD method in the software. (S)TEM. The thin specimens of the CsMPs were observed using a TEM (JEM-2010, JEOL) operated at 200 kV to characterise the inner structures of the CsMPs. EDS analysis was conducted using a STEM (JEM-2800, JEOL) operated at 200 kV equipped with a silicon drift detector (X-Max N 100 TLE, Oxford Instruments). The probe size and current were ~1 nm and ~1 nA, respectively, and the collection angle (2θ) of the annular dark-field (ADF) detector was 46-208 mrad. Sample thickness measurement using EELS was also conducted using a spectrometer (Enfina, Gatan) attached to the STEM. The collection semi-angle was ca. 27 mrad and the full-width at half-maximum of the zero-loss peak was ca. 1.2 eV.

chemsum

{"title": "Reactor environment during the Fukushima nuclear accident inferred from radiocaesium-bearing microparticles", "journal": "Scientific Reports - Nature"}

discovery_of_cryptic_allosteric_sites_using_reversed_allosteric_communication_by_a_combined_computat

## Abstract: Allostery, which is one of the most direct and efficient methods to fine-tune protein functions, has gained increasing recognition in drug discovery. However, there are several challenges associated with the identification of allosteric sites, which is the fundamental cornerstone of drug design. Previous studies on allosteric site predictions have focused on communication signals propagating from the allosteric sites to the orthosteric sites. However, recent biochemical studies have revealed that allosteric coupling is bidirectional and that orthosteric perturbations can modulate allosteric sites through reversed allosteric communication. Here, we proposed a new framework for the prediction of allosteric sites based on reversed allosteric communication using a combination of computational and experimental strategies (molecular dynamics simulations, Markov state models, and site-directed mutagenesis). The desirable performance of our approach was demonstrated by predicting the known allosteric site of the small molecule MDL-801 in nicotinamide dinucleotide (NAD + )-dependent protein lysine deacetylase sirtuin 6 (Sirt6). A potential novel cryptic allosteric site located around the L116, R119, and S120 residues within the dynamic ensemble of Sirt6 was identified. The allosteric effect of the predicted site was further quantified and validated using both computational and experimental approaches. This study proposed a state-of-the-art computational pipeline for detecting allosteric sites based on reversed allosteric communication. This method enabled the identification of a previously uncharacterized potential cryptic allosteric site on Sirt6, which provides a starting point for allosteric drug design that can aid the identification of candidate pockets in other therapeutic targets. ## Introduction Protein allostery, which is considered as the second secret of life, has paved a novel avenue for modern drug discovery. Allostery involves fne-tuning the functions of conserved orthosteric sites by topologically distinct allosteric sites. Allosteric ligands exhibit higher selectivity and better physiochemical properties than orthosteric ligands, which offers a new paradigm for innovative therapeutic development. Allostery has enabled the successful modulation of intractable drug targets, such as oncoprotein Ras, recalcitrant kinases, and G-protein-coupled receptors (GPCRs). The prerequisite for allosteric drug design is the precise identifcation of allosteric sites. However, most characterized allosteric sites were serendipitously discovered through exhaustive and tedious experimental approaches, such as high-throughput screening and alanine-scanning mutagenesis. 2, Computational tools, such as AlloFinder and Ohm can overcome the limitations associated with allosteric site identifcation. 17,18 However, the computational methods generate artifacts due to the use of potentially biased structural data for model training and nonphysiological dummy atoms or ligands for site detection and docking. 14,15,19,20 Moreover, the detection of cryptic allosteric sites, which can expand the repertoire of allosteric drug discovery, is not possible using these computational methods. The cryptic allosteric sites are exclusively detected in specifc intermediate states within protein conformational ensemble. In addition to the classical allosteric communication theory according to which the signals propagate from the allosteric sites to the orthosteric sites, a bidirectional allosteric communication theory has recently been proposed. According to this theory, the communication signals can also propagate from the orthosteric sites to the allosteric sites, which is referred to as reversed allostery or reversed allosteric communication. Several studies have validated this theory by examining a set of classical allosteric proteins. For example, novel allosteric sites on 15lipoxygenase and allosteric modulators targeting phosphoinositide-dependent protein kinase 1 (PDK1) were successfully identifed based on the bidirectional allosteric communication theory. 24,28,29 The reversed allosteric communication model facilitates the ab initio identifcation of allosteric sites without prior knowledge of the orthosteric sites. However, the limitation of this model is the artifcial nature of manually introduced orthosteric perturbations. Following a similar stream, we propose a computational framework for allosteric site predictions through the reversed allostery model. This framework is based on orthosteric ligand binding, which is the most common and natural orthosteric perturbation, to minimize the generation of potential artifacts. The developed framework can facilitate unbiased detection of allosteric sites under physiological conditions and delineate the crosstalk between orthosteric and allosteric sites. This model was further applied to investigate Sirt6, a key member of the sirtuin deacetylase family for which there are limited regulatory molecules. Sirtuins, which are nicotinamide adenine dinucleotide (NAD + )-dependent class III histone deacetylases (HDACs), deacetylate histones and mediate various cellular events ranging from maintenance of genomic stability to metabolic activities. Sirt6, a member of the sirtuin family (Sirt1-7), catalyzes the removal of acetyl groups from the lysine residues 9, 18, and 56 on histone 3 (H3K9, H3K18, and H3K56) and transfers them to NAD + cofactor. This reaction yields nicotinamide and a combination of 2 0 -and 3 0 -O-acetyl-ADP ribose as products. 31, Sirt6, which is involved in various biological processes, such as DNA repair and aging, is expressed in different mammalian organs. Thus, Sirt6 is a potential therapeutic target. 38 There are limited selective compounds for drugging Sirt6 owing to the highly conserved structural architecture among its paralogues. This has prompted the development of Sirt6 allosteric modulators for specifc therapeutic applications. 34, Sirt6 shares an evolutionarily conserved catalytic core with other sirtuin members, which comprise eight a-helices and nine b-sheets with a large Rossmann fold and a small zincbinding domain (ZBD) (Fig. 1). 40,42 These two domains flank the Sirt6 active cleft, where a long hydrophobic channel loads acetylated substrates and NAD + . Recently, we identifed an allosteric site behind the orthosteric site of Sirt6 and discovered a frst-in-class Sirt6 allosteric activator (MDL-801) for this site (Fig. 1). 43,44 Biochemical analysis of this activator revealed that it is robustly suppressed in cancer models and that it improves the quality of induced pluripotent stem cells derived from aging mice. These fndings highlight the potential of Sirt6 allosteric drug development. In this study, we developed a novel hybrid computational pipeline based on reversed allosteric communication and successfully implemented it for the prediction of Sirt6 allosteric sites. We hypothesized that orthosteric ligand binding shifts the conformational ensemble of Sirt6 and fne-tunes the conformation of the surface cavities that do not overlap with its orthosteric site, which results in the emergence of cryptic allosteric sites. Large-scale molecular dynamics (MD) simulations and Markov state model (MSM) analysis revealed that the dynamic landscape of Sirt6 is markedly altered upon binding of NAD + to the orthosteric site with the emergence of two novel intermediate substrates. Structural analysis revealed that the orthosteric NAD + site modulates the known allosteric MDL-801 site, which supported the rationale of our model. Importantly, NAD + -loaded Sirt6 exhibits a novel potential cryptic allosteric site (Pocket X) with marked allosteric effects in its intermediate conformation. The allosteric crosstalk between Pocket X and the orthosteric NAD + site was elucidated using energetic calculations and correlation analysis. 18,48 Furthermore, the results of site-directed mutagenesis demonstrated that mutations in Pocket X allosterically affect the deacetylation activity of Sirt6, which confrmed the accuracy and reliability of our proposed computational framework. Thus, we propose a combined computational framework based on reversed allosteric communication for allosteric site prediction. The method developed in this study revealed a potential cryptic allosteric site on Sirt6, which can aid future rational allosteric drug design. ## Orthosteric NAD + binding alters Sirt6 dynamic ensemble and generates new intermediate substrates To initiate our computational pipeline for reversed allosteric communication, a 500 ns accelerated MD (aMD) simulation was performed for apo-and holo-Sirt6 in fve replicas to enhance sampling and cover broader conformational spaces (see Materials and methods). The Rossmann fold and the ZBD are critical for the functions of Sirt6. 42 Hence, they were selected as two structural features to capture the conformational dynamics of Sirt6 throughout the simulations. The root-mean-square deviations (RMSDs) in these two domains were calculated and projected onto a two-dimensional (2D) space to visualize the overall free energy landscapes (Fig. 2). The apo-Sirt6 form exhibited two dominant conformations (Fig. 2A) with RMSDs of approximately 0.8 in the Rossmann fold and 1.5 and 3 in the ZBD. Orthosteric NAD + binding markedly altered the Sirt6 conformational ensemble (Fig. 2B). The basins for conformational clusters in NAD + -bound holo-Sirt6 slightly shifted along the X-axis. The RMSDs (approximately 0.7 ) in the Rossmann fold of NAD + -bound holo-Sirt6 were lower than those in the Rossmann fold of the apo-protein. Two novel intermediate states were observed with RMSDs of approximately 4 and 6.5 in the ZBD. The high RMSD values in the ZBD of the holoprotein suggested conformational rearrangements and indicated that orthosteric binding triggered reversed allosteric communication. To examine the key conformational states of Sirt6 during simulations, MSM, which has been widely used to probe protein conformational dynamics, 22,49,50 was constructed for our simulated systems using PyEMMA 51 (see Materials and methods). The results of the implied timescale test (Fig. S1 †) and the Chapman-Kolmogorov test (Fig. S2 †) confrmed that our models were Markovian (see Materials and methods). Conformational ensembles for apo-and holo-Sirt6 were then clustered into 2 (M1 and M2) and 4 (M1 0 , M2 0 , M3 0 , and M4 0 ) MSM metastable states, respectively (Fig. 2A and B). Comparative analysis of these dominant conformations revealed that the overall Sirt6 backbone structure was stable throughout the simulations irrespective of NAD + binding. However, the a3-helix in both systems underwent remodeling (Fig. 2C and D). The most prominent conformational rearrangements were observed in the ZBD, which was consistent with our conformational landscape analysis. The structural feature of ZBD, which mainly comprises coils, loops, and three b-sheets (b4-b6), indicated its high plasticity during simulations. In apo-Sirt6, the MSM metastable states M1 and M2 positioned ZBD with outward and inward orientations relative to the overall protein (hereafter designated as ZBD-out and ZBD-in conformations, respectively) (Fig. 2C), respectively. In holo-Sirt6, the conformations of M1 0 and M2 0 were similar to those of M1 in apo-protein with both exhibiting a ZBD-out conformation. The newly formed substates M3 0 and M4 0 exhibited ZBD-in conformation but projected ZBD further inward. The loop connecting b5 and b6 was transformed into a short a-helix (Fig. 2D). The conformational landscapes and MSM analyses highlighted that orthosteric NAD + binding induced marked conformational changes in the Sirt6 ensemble with predominant changes in the ZBD. ## Detecting potential allosteric pockets using the reverse allosteric communication model The motions and dynamics of the orthosteric and allosteric sites are highly correlated. 26,27 Perturbations at orthosteric sites can fne-tune the overall protein conformational dynamics and influence distal regions, which may result in the emergence of novel allosteric sites. 26,27,52 As the conformational ensemble of Sirt6 was markedly altered upon orthosteric NAD + binding, we investigated its surface cavities to identify potential allosteric sites. Fpocket was used for pocket prediction on all six MSM metastable state structures (i.e., M1, M2 and M1 0 -M4 0 ) 53,54 to retrieve novel binding cavities. In NAD + -bound Sirt6, we frst identifed the previously reported pocket for the allosteric activator MDL-801 (ref. 43) (Fig. 3A). Surprisingly, this pocket, which was detected as Pocket 1, was not completely formed during the simulations (Fig. 3B). The entrance of the pocket was partially blocked by the flexible N-terminus, which may be only displaced upon allosteric modulator loading through an "induced-ft" mechanism. These observations underscored the dynamic nature of the MDL-801 allosteric site, which underwent structural rearrangement through ligand-pocket interaction and was transformed into a suitable conformation to accommodate its corresponding allosteric ligands. These results also demonstrate the rationale of our prediction model. Interestingly, a novel binding site, named Pocket X, was iden-tifed, which emerged because of conformational remodeling caused due to NAD + binding (Fig. 3C). This novel cavity mainly comprises a5-helix, a loop linking a3and a4-helixes, and ZBD residues. This new pocket was not visible in the Sirt6 crystal structure in which a part of the C-terminus was covered above the pocket that hindered its formation (Fig. 3D, S3B and C). Pocket X was detected only in the newly emerging M4 0 MSM metastable state but not in the M1 0 -M3 0 metastable states of holo-Sirt6. The C-terminal coil structure was induced into an ahelix and displaced, which resulted in the exposure of this novel pocket. This indicated the dynamic nature of Pocket X (Fig. 3C, D and S3D-F). Pocket X was also not detected in the apo-Sirt6 MSM metastable structures (M1 and M2), which indicated that orthosteric binding was a prerequisite for the formation of this pocket (Fig. 4A and B). This further provided evidence for the reversed allostery theory. Pocket X, which is situated behind the orthosteric pocket for NAD + , exerted allosteric effects (Fig. 4C). The dynamic behavior of Pocket X emphasizes the feasibility of reversed allosteric communication triggered by orthosteric ligand-protein interactions. Additionally, reversed allosteric communication resulted in the emergence of novel cryptic allosteric sites, whose formation was induced by conformational rearrangements resulting from reversed allosteric communication caused by orthosteric NAD + binding. ## Characterizing the coupling between orthosteric and allosteric sites Next, we established a computational scheme to quantify the reversed allosteric communication caused by orthosteric binding and confrmed the coupling between the orthosteric site and the predicted allosteric sites. A previous study established the residue-residue interaction model in which some residue pairs within the allosteric sites exhibited substantial interaction energy changes upon ligand binding. 52,60 Based on this model, we derived a quantitative model to evaluate allosteric coupling between the orthosteric and allosteric sites. In the reversed allostery model, we speculated that the potential allosteric sites exhibited similar energetic properties in response to orthosteric ligand binding. The residue-residue interaction free energy of a proportion of residue pairs within the allosteric sites would undergo considerable changes before and after orthosteric loading, which suggested reversed allosteric communication due to orthosteric perturbations. To utilize the differences in residue interaction energy as predictors of potential allosteric sites, a quantitative measurement was needed to set up the threshold to differentiate between the allosteric and non-allosteric sites. Based on a previous workflow, 52,60 we divided all residue pairs within a given pocket into the following three groups according to their interaction energy differences before and after orthosteric binding: small energy difference (I), medium energy difference (II), and large energy difference (III). The energy differences between the residue pairs in the small energy difference (I) group are within one standard deviation of the overall average value (between pink and cyan lines in Fig. 5A and B), whereas those in the medium energy difference (II) group were within three standard deviations (between cyan and green lines in Fig. 5A and B). In the large energy difference group (III), the energy differences between the residue pairs were distributed at least three standard deviations away from the average interaction energy changes (outside the green lines in Fig. 5A and B). The ratio of the number of residue pairs in the (III) group to that in the (II) and (III) groups was calculated as the energy coupling score, which indicated the coupling between orthosteric and allosteric sites through energetic terms. The energy coupling score was used to identify potential allosteric sites with 0.25 as the threshold, which was selected based on previous studies. 52,60 Pockets with energy coupling scores higher than 0.25 were considered as allosteric sites, while those with scores lower than 0.25 were considered as non-allosteric sites. Among the prediction results from Fpocket based on the conformational ensemble obtained after orthosteric binding, we tested our method with the recently reported MDL-801 allosteric site on Sirt6. The energy coupling score for the allosteric MDL-801 site was 0.28, which was higher than the cutoff score of 0.25 (Fig. 5A). This was consistent with the results of a previous experimental study that reported the allosteric nature of the MDL-801 binding site. 43 Thus, our fndings demonstrating the reliability and accuracy of our "energy coupling" approach for identifying novel allosteric sites and quantifying reversed allosteric communication. This approach was further applied to analyze the newly predicted Pocket X. The energy coupling score for Pocket X was 0.33 (Fig. 5B), which was higher than the threshold score for the allosteric site. This confrmed that Pocket X was strongly coupled with the orthosteric NAD + binding site and that Pocket X is allosterically regulated. Most cavities predicted by Fpocket had an energy coupling score lower than the threshold score (0.25) and thus were excluded from further analyses (Fig. S4 †). This indicated the sensitivity of our approach. The energy coupling score only deciphered the crosstalk between orthosteric and allosteric sites through energetics. Next, we delineated the structural basis underlying the reversed allosteric communication between the allosteric Pocket X and the orthosteric NAD + binding site. Several cutting-edge computational algorithms were applied for allosteric analyses of Pocket X. Among them, Ohm, a newly developed networkbased method, 18 was used to characterize the allosteric signals originating from Pocket X. The allosteric coupling intensity of the residues in Pocket X was high (0.7866 AE 0.02609). Thus, these residues were considered allosteric hotspots. This indicated that Pocket X exerted a marked allosteric effect on the orthosteric NAD + cavity, which concurred with our results. Furthermore, D3Pocket, a webserver for cavity dynamic analysis, 48 was employed for structure-based evaluation of allosteric coupling. During the simulations, a strong positive volume correlation (0.65) was observed between the orthosteric NAD + site and Pocket X. The dynamics of Sirt6 involving increased volume of the orthosteric NAD + site were associated with a concomitant increase in the volume of allosteric Pocket X (Fig. 5C). This was consistent with our thermodynamic analysis of the energy coupling score. Additionally, this fnding indicated the correlation between the orthosteric binding site and allosteric Pocket X, which suggested bidirectional allosteric communication. ## Pocket X negatively modulates the activity of Sirt6 The aforementioned computational analysis of reversed allosteric communication from an orthosteric NAD + site identifed Pocket X as a novel allosteric site on Sirt6. To further corroborate the allosteric regulation of Sirt6 deacetylation, we performed site-directed mutagenesis experiments. The Fluor de Lys (FDL) assay 36,61 was conducted to quantify the effect of different single mutations (S86A, R88A, P89A, T90A, L116A, R119A, S120A, D149A, and T150A; Fig. 6A) on the Sirt6 deacetylation activity on the acetylated substrate peptide (RHKK-ack- AMC). Sirt6 with mutant Pocket X exhibited approximately twofold lower deacetylation activity than wild-type Sirt6 (Fig. 6B). Additionally, Sirt6 with mutations at the L116, R119, and S120 residues in Pocket X exhibited 4-8-fold lower deacetylation activity than wild-type Sirt6 (Fig. 6A and B). These results validated that Pocket X was an allosteric pocket that negatively modulates the deacetylation activity of Sirt6. Thus, Pocket X can be potentially utilized for designing Sirt6 allosteric modulators in the future. Moreover, these experimental fndings demonstrated the desirable predictive efficacy of our computational reversed allostery model, which represents a novel promising methodology for allosteric site identifcation and allosteric drug design. ## Discussion In this study, we proposed a novel computational pipeline based on reversed allosteric communication for allosteric site prediction. HDAC Sirt6 was used to demonstrate the ability of the computational pipeline to predict the previously reported allosteric MDL-801 site. Additionally, a potential novel cryptic allosteric pocket was identifed, which only formed within a newly induced MSM metastable state upon orthosteric NAD + binding. Bidirectional signals between orthosteric and allosteric sites were confrmed using both energetic and structural dynamics computations, as well as site-directed mutagenesis. This approach provides evidence for the protein allostery theory and enables allosteric drug discovery for intractable drug targets, including Sirt6. Modern medicine is based on improved selectivity and enhanced pharmacological performances. Thus, allosteric modulators are applied for drug discovery 1,4,20,62,63 with multiple allosteric candidates targeting the historically "undruggable" targets, such as K-Ras, and Src homology 2 phosphatase (SHP-2), which may have clinical applications. 8,9,11, Previously, allosteric drugs were identifed mainly through serendipitous discovery due to difficulties in the identifcation and characterization of their binding sites. 2,14,15,20 The rapid development of computational tools to predict allosteric sites partially enabled overcoming these limitations. 17,68-70 However, most sites predicted using these computational tools were associated with several limitations due to the nature of their computational models or algorithms. For example, data-driven prediction methods are frequently limited due to the use of biased training datasets in which most allosteric sites are identifed from kinases and GPCRs, which hinders the performance of the prediction methods. Normal mode analysis-based approaches are restricted by their relatively coarse analysis procedures with limited accuracy and specifcity during approximations. 2,15,17,19,69,70 Traditionally, the allosteric effect is defned as the modulation of the orthosteric sites by the allosteric sites. Recently, a bidirectional model for allostery was proposed according to which the signals propagate from the orthosteric sites to the allosteric sites. This theory has been validated through various studies that discovered a novel allosteric site on 15-lipoxygenase and designed allosteric modulators for PDK1. 24,28,29 The possibility of generating artifacts from nonphysiological orthosteric perturbations and other ligandpocket interactions cannot be ruled out. To address these issues, we developed a combined computational framework exploiting the reversed allosteric communication triggered by natural orthosteric ligand binding to detect potential allosteric sites within the protein dynamic ensembles. As shown in Fig. 7, we began with the apo-and holo-protein structures. MD simulations and MSM analysis were performed to probe the conformational ensemble of proteins in the absence and presence of orthosteric ligand binding. The dominant metastable state conformations were clustered and extracted. Potential allosteric sites can be predicted based on the holo-protein conformations, and their correlations with orthosteric sites can be further quan-tifed through both structural and energetic analyses. Finally, the computational predictions were experimentally validated through site-directed mutagenesis. Compared with the previous reports, the workflow of this study precludes most artifacts by physiologically manipulating orthosteric sites through their corresponding ligand binding instead of mutations or manual perturbations. Additionally, large-scale MD simulations and cutting-edge techniques, such as MSM analysis, facilitated the identifcation of subtle metastable states within a protein dynamic ensemble. Thus, this approach can provide comprehensive insights into protein conformations. The detailed analysis of the alterations of protein conformation upon orthosteric ligand binding can reveal cryptic allosteric sites, which are not detected in most crystal structures and are present only in some conformational metastable states. The identifcation of cryptic allosteric sites, which are potential targets for allosteric drug discovery, 21,23,71 will expand the spectrum of available druggable targets. Our approach successfully predicted a cryptic allosteric site on Sirt6, a critical target that is involved in various physiological and pathological processes, such as longevity and carcinogenesis. However, there is a lack of specifc therapeutic agents for Sirt6. 38 Sirt6 belongs to the "magnifcent seven" sirtuin family (Sirt1 to Sirt7). The major challenge to identify drugs for Sirt6 is to achieve specifcity among the seven different isoforms. Thus, the identifcation of selective Sirt6 allosteric modulator has piqued the interest of the scientifc community. 34, Previous studies have achieved, limited success in identifying allosteric drugs targeting Sirt6. In this study, we identifed a potential new exosite Pocket X with allosteric potential based on reversed allosteric communication. This will enable the identifcation of Sirt6 allosteric modulators in the future. Previous studies have demonstrated that Sirt6 inhibitors can be used to treat metabolic syndromes, such as diabetes, whereas Sirt6 activators can be used to treat cancers and aging-related diseases. 33,34,43, The novel allosteric site discovered in this study markedly regulated the deacetylation activity of Sirt6. This indicated that this site could be a potential target for Sirt6 allosteric drug design. These fndings illustrate the potential of the novel Sirt6 allosteric regulatory site identifed using the reversed allosteric communication model for allosteric drug discovery. Although the novel pocket identifed using the workflow developed in this study exhibited allosteric activity, the corresponding allosteric ligands for this site were not identifed. Hence, further studies on the allosteric site structure and therapeutic agents that target this site using compound screening and crystallography are needed for allosteric modulator discovery. This will complement the other half of allosteric drug design and enhance its translational signifcance. ## Conclusions Here, we described a novel computational workflow using the reversed allosteric communication model based on orthosteric ligand binding to predict the allosteric sites from the protein conformational ensemble. The accuracy of this methodology was validated using the previously known allosteric site for MDL-801 on Sirt6. A potential new allosteric site for Sirt6 was identifed for further investigation. The method described in this study represents an up-to-date advance for protein allostery, which has potential applications in designing novel therapeutics. Moreover, the exosite on Sirt6 is a potential starting point for future rational allosteric drug design. ## Construction of simulated systems The co-crystal structure (2.53 ) of human Sirt6 in complex with cofactor ADPR, substrate H3K9 myristoyl peptide, and allosteric activator MDL-801, which was previously solved by our research group, was downloaded from the RCSB Protein Data Bank (PDB ID: 5Y2F). 43 The myristoylated substrate peptide in 5Y2F was truncated to represent the acetylated substrate peptide after removing its 12-carbon long aliphatic chain. To model the active reactant NAD + -bound Sirt6, the atomic coordinates of NAD + , which were extracted from the crystal structure of the Sirt1-NAD + complex (PDB ID: 4I5I), 72 were docked to replace the ADPR coordinates in 5Y2F. ## Docking simulations Molecular docking was performed using AutoDockFR-Auto-Dock. 73 Briefly, the structures of Sirt6-acetylated substrate complex and NAD + were converted to PDBQT fles using Auto-Dock Tools and loaded into AutoGridFR separately. The centroids of ADPr were set as the grid center with a search space of 60 60 60 using a spacing value of 0.375 between every two grids. The atomic affinity maps, which were generated using AutoGridFR, were used as input into AutoDockFR for protein-ligand docking. Fifty independent docking runs were performed to obtain multiple binding modes. During these docking simulations, the side chain of acetylated Lys9 (K9Ac) in the substrate peptide was set as a flexible receptor residue. The resulting docking poses were clustered with an RMSD threshold of 2 . The lowest energy-docking pose from each cluster was then visually determined by comparing the distance of the anomeric carbon of the nicotinamide ribose group from NAD + and the acetyl oxygen atom of K9Ac. Additionally, the conformational arrangement was assumed necessary for promoting an S N 2 nucleophilic substitution reaction. ## Conventional MD (cMD) simulations The Amber ff14SB force feld was employed to model the Sirt6 structure 74 and revised parameters were assigned for the Sirt6binding Zn 2+ . 75 A TIP3P truncated octahedral water box (10 ) was then applied for solvation, while the counterions were added for neutralization. 76 To mimic the physiological conditions for proteins, 0.15 mol L 1 NaCl was randomly added to the complex system. Complex systems underwent two rounds of energy minimization after preparation. The energies of water and counterions were minimized in 5000 steps (2000 steepest descent and 3000 conjugate gradient minimization steps) with a position restraint of 500 kcal (mol 1 2 ) on the protein backbone. Next, 4000 steepest descent and 6000 conjugate gradient cycles of minimization were applied to the whole system without any position restraint. Subsequently, under NVT condition and restraint of 10 kcal (mol 1 2 ) on Sirt6, each system was heated to 300 K in 300 ps after which 700 ps equilibration runs were performed. Finally, to obtain the energy data for aMD, 50 ns cMD simulations with a timestep of 2.0 fs were performed under 1 atm and 300 K. Langevin dynamics with a collision frequency of 1 ps 1 were applied to regulate the system temperature, while the system pressure was regulated by a Berendsen barostat with isotropic position scaling. During cMD, the particle mesh Ewald method was used to model the long-range electrostatic interactions, 77 whereas a cutoff value of 10 was set for short-range electrostatic interactions and van der Waals force calculations. Additionally, the SHAKE algorithm was applied to restrain all covalent bonds containing hydrogen atoms. 78 ## aMD simulations The aMD simulations were performed after a 50 ns cMD simulation to enhance sampling. The principle of aMD is that "boost energy" was introduced into the system potential energy surface when the corresponding potential energy is lower than the predetermined threshold. The boosting energy decreases energy barriers between conformations and flattens the free energy landscape, which accelerates the conversions between different stable conformers. 79 As shown in eqn (1) and (2), the system potential energy was V(r) and E thresh was the threshold determined by the system size and normal energy value. If the instantaneous potential energy was lower than E thresh , it would increase by DV(r) and reach V*(r). In other cases, the origin value V(r) was used as V*(r) in aMD simulation. V*(r) ¼ V(r); V(r) $ E thresh (1) where DV(r) is defned in eqn (3) depending on E thresh and the acceleration parameter a During aMD simulation, both potential energy and dihedral energy surfaces were boosted, which is referred to as the "dual boost" protocol. 80 Eqn (4)-( 7) determined the parameters of potential and dihedral energy. Among them, E threshP denotes the energy threshold for potential energy, while E threshD represents the energy threshold for dihedral energy. The a P and a D values represent the acceleration parameters for potential and dihedral energy, respectively. cMD simulations retrieved the average total energy (E tot avg ) and the average dihedral energy (E dih avg ) for aMD. N atoms and N residues represent the number of atoms and residues in the systems, respectively. a P ¼ N atoms 0.16 kcal mol 1 (5) The last structures of the cMD simulations were set as the starting structure of the aMD simulations. Five independent aMD replicas of 500 ns with random initial velocity were performed for each system (apo versus holo Sirt6), which resulted in a cumulative timescale of 5 ms. The calculation of electrostatic interaction and van der Waals forces, as well as other systemic environmental parameter settings, such as the thermostat and barostat values were consistent with those used during cMD processes. After simulations, trajectory analysis was performed using CPPTRAJ. ## Construction of MSM Based on the coordinates of apo Sirt6, we calculated the RMSDs in the Rossmann fold and ZBD during the apo and holo Sirt6 simulations, which were then used as inputs for the PyEMMA MSM analyses. 51 After validating the Markovian property (please see next part), we used the "coordinates.cluster_kmeans" method to cluster our structural data into 200 microstates with a maximum k-means iteration number of 200. Next, MSMs for apo and holo Sirt6 were constructed using the "msm.estimate_markov_model" function with a lag time of 2 ns. These 200 microstates were further clustered into multiple metastable states using the Perron cluster analysis (PCCA+) algorithm, which was further confrmed by the Chapman-Kolmogorov test (please see next part). In each metastable state, the "coordinates.save_traj" algorithm was used to extract trajectories comprising more than 50% snapshots for the corresponding state. Next, the MDTraj package was applied to capture the representative structures corresponding to each intermediate state according to the similarity score S ij . As shown in eqn (8), d ij is the RMSD between structures i and j, while d scale is the standard deviation of d. 81 After iterating each pair of structures in the trajectory, the structure with the highest S ij was considered as the most representative conformation. Validation of the MSM Based on our 2D free energy landscape depicting the conformation distributions throughout the simulations, an implied timescale test was applied to validate the Markovian properties of our simulation systems. First, we specifed the number of kmeans centers and maximum iteration and then constructed the corresponding multiple transition probability matrixes (TPMs) with different lag times, which represent the time interval between transitions. TPMs account for the possibility of transition between all pairs of microstates and determine the implied timescale (also known as relaxation timescale) through eqn (9). where s is the lag time for the construction of TPMs, l i represents the i th eigenvalue of TPM with a lag time s. When the transition of microstates relaxes for i th time, the implied timescale s i is calculated using eqn (9). Under practical conditions, the implied timescale comes from the eigen decomposition of TPMs, and the sequence of eigenvalues corresponds to the sequence of transition. Thus, the frst eigenvalue (l 1 ) represents the slowest transition. To verify the Markovian properties, we must calculate different s i based on increasing s. If s i (notably for s 1 ) starts to become steady as s increases, the transitions between microstates become independent of the lag times, which concurs with the Markovian model. 82 As shown in Fig. S1, † we calculated the changes in the implied timescale of apo (Fig. S1A †) and holo (Fig. S1B †) Sirt6 with lag time. Both the k-means cluster centers and the maximum iteration times were set to 200. In Fig. S1A, † the s i curves of apo Sirt6 converge to reach the platform from a lag time of approximately 1.25 ns, while those for holo Sirt6 are flattened from a lag time of approximately 0.5 ns (Fig. S1B †). Hence, the lag times longer than 1.25 ns will guarantee the Markovian properties of our systems. We used 2 ns as the lag time for further analysis. According to the Sirt6 conformation landscape distributions, we clustered the microstates of apo Sirt6 into 2 MSM metastable states, while those of holo Sirt6 were clustered into 4 MSM metastable states using the PCCA+ method. The Chapman-Kolmogorov test shown in Fig. S2 † demonstrated that the transition probability estimated based on MSM is highly similar to the practical transition process. 83 Thus, our MSM estimation was also validated. ## Energy coupling score calculation Based on the MD simulation trajectories, molecular mechanics generalized Born surface area (MM-GBSA) energy calculation was performed for the corresponding cavities on Sirt6 in both apo and holo systems. The interaction free energy with a pocket was calculated for energy residues separated by at least three amino acids in the sequence based on the following eqn (10): where E int represents internal energy, E eel denotes electrostatic energy, E vdw indicates van der Waals energy, and G pol and G sas are the polar solvation energy and solvent-accessible surface energy, respectively. Energy differences, which were calculated for pockets in the NAD + -bound (holo) and NAD + -unbound (apo) systems, were used to evaluate the energy coupling scores. ## Mutant plasmid construction and protein purication The full-length wild type (1-355 amino acids) human Sirt6 was cloned into the PET28a vector to express the fusion protein with an N-terminal His-tag. Site-directed mutagenesis (S88A, R90A, P91A, T92A, L118A, R121A, S122A, D151A, and T152A) was performed using the In-Fusion® HD Cloning Kit (Takara), following the manufacturer's instructions. All recombinant plasmids were transfected into the Escherichia coli DH5a cells and extracted using the TIANprep Rapid Mini Plasmid Kit (TianGen). The quantity and quality of the plasmids were measured based on the absorbances at 260 and 280 nm using NanoDrop (Thermo). The plasmid sequences were further confrmed using DNA sequencing. The recombinant plasmids were transformed into Escherichia coli Rosetta (DE3) chemically competent cells (Weidi Biotechnology) for protein purifcation. The recombinant colony was inoculated in 2xYT medium with 50 mg mL 1 kanamycin at 37 C. The expression of the recombinant protein was induced using 0.5 mM isopropyl-beta-D-thiogalactopyranoside when the optical density of the culture at 600 nm (OD600) reached 0.6-0.8. The induction was continued at 16 C for 16-18 h. Next, the cells were lysed at high pressure using Sirt6 lysis buffer (1 phosphate-buffered saline [pH 7.5], 300 mM NaCl, and 5% glycerol). The lysate was loaded onto a nickel column (GE Healthcare) with 1 mM dithiothreitol. A protein inhibitor cocktail was added before high-pressure crushing. After washing with 20, 40, 100 mM of imidazole, the protein was eluted using 250 mM imidazole. The elution was dialyzed into Sirt6 assay buffer (50 mM Tris-HCl, pH 8.0, 137 mM NaCl, 2.7 mM KCl, and 1 mM MgCl 2 ) and used in all in vitro assays in this study. ## FDL assay To measure the deacetylation activity of both wild-type and mutant Sirt6, 5 mM purifed proteins were incubated in a 50 mL reaction mixture (75 mM RHKK-ack-AMC, 2.5 mM NAD + , 10% dimethyl sulfoxide, and Sirt6 assay buffer) at 37 C for 2 h. The reaction mixture was incubated with 50 mL stop buffer mixture (40 mM nicotinamide, 6 mg mL 1 trypsin, and Sirt6 assay buffer) at 25 C for 30 min to quench the reaction. The fluorescence intensity was quantifed using a microplate reader (Synergy H4 Hybrid Reader, Bio Tek) at excitation and emission wavelengths of 360 nm and 460 nm, respectively. Each experiment was independently repeated at least three times. ## Conflicts of interest The authors declare no conflict of interest.

chemsum

{"title": "Discovery of cryptic allosteric sites using reversed allosteric communication by a combined computational and experimental strategy", "journal": "Royal Society of Chemistry (RSC)"}

a_dnazyme-augmented_bioorthogonal_catalysis_system_for_synergistic_cancer_therapy

## Abstract: As one of the representative bioorthogonal reactions, the copper-catalyzed click reaction provides a promising approach for in situ prodrug activation in cancer treatment. To solve the issue of inherent toxicity of Cu(I), biocompatible heterogeneous copper nanoparticles (CuNPs) were developed for the Cu-catalyzed azide-alkyne cycloaddition (CuAAC) reaction. However, the unsatisfactory catalytic activity and off-target effect still hindered their application in biological systems. Herein, we constructed a DNAzyme-augmented and targeted bioorthogonal catalyst for synergistic cancer therapy. The system could present specificity to cancer cells and promote the generation of Cu(I) via DNAzyme-induced value state conversion of DNA-templated ultrasmall CuNPs upon exposure to endogenous H 2 O 2 , thereby leading to high catalytic activity for in situ drug synthesis. Meanwhile, DNAzyme could produce radical species to damage cancer cells. The synergy of in situ drug synthesis and chemodynamic therapy exhibited excellent anti-cancer effects and minimal side effects. The study offers a simple and novel avenue to develop highly efficient and safe bioorthogonal catalysts for biological applications. ## Introduction As an alternative to conventional chemotherapy, prodrug approaches are considered a promising solution to the drug side effects caused by off-target effect. The CuAAC reaction as one of the representative bioorthogonal reactions, has now become a powerful tool for prodrug activation or drug synthesis in situ. However, the intrinsic toxicity of Cu(I) to organisms hampered its application in biological systems. 17 To circumvent the limitation, biocompatible heterogeneous catalysts based on CuNPs were developed for the CuAAC reaction. 18 Unfortunately, the inherent poor activity of Cu(0) makes these reactions often require a long time and the assistance of high temperature, microwave irradiation and additives. 19,20 It has been reported that the surface of CuNPs was more reactive due to the unflled valences of the surface atoms and a large surface-to-volume ratio. 21 Therefore, promoting the conversion of surface Cu(0) to Cu(I) and regulating their size would accelerate the catalytic process. 22 Recently, we designed biocompatible heterogeneous CuNPs for highly efficient CuAAC reaction, in which the valence state change of Cu was achieved by near-infrared light (NIR)induced reactive oxygen species (ROS). 23 However, the dependence of external excitation inevitably suffers from limited tissue-penetrating depth, and has adverse effects on the surrounding normal tissues. Therefore, the endogenous control of the valence state of CuNPs is highly desirable. In addition to unsatisfactory efficiency, the lack of targeting ability of these nanocatalysts may compromise the efficacy of the CuAAC reaction in specifc regions and cause damage to healthy tissues. Although modifcation of targeting molecules, such as triphenylphosphonium, 24 could endow nanocatalysts with increment of specifc localization, it is still a challenge to develop highly efficient and precise targeted bioorthogonal catalysts for biological applications. Peroxidase-mimicking DNAzymes are single-stranded DNA molecules, which adopt a G-quadruplex structure and bind hemin molecules to exhibit high catalytic capability in the presence of H 2 O 2 . They have been widely explored in biology, sensing, and material sciences. Recently, it was reported that the peroxidase-mimicking DNAzyme could produce strong radical species to cut single-wall nanotubes (SWNTs) and degrade graphene oxide into small fragments. 38,39 Inspired by the properties of DNAzyme, herein, we construct a DNAzyme-augmented bioorthogonal catalysis system for synergistic cancer therapy. Using the high local concentration of H 2 O 2 in cancer cells, the DNAzyme could produce active radical species to facilitate the conversion of Cu(0) to Cu(I) on the surface of DNA-templated ultrasmall CuNPs, resulting in much enhanced bioorthogonal catalytic activity and enabling in All publication charges for this article have been paid for by the Royal Society of Chemistry situ prodrug activation. In addition, the aptamer AS1411/heminbased DNAzyme exhibited cancer cell-targeting ability through specifc recognition of nucleolin overexpressed on the surface of cancer cells. Meanwhile, the cancer-killing radical species produced by DNAzyme could offer synergistic chemodynamic effects. The combination of in situ drug synthesis with chemodynamic therapy successfully achieved highly specifc and efficient therapeutic efficacy with minimal side effects. ## Results and discussion The design principle of the system is shown in Scheme 1. Featuring a G-quadruplex structure and specifc recognition of nucleolin overexpressed by cancer cells, 43 the aptamer AS1411 can form peroxidase-mimicking DNAzyme with cancer cell targeting ability. AS1411 is selected and linked with a polythymine (polyT) sequence, which serves as a template for CuNP formation. 47,48 The obtained DNA sequence AS1411linker-polyT was mixed with CuSO 4 and sodium ascorbate to synthesize CuNPs (G-Cu), followed by incorporating hemin to form peroxidase-mimicking DNAzyme (D-Cu) (Fig. S1 †). The combination of DNAzyme with CuNPs yielded a system of predesigned functionalities. The DNAzyme features triple functions: targeting cancer cells, endogenous control of the valence state of CuNPs, and production of cancer-killing radical species. High-resolution transmission electron microscopy (TEM) images show that the as prepared DNAzyme-CuNPs had excellent monodispersity with an average size of 3.5 nm (Fig. 1a-c). D-Cu still maintained a stable particle size distribution in the cell culture medium (DMEM) with 10% fetal bovine serum (FBS), suggesting its stability under physiological conditions (Fig. S2 †). Zeta potentials showed that G-Cu and D-Cu were negatively charged in H 2 O, PBS, and DMEM with 10% FBS (Fig. 1d). The inductively coupled plasma (ICP) data showed the accurate copper content of D-Cu (Table S2 †). D-Cu showed fluorescence emissions at 600 nm (Fig. 1e). X-ray photoelectron spectroscopy (XPS) was conducted to examine the valence states of D-Cu. The spectrum of D-Cu exhibited two typical peaks at 932.6 eV (Cu 2p 3/2 ) and 952.3 eV (Cu 2p 1/2 ), corresponding to Cu(0) (Fig. 1f). These results indicated that the DNA-templated CuNPs were successfully formed. Subsequently, we investigated the peroxidase-mimicking activity of D-Cu via the 3,3 0 ,5,5 0 -tetramethylbenzidine (TMB) oxidation reaction. D-Cu catalyzed TMB into oxTMB in the presence of H 2 O 2 , leading to a color change from colorless to blue and an increase in absorbance at 652 nm. In contrast, D-Cu or H 2 O 2 alone did not produce a color change, suggesting the peroxidase-like activity of D-Cu (Fig. 1g). Furthermore, we systematically studied the catalytic activity of D-Cu under various reaction conditions, including pH, temperature, and different concentrations of D-Cu, H 2 O 2 , and TMB (Fig. S3-S7 †). These results indicated that D-Cu could effectively transform H 2 O 2 into ROS. To confrm the value state conversion effect of DNAzyme, the catalytic performance of D-Cu in the CuAAC reaction in solution was examined. Pro-fluorophore 1 and 2, which were nonfluorescent, were chosen as models (Fig. 2a and S8 †). As shown in Fig. S9, † the mixture of 1 and 2 had no fluorescence (l ex ¼ 340 nm; l em ¼ 460 nm). Upon addition of D-Cu to 1 + 2 for 30 min, the sample emitted cyan-blue fluorescence, indicating that triazole 3 was produced. When D-Cu and H 2 O 2 were added to 1 + 2 for 30 min, the product presented stronger fluorescence, implying higher catalytic efficiency (Fig. 2b). It could be attributed to the higher content of Cu(I) on the surface of D-Cu. The time-dependent fluorescence changes of products generated Scheme 1 Schematic illustration of a DNAzyme-augmented bioorthogonal catalytic system for synergistic cancer therapy. S3 †). The catalytic transformation performance of D-Cu was further measured by high-performance liquid chromatography (HPLC) at 0, 1, 10, 30 and 60 min (Fig. S12 †). The transformation rates were consistent with the results of fluorescence analysis (Fig. 2f). D-Cu maintained catalytic activity under different physiological conditions (Fig. S13 †), indicating its adaptability in biological systems. Furthermore, a nanocatalyst based on a DNA strand without the AS1411 aptamer (WA-Cu) was tested for comparison. The fluorescence spectra showed that the presence of aptamer did not affect the catalytic behavior in vitro (Fig. 2g). The ultrasmall size of CuNPs and the valence state conversion of DNAzyme endowed D-Cu with high catalytic activity. In addition, we used molecular dynamics simulation (MDS) to study whether DNA had an additional contribution to catalytic activity. 49 The DNA structures were frst relaxed via MDS. The root-mean-square deviation (RMSD) 50 values during the simulation suggested that the linker (15-base motif) and AS1411 could reach a stable state in 25 ns, while polyT kept free string style (Fig. S14 †). The two precursors 1 and 2 were docked with the DNA structure in the last frame of simulations. The hydroxyl group of 1 could form two hydrogen bonds with A33 and C40 of the linker sequence. And the compound was stabilized through hydrophobic interactions between 1 and A37 and T45 and T47 (Fig. 2h). The hydrophobic interactions between 2 and C36, A37 and T45 of the linker could stabilize the complex (Fig. 2i). The results indicated that reaction substrates pulled close by DNA contributed to the high catalytic activity. To investigate the targeting capability of D-Cu, a nucleolinpositive cell, HeLa (human cervical cancer cells) was used as a model and a normal cell, HEK-293 (human embryonic kidney cells) was used as a control. First, 3-(4,5-dimethyl-2-yl)-2,5diphenyltetrazolium bromide (MTT) assay was used to test the cytotoxicity of D-Cu. The results revealed that D-Cu was biocompatible for both HeLa and HEK-293 cells evenly at 50 mM (Fig. S15 and S16 †). Then D-Cu and WA-Cu were incubated with HeLa and HEK-293 cells, respectively. Their cellular internalization was monitored using a confocal laser scanning microscope (CLSM) and quantifed using inductively coupled plasma mass spectrometry (ICP-MS). The results indicated that D-Cu could be efficiently endocytosed by HeLa cells and reached the maximum within 4 hours, while a relatively low amount of WA-Cu was endocytosed even within 24 hours (Fig. S17-S19 †). For HEK-293 cells, weak red fluorescence was observed within 24 hours, indicating that very little D-Cu was endocytosed (Fig. S20 and S21 †). These results indicated that AS1411 could boost cellular internalization of D-Cu via receptor-mediated endocytosis. After exploring the specifc recognition capability of D-Cu, we evaluated its catalytic ability in the CuAAC reaction in HeLa and HEK-293 cells. The signals of cell membrane dye-DiO (green), D-Cu (red) and triazole 3 (blue) were monitored by CLSM. As shown in Fig. 3a, HeLa cells treated with precursors and D-Cu presented signifcant red and blue fluorescence enhancement compared with the cells treated with precursors only. The red and blue fluorescence was also much stronger than the cells treated with precursors and WA-Cu. HEK-293 cells treated with precursors and D-Cu showed weak red and blue fluorescence (Fig. 3b). These results indicated that a higher amount of CuNPs as a catalyst was accumulated in HeLa cells via the targeting effect of the aptamer, resulting in a higher transformation rate of precursors into triazole 3. The results of flow cytometry analysis were consistent with those of CLSM. As shown in Fig. 3c, HeLa cells treated with precursors and D-Cu presented about a 60-fold fluorescence enhancement compared with the cells treated with precursors only. However, only 1.6-fold fluorescence enhancement was shown in the HEK-293 cells under the same conditions (Fig. 3d). Furthermore, LC-MS analysis indicated that triazole 3 was synthesized in HeLa cells (Fig. S22 †). The above results demonstrated that D-Cu could achieve DNAzyme-augmented and targeted CuAAC reactions in living cells. Subsequently, we explored whether D-Cu could catalyze prodrugs 4 and 5 to produce the resveratrol (3,5,4 0 -trihydroxystilbene; Rsv) analog 6 (Fig. 4a), a model drug with antitumor activity. It is reported that the Rsv analog 6 can restrain tumor initiation, promotion, and progression via inhibiting the cyclooxygenase activity of COX-1. 51,52 MTT assay showed that the prodrugs 4 and 5 had little cytotoxicity to HeLa and HEK-293 cells even at 50 mM (Fig. 4b and S23 †). The half maximal inhibitory concentration (IC 50 ) values of 4 and 5 against HeLa cells were 74.8, and 77.3 mM, respectively (Fig. S24 †), indicating that the prodrugs were biocompatible at the concentration used in our experiment. After being treated with prodrugs 4 + 5 and G-Cu, the viability of HeLa cells decreased signifcantly compared with those treated with prodrugs 4 + 5. The viability further decreased when the cells were treated with prodrugs 4 + 5 and D-Cu (Fig. 4c). The IC 50 value of prodrugs 4 + 5 and D-Cu against HeLa cells was 18.6 mM (Fig. S24 †), indicating that the synthesized drug 6 had a strong cytotoxic effect. These results demonstrated that D-Cu could effectively catalyze the in situ synthesis of 6 to kill cancer cells. However, HEK-293 cells treated with prodrugs, prodrugs + G-Cu, and prodrugs + D-Cu showed a slight decrease in cell viability (Fig. 4d and S25 †). To further explore the self-augmented properties of D-Cu, we primarily detected the generation of ROS from the high local concentration of H 2 O 2 in cancer cells using the DCFH (2,7-dichlorofluorescein) fluorescent probe. As shown in Fig. S26, D-Cu † could produce a large number of active radical species that boosted the transformation of Cu(0) into Cu(I). The cells were further examined using flow cytometry analysis by double staining. The apoptosis rate of HeLa cells was much higher than that of HEK-293 cells (Fig. 4e). The viability obviously decreased when the HeLa cells were treated with prodrugs and D-Cu. The results demonstrated that cancer cells could be killed by Rsv analogue 6 catalyzed by D-Cu and ROS produced by DNAzyme. Moreover, the synthesis of Rsv analogue 6 in cancer cells was confrmed by HPLC (Fig. S27 †) and LC-MS analysis (Fig. S28 †). The above research proved that D-Cu could promote the intracellular CuAAC reaction, enhancing anticancer efficacy in tumor cells and simultaneously minimizing the side-effects in normal cells. We next monitored the catalytic performance of D-Cu in vivo. Caenorhabditis elegans (C. elegans) was used as a model to investigate the toxicology and phenotype alteration of D-Cu. After being treated with D-Cu, the N 2 wild-type strain worms were totally paralyzed in 16 days, which was consistent with the control group (Fig. 5a and b), suggesting the biosafety of D-Cu in vivo. To systematically investigate the toxicology of D-Cu, hemolytic assay was performed. D-Cu showed little hemolysis even at a concentration of 50 mM (Fig. S29 †). After that, the pharmacokinetics of D-Cu was investigated using HeLa tumor-bearing nude mice via ICP-MS analysis. The results showed that D-Cu had favorable accumulation and retention in the tumor region due to AS1411 with high specifcity and affinity to nucleolin (Fig. S30 †). Meanwhile, the growth of the mice was not affected (Fig. S31 †). Furthermore, the results of hematological analysis, blood biochemical assay and hematoxylin and eosin (H&E) staining of major organs show that D-Cu was nontoxic, which favored its further biological applications (Fig. S32 and S33 †). We further evaluated the therapy efficacy of D-Cu in vivo using the HeLa tumor-bearing athymic nude mouse model (HTANM). The HTANMmice were randomly divided into seven groups: (1) control; 5c) of mice in each group were monitored during the treatment. By comparing the excised tumor mass of each group, admirable antitumor efficacy was observed in group 7 (Fig. 5d). The H&E staining of tumor tissue revealed that the cancer cells in group 7 were extremely damaged. The apoptotic cells increased signifcantly, implying that the tumor was sufficiently suppressed (Fig. 5e). Taken together, based on targeting recognition and self-augmented capability, D-Cu could accumulate at the tumor site and produce ROS, which catalyzed drug synthesis in situ and directly kill cancer cells for synergistic cancer therapy. ## Conclusions In summary, a DNAzyme-augmented and targeted bioorthogonal catalyst was successfully designed and constructed for synergistic cancer therapy. Combining the peroxidasemimicking activity of DNAzyme, specifc recognition of the aptamer and DNA-templated CuNPs, the bioorthogonal catalysis system could be actively delivered to tumor cells and then produce ROS to convert the valence state of CuNPs for enhanced bioorthogonal drug synthesis and to kill cells directly. Compared with the common catalyst CuSO 4 /sodium ascorbate, the catalytic transformation of the system in the presence of H 2 O 2 was increased by more than 20 times in 30 minutes. The system combined in situ drug synthesis with chemodynamic therapy, offering highly specifc and efficient cancer therapy effects with minimal side effects. The study offers a simple and novel avenue to develop efficient bioorthogonal catalysts for biomedical applications.

chemsum

{"title": "A DNAzyme-augmented bioorthogonal catalysis system for synergistic cancer therapy", "journal": "Royal Society of Chemistry (RSC)"}

production_of_monodisperse_polyurea_microcapsules_using_microfluidics

## Abstract: Methods to make microcapsules -used in a broad range of healthcare and energy applicationscurrently suffer from poor size control, limiting the establishment of size/property relationships. Here, we use microfluidics to produce monodisperse polyurea microcapsules (PUMC) with a limonene core. Using varied flow rates and a commercial glass chip, we produce capsules with mean diameters of 27, 30, 32, 34, and 35 µm, achieving narrow capsule size distributions of ±2 µm for each size. We describe an automated method of sizing droplets as they are produced using video recording and custom Python code. The sustainable generation of such size-controlled PUMCs, potential replacements for commercial encapsulated systems, will allow new insights into the effect of particle size on performance.Microcapsules -that is, sub-mm size capsules with a solid shell and a solid or liquid core -have diverse applications across sustainability and energy, in healthcare, and in consumer products [1][2][3][4][5][6][7][8][9] . For example, microcapsules have been used for self-healing anticorrosion coatings 10,11 , energy storage materials 12,13 , and in catalysis 14 . By using encapsulation technology, manufacturers and researchers are able to use much smaller quantities of expensive or harmful ingredients 15,16 , or achieve a controlled release of the liquid core on response to a stimulus [17][18][19] . Despite the utility of polymer microcapsules, there are considerable environmental concerns regarding their persistence in the environment 20 or use of harmful additives such as formaldehyde 21,22 . As the properties of microcapsules -such as release profile, permeability, and stability over time -often depend on particle size [23][24][25] , there is a strong drive to produce monodisperse microcapsules such that robust size/property relationships can be established. However, commonly used industrial methods of microcapsule production result in polydisperse populations, limiting the information that can be gained concerning the effect of size on their properties. A sustainable method of generating monodisperse, size-controlled polymer microcapsules is therefore highly desirable for research and development into the next generation of environmentally benign microcapsules.The most common method of polymer microcapsule production is interfacial polymerisation (IFP) at the interface of an oil-in-water (o/w) or water-in-oil (w/o) emulsion produced by high-shear mixing with a homogenizer 24 . First, a stable emulsion must be produced with the required droplet diameter; the droplet will form the core of the microcapsule. Polymerization occurs only at the boundary of the emulsion, ensuring that a thin film is formed around the droplet template. The size regime of the droplets produced is chiefly dependent on the emulsification device, surfactants present, and the energy applied to the system 26 ; standard batch high-shear mixing methodologies result in poor control over the size distribution of the droplets (Fig. 1a), and hence the polymer microcapsules formed.Microfluidic methods 27 -that is, where reagents are flowed through micrometre-sized channels and mixed at a junction -are, by contrast, capable of extremely precise control over both droplet size and dispersity [28][29][30] . At the point of mixing, a high shear force is generated between the two immiscible fluids, resulting in droplet formation 29,31 . The shear force can be adjusted by altering the relative flow rates of the two input streams, which, along with channel size, controls the size of the resultant droplet. Thus, it is possible to continuously produce monodisperse emulsions of a desired size. Examples of such droplet production methods have been used in the synthesis of multifunctional magnetoresponsive microcapsules 32 , core-shell organosilicon capsules 33 , and biopolymer hydrogels 34 . Monodisperse polymer microcapsules of between 10-50 microns in diameter are of particular interest to the personal care industry 35 to avoid the potential hazards of nanoparticles; microfluidic methods can readily access this size regime. Furthermore, microfluidic methods require very small amounts of material, and are sustainable due to the short reaction times, low energy usage, waste minimization, and energy and cost efficiency achievable with a continuous process 36,37 . The choice of polymer, oil, and surfactant have a significant effect on the kinetics of polymerisation and the resultant microcapsule shell thickness 24, . Here, we also consider sustainability of the raw materials and the resultant polymer capsules. The ideal polymer shell from an industrial perspective is stable over the shelf-life of the product, capable of payload delivery at the required moment or rate, and not hazardous to the environment. Polyurea microcapsules (PUMCs) have been suggested as candidate materials that fulfil these criteria 38,41 . In terms of feedstocks, the oil used should ideally be from a cheap, renewable source, and function as an active ingredient in a product formulation. Limonene, commonly used in fragrance and foods, can be used as both a template and a payload, and is a sustainable by-product of the citrus industry found in peel 42 . Although polydisperse limonene microcapsules have been previously prepared , there are no studies using microfluidic chips to generate limonene-containing microcapsules. In this research, we use a microfluidic chip to generate monodisperse emulsion microdroplets of limonene containing diisocyanate monomer in an aqueous carrier fluid containing sodium dodecyl sulfate (SDS) and NaCl. We compare the size and polydispersity of the resultant droplets to samples produced by homogenization methods. By systematically varying the flow rate of the oil, we generate emulsions of tunable droplet diameter. Droplet formation is monitored in situ; the droplet size and polydispersity is measured from both still images and videos, the latter using an automated method. Methods of video processing of droplets using Labview have been previously reported 46,47 ; here we use Python code for ease of accessibility. Interfacial polymerisation is achieved offline by collecting the droplets in a stirred solution of aqueous polyamine, which reacts with the diisocyanate to form a polyurea shell (see SI for reaction scheme). The resultant microcapsules are characterised by optical microscopy, SEM, and fluorescence microscopy, and found to have narrow size dispersity, high stability in air over at least 24 h, and the ability to carry a fluorescent payload. Having developed a sustainable method to produce size-controlled microcapsules on demand, we now seek to exploit this to understand the effect of size and dispersity on the performance of microcapsules in product formulations. ## Methods Batch synthesis. An aqueous solution of SDS and NaCl (1.0 wt. % and 1.5 wt. % respectively in 200 ml) and a solution of methylene diisocyanate (MDI) in limonene (0.3 wt. %, 10 ml) were prepared, mixed and homogenised at 8000 RPM for 2 minutes using an ULTRA-TURRAX T-25 homogeniser. The use of SDS and NaCl in these quantities resulted in the formation of emulsion droplets that were stable for at least 24 h. To form capsules, the resulting emulsion was allowed to stand for 10 minutes before a portion (1 mL) was injected into an aqueous solution of tetraethylenepentamine (TEPA), SDS, and NaCl (3.0 wt. %, 1.0 wt. %, and 1.5 wt. % respectively in 10 mL) and stirred at 100 RPM with a magnetic flea for 15 minutes. Capsules were left unstirred for 24 hours before being isolated via pipette and dried in air on a glass slide for imaging and SEM analysis. pressed air pumps was used to generate flow rates of between 1-100 µL min −1 . MDI dispersed in limonene (0.3 wt. %) and an aqueous solution of SDS and NaCl (1.0 and 1.5 wt. % respectively) were delivered to a Dolomite glass 2-reagent droplet chip with a junction size of 50 µm to generate monodisperse emulsion droplets (see SI for full details). The flow rate of the dispersed oil phase, Q d , was varied; the flow rate of the continuous water phase, Q c , was kept constant at 100 µL min −1 . To avoid the potential for blockages, polymerisation was accomplished offline. Droplets were collected in a stirred (100 rpm, magnetic flea) solution of TEPA, SDS, and NaCl in water (3.0, 1.0, and 1.5 wt. % respectively in 10 mL). Droplets were collected for 15 minutes, after which time stirring was stopped and the solution left undisturbed for 24 h prior to being collected via pipette and dried in air on a glass slide for analysis. characterisation of droplets and microcapsules. Droplets were imaged at the junction with a high speed optical microscope capable of capturing both still images and videos. Samples of emulsion and microcapsules were collected before and after polymerisation and imaged using offline optical microscopy. Typically, droplets and microcapsules in the 10-100 µm range are characterised by image analysis, either manually or using image processing software 48 . This analysis is normally limited to 50-100 particles per sample. Laser scattering methods can be unreliable in this size regime, particularly for core-shell particles, as several assumptions about density, refractive index, particle shape, and stability under measurement conditions must be made that do not generally hold for such materials 49 . We therefore decided to explore video processing as an alternative method that would allow the analysis of many more droplets per sample, taking advantage of the continuous production and inline video monitoring of droplets. Still images of droplets and microcapsule were analysed using ImageJ. Videos of droplet production were processed using custom Python code; methodology and limitations are discussed in the SI. Version 1.0 of this code is available under https://github.com/fsimkovic/droplet-assessment. ## Results and Discussion Homogenized emulsions were found to contain droplets from 1-16 µm, with a broad, bimodal distribution of droplet sizes (Fig. 1a,c), as typical for droplets produced by this method 43 . Larger or smaller average droplet sizes can be generated by changing the stirring speed, but high polydispersity always results owing to the variable shear forces experienced in batch processes. By contrast, droplets produced in the microfluidic chip were characterised by narrow dispersity (Fig. 1b,d). Relative flow rates that produced single streams of droplets were targeted to avoid the production of aggregated particles in the polymerisation step; however, smaller or larger droplet sizes could readily be produced by widening the range of flow rates used (Fig. 2a-c). Average droplet diameters between 20-26 µm were measured via video analysis when Q d was varied from 1-5 µL min −1 (Fig. 2d,e); an increase in droplet diameter was observed with increased Q d . www.nature.com/scientificreports www.nature.com/scientificreports/ Narrow dispersity was also observed in the microcapsules produced via subsequent polymerisation of the droplets (Fig. 3a-c). Again, we observed an increase in PUMC size with an increase of Q d, allowing rapid access to 'learning sets' of size-controlled microcapsules. Average PUMC diameters of 27, 30, 32, 34, and 35 µm produced at Q c = 100 µL min −1 and Q d of 1, 2, 3, 4, and 5 µL min −1 respectively were measured by still image analysis. Although it is tempting to compare the sizes of the droplets (20-26 µm, Fig. 2e) and the resultant PUMCs (27-35 µm Fig. 3b), these sets of results are not directly comparable due to the different image processing techniques, and conditions under which the images were obtained (through a glass chip vs. on a glass slide -see SI for detailed discussion). The clear advantage of automated video processing is in enabling the easy processing of tens of thousands of droplets; in this iteration, we sacrifice some accuracy to enable rapid processing (see SI for detailed discussion of the origin of this inaccuracy). In future, a more sophisticated approach, such as a supervised Machine Learning algorithm, could be trained to detect droplets; we anticipate this approach would greatly improve accuracy and enable high quality, automated measurement of droplet sizes as they are generated 50 . Microcapsules were characterised by SEM (Fig. 4a,b); both intact and burst particles were observed after exposure to the high vacuum conditions required for SEM imaging, confirming their hollow nature. The microcapsules were observed to have poor stability under the electron beam, eroding during extended exposure and therefore making it difficult to accurately assess shell thickness. From the images obtained, we estimate a shell thickness of ~100 nm. To visualise the liquid core of the microcapsules, an emulsion containing fluorescent dye (Hostasol yellow 3 G) was generated in the microfluidic chip and subjected to encapsulation via IFP using the protocols described above. The resultant microcapsules were dried on a glass slide for 24 h before imaging with confocal fluorescence microscopy (Fig. 4c). To release the fluorescent payload, a gentle pressure was then applied via a second glass slide (Fig. 4d), indicating that these microcapsules may have utility in applications where pressure-sensitive release is desirable -for example, fragrance release in deodorants. ## conclusions A series of monodisperse polymer microcapsules was produced using microfluidic methods and using sustainable materials. By using video processing to analyse the size distributions of the droplets produced, we can rapidly and automatically establish narrow dispersity, and measure changing droplet size when using different flow rates. Such straightforward and adaptable methodologies are readily extendable to other chemistries, different particle sizes, and new payloads for diverse applications. It has been previously demonstrated that shell thickness and

chemsum

{"title": "Production of monodisperse polyurea microcapsules using microfluidics", "journal": "Scientific Reports - Nature"}

photoinduced_hydrocarboxylation_via_thiol-catalyzed_delivery_of_formate_across_alkenes

## Abstract: Herein we disclose a new photochemical process to prepare carboxylic acids from formate salts and alkenes. This redox-neutral hydrocarboxylation proceeds in high yields across diverse functionalized alkene substrates with excellent regioselectivity. This operationally simple procedure can be readily scaled with low photocatalyst loading (0.01% photocatalyst) without the need for a flow reactor or any precautions to exclude air or moisture. Furthermore, this new reaction can leverage commercially available formate carbon isotologues to enable the direct synthesis of isotopically labeled carboxylic acids. Mechanistic studies support the working model involving a thiol-catalyzed radical chain process wherein the atoms from formate are delivered across the alkene substrate via CO2 •-as a key reactive intermediate. The selective and efficient transformation of alkenes into polar functional groups is a fundamental synthetic strategy. Tremendous progress has been made in alkene functionalization; however, many seemingly simple transformations remain challenging to accomplish. Our group has a growing interest in advancing new alkene functionalization strategies designed to leverage appealing chemical building blocks rather than high energy reagents. 12 Following this line of inquiry, we questioned whether synthetically valuable carboxylic acids could be prepared by delivering formate salts across alkenes (Figure 1, top). In principle, this thermodynamically favorable transformation could occur with perfect atom 22 and redox 23 economy. However, while formate is a common reductant in transition metal catalysis, it is rarely used as a C1 source despite the practical appeal of such an approach relative to gaseous alternatives. Notably, Shi and coworkers leveraged in situ generated formate ahydrides as CO-surrogates for hydrocarboxylation using palladium-catalysis at elevated temperature. We envisioned that a mild approach to directly add formate salts across alkenes would constitute an attractive alternative to this strategy as well as established CO-based hydrocarboxylation processes 33 and emerging methods that rely on reductive activation of CO2. Our reaction design was guided by the recognition that formate is formally comprised of CO2 •-and a hydrogen atom. A strategy to elicit this reactivity from formate could tap into recently developed hydrocarboxylation manifolds that proceed via SET reduction of CO2 (Figure 1, middle). Unfortunately, reductive approaches to access CO2 •-require deeply reducing conditions due to the thermodynamic stability of CO2 (E1/2(CO2/CO2 •-) = -2.2 V vs SCE). 51 To access this requisite driving force for CO2 reduction, prior efforts required <300 nm UV light, 46,47 visible light with stoichiometric thiolate-promoters, 48 or deeply reducing electrodes. 49,50 Furthermore, the photochemical approaches require external hydrogen atom sources while high electrode overpotentials for CO2 reduction 51,52 erode functional group tolerance in electrochemical strategies. In contrast to CO2-reduction strategies, formate enters the reaction in the appropriate oxidation state for alkene hydrocarboxylation without need of a sacrificial electron donor. We suspected that redox-neutral hydrocarboxylation by delivery of formate across alkenes would not only improve atom economy relative to net-reductive strategies, ## + e -+ H• but would also provide an appealing chemoselectivity profile by circumventing the need for strong reductants. 53 Our group 54 and others 55,56 recently introduced a collection of photoredox strategies to generate CO2 •-in situ via cleavage of the formate C(sp 2 )-H bond. In these prior studies, the nascent CO2 •-was primarily employed as an SET reductant. 66 We hypothesized that our catalytic system could be repurposed as a general and functional group tolerant strategy to access CO2 •-for C-C bond-forming reactions. Overall, this would introduce a mechanistically distinct approach to promote hydrocarboxylation reactions via a mild oxidation event (Eox(CHO2 -) = +1.25 V vs SCE) or hydrogen atom abstraction (BDE = 86 kcal/mol) 20 instead of the difficult SET reduction of CO2. Herein, we report a redox-neutral approach to hydrocarboxylation via addition of formate across alkene substrates (Figure 1, bottom). We selected styrene as a model alkene substrate as diverse analogs are commercially available and 3-aryl propionic acids are well-represented in bioactive molecules. 14,15 Of note, the anticipated linear selectivity will complement CO2-based transition-metal-catalyzed processes that furnish branched products from alkenylarenes 35,37,38,40,42,44,45 with one notable recent exception from König and co-workers. 40 Accordingly, we evaluated our previously developed conditions for CO2 •-generation from formate 54 in the presence of styrene. These conditions fully converted styrene in 20 hours and provided a 25% yield of the linear carboxylate, 3, along with 37% ethylbenzene. Reaction optimization-including adjusting the irradiation wavelengths away from those that excite reduced 4DPAIPN 67 -resulted in improved conditions that furnish nearly quantitative yield of 3 without observable ethylbenzene. Furthermore, these reaction conditions provided complete conversion in two hours with low loadings of photoredox and thiol catalysts (Table 1, entry 1). Control experiments confirmed that no conversion of styrene was observed in the absence of the photoredox catalyst (entry 2). 68 The structure of the thiol hydrogen atom transfer (HAT) catalyst was identified as a key parameter. Omission of the thiol from the reaction resulted in diminished rate and, consequently, reduced chemical yield (entry 3). The alkyl thiol we employed in related, formate-based hydroarylation processes, 54 CySH, was similarly ineffective (entry 4). While several thiolphenols and electron-deficient thiols performed comparably to T1 (see Table S1), when T1 was substituted for an electron-rich analog, T2, the yield was substantially diminished, reverting to nearly that of the reaction performed without thiol (entry 5). Overall, these data cannot be rationalized using thermodynamic parameters such as BDEs 69,70 but are fully consistent with substantial polar effects on the HAT transition structures. 71 The formate counterion also had an impact on the reaction. Substitution of potassium for sodium slows the reaction and results in lower yield (entry 6). Replacement of potassium with cesium delivers a similar yield and modestly accelerates the rate (entry 7, see Table S6 for details regarding rate changes). We attribute this effect to differential solubility of the formate salts in DMSO. Potassium formate was selected for further study as it furnishes nearly quantitative product in only two hours and is inexpensive. 72 The photoredox catalyst identified (4DPAIPN) was particularly effective; however, a variety of other photocatalysts promote the reaction. For example, iridium-based photocatalysts could be used in place of 4DPAIPN, albeit with extended reaction times (entry 8). 73 We found that the generation of these carboxylic acid products is robust; no precautions to exclude air or moisture are necessary and the process tolerates the deliberate addition of water (entry 9). We next examined the scope of this new alkene hydrocarboxylation reaction. We found that these simple conditions promote the delivery of formate across a wide range of alkenylarene substrates with exquisite functional group tolerance (Table 2). Diverse electron-donating and electron-withdrawing substituents could be introduced on the arene (3-8) without a substantial impact on reaction efficiency. Since the reaction conditions are only mildly basic, protic substrates were well-tolerated. Alkenylarenes bearing carbamates (5), carboxylic acids (7 and 14), 74 and unprotected alcohols (9 and 13) each underwent hydrocarboxylation in high yield. Furthermore, a substrate containing a synthetically versatile but Lewis acidic boronic acid pinacol ester (8) was efficiently converted into the linear carboxylic acid. This redox-neutral process also tolerates reductively sensitive functional groups, such as aryl chlorides (6). This substrate was of particular interest because we have previously engaged aryl chlorides in reductive radical coupling reactions under similar formate-based reaction conditions. 75 Hydrocarboxylation of 2-vinylpyridne (10) illustrated the suitability of vinyl heterocycles as substrates. The reaction was insensitive to additional alkene substituents; aand b-methyl styrenes were each converted to carboxylic acids (11 and 12). Additionally, functionalized alkyl groups such as unprotected alcohols or acids in the b- ## H position did not disrupt the hydrocarboxylation process (13 and 14). We found that the reaction was not limited to alkenylarene substrates; electron-deficient alkenes successfully underwent hydrocarboxylation to furnish succinate derivatives (Table 3). Hydrocarboxylation of these substrates was unperturbed by substitution at either the aor b-position (15-17), providing access to differentially substituted 1,4-dicarbonyl compounds. Hindered b,b-disubstituted substrates converted at diminished rate under the standard conditions; however, substitution of potassium formate for an excess of more soluble cesium formate accelerated the rate and delivered product (18) in high yield in two hours. Further investigation of this sterically-congested substrate class illustrated that this new reaction provides efficient access to heterocyclic building blocks bearing aquaternary carboxylic acids (19-21). Since aliphatic alkenes did not undergo hydrocarboxylation, a substrate bearing both an unactivated and activated alkene (22) underwent chemoselective hydrocarboxylation at the electron-deficient alkene to provide 23. Given that scaling photochemical reactions can be technically challenging, 76 we evaluated the viability of performing this process on preparative scale. Using a simple batch setup with no precautions to exclude air or moisture, the carboxylate salt 24 was synthesized in 79% yield (7.4 g, 39.5 mmol) in under nine hours using only 0.01 mol% of the photocatalyst, 4DPAIPN (Scheme 1). This preparative scale reaction is not only technically simple to execute but the carboxylate salt can be purified from the reaction mixture by crystallization with neither chromatography nor extensive aqueous washes to remove non-volatile DMSO. Overall, these results illustrate the immediate practical utility of this new formate-based hydrocarboxylation reaction. Based on our working mechanistic model, we anticipate that the formate salt is incorporated as the carboxylic acid in the final product. This opens up an appealing avenue to prepare isotopically labeled molecules because both 13 Cand 14 C-labeled sodium formate salts are commercially available. We first adjusted the reaction conditions to employ limiting sodium formate as this would be particularly attractive for 14 C-radiolabeling applications. 77 Under these modified conditions, we found three distinct bioactive molecules (25-27) could be produced with near perfect 13 C-incorporation (Table 4). This offers a simple but effective complement to recently developed carboxylic acid isotopic exchange reactions. H simultaneously deliver high specific activity and radiochemical yield whereas this hydrocarboxylation process offers both. Finally, each of these bioactive carboxylic acids can also be synthesized efficiently under our standard limiting-alkene conditions in high yields (see Figure S9). We next questioned whether the hydrogen atom incorporated in the final product is derived from formate. To this end, we subjected D-formate to our reaction conditions and observed high but incomplete deuteration (Table 5, entry 1). This is consistent with formate acting both as a C1 and hydrogen atom source but also suggests a secondary process to account for the incomplete deuterium incorporation. We questioned whether a parallel electron-transfer-protontransfer (ETPT) pathway may also occur. In this process, the benzylic radical would be reduced (Ered = -1.4 V vs SCE) 87 and subsequently protonated, by solvent or adventitious water. 88 However, running the reaction in DMSO-d6 resulted in no measurable deuterium incorporation (entry 2). Given that both DMSO and water would spontaneously quench a benzylic anion and DMSO is present in vast excess, these data indicate that ETPT is not responsible for incomplete transfer of the deuterium label from formate to the carboxylic acid product. Given our working model that the thiol catalyzes the reaction by shuttling hydrogen atoms, we suspected that proton exchange between the thiol co-catalyst and adventitious water could explain these data. Addition of D2O to otherwise standard conditions resulted in substantial deuterium incorporation (entry 3). Conversely, when a D-formate reaction was conducted in the presence of H2O, deuterium incorporation was not observed (entry 4). Taken together, these data are consistent with our working model wherein the thiol catalyzes HAT from formate to transient C(sp 3 )-radical intermediates. 89 Based on these results, we propose a working mechanistic model based on a thiol-catalyzed radical chain process (scheme 2). 90 Thiyl radicals generated in situ could abstract a hydrogen atom from formate to generate CO2 •-. This radical anion intermediate, in turn, reacts with the alkene substrate to furnish a new C-C s-bond and a C(sp 3 ) radical. This radical intermediate is quenched by HAT from the thiol, regenerating the thiyl radical. We envisioned two plausible mechanisms for initiation of this chain process. First, SET oxidation of formate in DMSO is known to result in a second order decomposition to formic acid and CO2 •-(initiation A). 91 Second, SET oxidation of the thiol followed by a proton transfer could directly generate the key thiyl radical intermediate (initiation B). Stern-Volmer analysis indicated that both a soluble formate salt (tetrabutylammonium formate) and the thiol catalyst, T1, quenched the excited state of the photoredox catalyst 4DPAIPN. However, the rate of photocatalyst quenching by T1 is approximately an order of magnitude faster than formate and potassium formate is sparingly soluble in DMSO. 92 Accordingly, we favor SET oxidation of the thiol catalyst (initiation B) as the primary initiation mechanism but suspect both occur in parallel under standard conditions. Overall, we have introduced a new strategy to access linear carboxylic acids from formate and activated alkenes by exploiting a photoinitiated, thiol-catalyzed chain reaction. This redox-neutral process is conducted under mild conditions without stoichiometric redox agents and, as a result, a wide variety of functional groups are tolerated by the procedure. The conditions are operationally simple; no precautions to exclude air or moisture are necessary. This transformation is also readily translated to preparative scale (50 mmol) using a straightforward batch reaction set up. This hydrocarboxylation method also provides a facile approach to leverage commercially available isotopically labeled formate salts to prepare labeled bioactive carboxylate ## isotopically-labeled bioactive molceules a Reactions were conducted under air and run for 8 h with 0.4 mmol Na 13 CHO 2 and 0.6 mmol styrene and yields were determined relative to Na 13 CHO 2 via 1 H NMR and 13 C incorporation was determined by mass spectrometry. See the SI for further details. products. Additionally, mechanistic investigations revealed that formate acts as both the C1 and hydrogen atom source. This study illustrates the potential benefits of a redox-neutral approach to alkene hydrocarboxylation and, more broadly, provides a roadmap to unlock formate as a general CO2

chemsum

{"title": "Photoinduced Hydrocarboxylation via Thiol-Catalyzed Delivery of Formate Across Alkenes", "journal": "ChemRxiv"}

annulative_π-extension_of_indoles_and_pyrroles_with_diiodobiaryls_by_pd_catalysis:_rapid_synthesis_o

## Abstract: A palladium-catalyzed one-step annulative p-extension (APEX) reaction of indoles and pyrroles that allows rapid access to nitrogen-containing polycyclic aromatic compounds is described. In the presence of palladium pivalate and silver carbonate, diverse indoles or pyrroles coupled with diiodobiaryls in a double direct C-H arylation manner to be transformed into the corresponding p-extended compounds in a single step. The newly developed catalytic system enables the use of various pyrroles and indoles as templates with a series of diiodobiaryls to provide structurally complicated and largely p-extended nitrogen-containing polycyclic aromatic compounds that are otherwise difficult to synthesize. ## Introduction With desirable electronic properties and diverse biological activities, nitrogen-containing fused aromatics have long been recognized as privileged structures in the felds of organic materials and pharmaceutical science. 1 As these properties can be readily tuned via skeletal modifcation of the core N-heteroarene structure, signifcant efforts have been devoted to develop new synthetic approaches for p-extended nitrogen-containing polycyclic aromatic compounds (N-PACs). Representative classical approaches include (i) intramolecular carbon-nitrogen bond formation of biaryl amines, 2 (ii) intramolecular carboncarbon bond formation of diaryl amines, 3 and (iii) stepwise functionalization and p-extension of indoles and pyrroles. 4 However, these methods require the use of prefunctionalized heteroaromatics such as halogenated pyrroles, anilines and indoles, and stepwise transformations from unfuctionalized (hetero)aromatics. To achieve maximum efficiency in N-PAC construction, a more direct and 'intuitive' method for p-extension of unfunctionalized pyrroles and indoles is called for. Recently, we have introduced several new one-step methods for the annulative p-extension (APEX) of unfuctionalized (hetero)aromatics (Fig. 1a). Because such APEX reactions directly transform easily available unfunctionalized (hetero) arenes to polycyclic aromatic hydrocarbons, nanographenes and p-extended heteroaromatics in a double direct C-H arylation manner, these protocols offer large benefts in the context of cost, simplicity, and step/atom economy. 8 Recently, we 7 and others 9-14 have reported transition-metalcatalyzed APEX reactions of indoles and pyrroles using various p-extension units such as alkyne, 9 alkene, 7a,10 1-vinylpropargyl alcohols, 11 a-diazocarbonyl compounds, 12 a-bromochalcone, 13 a-bromocinnamate, 13 cyclic diaryliodonium salts, 14 dibenzogermoles 7b and diiodobiphenyls 7c (Fig. 1b). However, these APEX reactions are limited in terms of lack of variety in pextending agents, narrow substrate scope, and low functional group tolerance. Herein, we report a new catalytic APEX reaction that allows efficient pyrrole-to-indole, pyrrole-to-carbazole and indole-to-carbazole p-extensions. Our newly established catalytic system featuring palladium pivalate and silver carbonate in a mixed DMF/DMSO solvent system enabled the rapid synthesis of structurally complicated N-PACs from readily available unfunctionalized pyrroles/indoles and diiodobiaryls. ## Results and discussion We began our study by optimizing the reaction conditions for indole-to-carbazole extension of N-methylindole (1a) using 2,2 0diido-1,1 0 -biphenyl (2a) as a p-extending agent (Table 1). After extensive screening, we discovered that 1a (1.0 equiv.) coupled with 2a (1.5 equiv.) in the presence of Pd(OAc) 2 (5 mol%) and Ag 2 CO 3 (3.0 equiv.) at 80 C in 7 : 3 mixture of dimethylformamide (DMF) and dimethylsulfoxide (DMSO) to provide N-methyldibenzo[a,c]carbazole (3aa) in 54% yield (entry 1). Use of palladium pivalate [Pd(OPiv) 2 ] instead of Pd(OAc) 2 improved the yield to 61% (entry 2), but other palladium sources such as PdCl 2 , PdI 2 , Pd(PPh 3 ) 4 , Pd(OCOCF 3 ) 2 and Pd(CH 3 -CN) 4 (BF 4 ) 2 failed to give more than trace amounts of product (entries 3-7). Decreasing the amount of Ag 2 CO 3 to 1.5 equiv. (relative to 1a) further increased the yield of 3aa to 78% (entry 8). The use of silver carboxylate salts (AgOAc, AgOPiv, or AgOCOCF 3 ) instead of Ag 2 CO 3 resulted in much lower yield (entries 9-11). The silver cation itself was essential for this reaction; the use of Na 2 CO 3 , K 2 CO 3 or Cs 2 CO 3 instead of Ag 2 CO 3 failed to give any product (see ESI † for details). The use of the DMF/DMSO mixed solvent system was important for obtaining maximum conversion; highly polar single solvents such as N,Ndimethylacetamide (DMAc), DMF, DMSO, CH 3 CN provided 3aa in diminished yield (29-10%, see ESI † for details), while less polar solvents such as 1,2-dichloroethane, 2,2,2-tri-fluoroethanol, 1,4-dioxane and toluene completely suppressed the reaction. Although higher reaction temperature accelerated the consumption of the starting material, the yield of 3aa was decreased (entries 12 and 13). Finally, we confrmed that the APEX reaction did not proceed in the absence of Pd catalyst or Ag 2 CO 3 (entries 14 and 15). Although the use of additional ligands for Pd and the use of dibromobiphenyl in place of diiodobiphenyl as the p-extension reagent were also investigated, these modifcations proved ineffective (see ESI † for details). Ultimately the conditions in entry 8 were deemed optimal for the present indole-to-carbazole APEX reaction. A possible reaction mechanism of current indole-tocarbazole APEX reaction is shown in Scheme 1. Oxidative addition of 2a to palladium species (Pd(0) or Pd(II)) occurs to form biphenylylpalladium intermediate A. 15 Then, the removal of iodide by silver salt may activate Pd complex A 16 to form electron-defcient aryl-Pd species, 17 which then react with indole at the C2 position to afford intermediate B. Through the control experiments on the C-H arylations of 1,2-dimethylindole and 1,3-dimethylindole with iodobenzene, the present APEX reaction seems to occur through the C2-arylation of indole rather than C3-arylation in the frst step (see ESI † for details). Final step would be well-established Pd-catalyzed Scheme 1 Proposed reaction mechanism for the Pd-catalyzed APEX reaction of N-methylindole (1a) with 2,2 0 -diiodo-1,1 0 -biphenyl (2a). intramolecular C-H/C-I coupling to afford the cyclized compound 3aa. 18 Under the optimized conditions, various types of p-extended carbazoles/indoles 3, 5 were prepared from the corresponding indole/pyrrole derivatives 1, 4 and diiodobiaryls 2. Scheme 2 illustrates the scope of applicable indole and pyrrole derivatives (1a-1m). N-Alkyl (2a, 2b), N-benzyl (2c), N-phenyl (2d) indoles and cross-linked lilolidine (2e) were converted smoothly to dibenzocarbazoles 3ba-3da in good to moderate yield, however the reaction of N-acetyl indole 2f did not provide the expected pextension product 3fa. The presence of substituents at the 5-, 6-, or 7-positions of the indole ring was well-tolerated, giving various nitro-(3ga), cyano-(3ha, 3la), bromo-(3ia), methoxy-(3ja, 3ka), and benzyloxy-substituted (3ma) dibenzocarbazoles in moderate yields (40-62%). These results suggest that substituents on the benzene ring of indole do not critically affect the reaction progress. Interestingly, we found that the current APEX reaction between N-substituted pyrroles and 2a was mono-selective for the formation of dibenzoindoles 5aa and 5ba in 39% and 43% yields; only trace amounts of di-APEX tetrabenzocarbazole products, the main products of our previous report, 7c,19 were observed. As synthetic methods to prepare the dibenzo [e,g]indole skeleton remain limited and inefficient, 20 the current APEX protocol provides a valuable, streamlined entry to this compound class. The scope of diiodobiaryls in the current APEX reaction is shown in Scheme 3. The reaction of N-methylindole (2a) with unsymmetrical 4-chloro-2,2 0 -diiodo-1,1 0 -biphenyl (2b) gave a 1 : 1.4 regioisomeric mixture of 3ab and 3ab 0 in 94% combined yield. APEX reactions of 1a and 4a with 4,4 0 -dibromo-2,2 0 -diiodo-1,1 0 -biphenyl (2c) smoothly occurred to give dibromodibenzocarbazole 3ac and dibromodibenozoindole 5cc in 81% and 44% yield, respectively. To our delight, the reaction of 1a with 2,2 0 -diiodo-1,1 0 -binaphthalene (2d) gave dinaphthocarbazole 3ad containing a helicene moiety in 25% yield, whose helical structure was confrmed by X-ray crystallographic Scheme 2 Substrate scope of indoles and pyrroles in the APEX reaction with 2,2 0 -diiodo-1,1 0 -biphenyl (2a). Scheme 3 Substrate scope of diiodobiaryls. a 0.20 mmol scale. Scheme 4 Sequential APEX reactions of N-methylpyrrole (4c) for the synthesis of unsymmetrically substituted tetrabenzocarbazole 6. analysis. As this example clearly demonstrates, the late-stage attachment of complex, extended polyaromatic units is one of the most remarkable characteristics in the present APEX reaction. To demonstrate the power of the current APEX reaction to build complex, unsymmetrical N-PACs from simple starting materials, we employed a two-step sequence to synthesize tetrabenzocarbazole 6, a compound difficult to prepare via known methods (Scheme 4). First, APEX reaction of N-methylpyrrole (4c) with 2a was carried out to give the corresponding N-methyldibenzoindole (5ca) in 37% yield. Notably, this reaction did not give double-APEX product which is the major product in the previously developed APEX reaction of N-phenylpyrrole. 8b Then, 5ca was further reacted with 4,4 0 -dibromo-2,2 0 -diiodo-1,1 0 -biphenyl (2c) by using Pd(CH 3 CN) 4 (BF 4 ) 2 / AgOPiv/TfOH catalytic system 8b to give the desired product 6 in 33% yield. 21 Rapid access to a new class of unsymmetrically substituted tetrabenzocarbazole is notable, and should contribute to the exploration of new compounds for organic electronics application. The current APEX reaction also provided a facile route to polycyclic aromatic compounds containing both nitrogen and sulfur (N-S-PACs) (Scheme 5). N-Methylindole (1a) coupled with 3,3 0 -diiodo-2,2 0 -bibenzothiophene (7) to give di(benzothieno)carbazole 8 in 32%. To our delight, the reaction of Nmethylpyrrole (4c) with diiodo-2,2 0 -bibenzothiophene 7 afforded a double APEX product, tetra(benzothieno)carbazole 9, in 15% yield. While the yields were low, the generation of these novel N-S-PAC structures, which are highly interesting from the viewpoint of optoelectronic properties yet otherwise difficult to synthesize by conventional organic reactions, is notable. The structural and electronic properties of 8 and 9 were elucidated via X-ray crystallography, UV-vis/photoluminescence spectroscopy, and DFT/TD-DFT calculations at the B3LYP/6-31G(d) level of theory (Fig. 2). Single crystal X-ray structures (Fig. 2a, b, S2 and S3 †) reveal that compound 8 adopts a relatively flattened structure in the solid state (Fig. 2a), while compound 9 possesses a twisted structure owing to the embedded heterohelicene moiety. DFT calculations for 8 (Fig. 2c) reveal delocalization of the HOMO (5.23 eV) over the entire molecule, while the LUMO (1.49 eV) localizes on a benzothienocarbazole wing. On the other hand, the HOMO and LUMO of 9 are delocalized over entire molecule, and thus the energy level of LUMO (1.72 eV) is slightly lower than that of 8. The UV-vis absorption spectra of 8 and 9 in CH 2 Cl 2 show that both compounds have broad absorption bands between 300 and 450 nm (Fig. 2e). Absorption maxima were found at 294, 317, 339, 357, 381 and 399 nm in 8, and the corresponding peaks were also found in 9 at 305, 332, 348, 393 and 412 nm. The TD-DFT calculations revealed that the longest-wavenumber absorptions in 8 and 9 (399 and 412 nm) are attributed to the allowed HOMO-LUMO transitions (see ESI † for details). The fluorescence spectra of 8 and 9 in CH 2 Cl 2 display broad emissions with emission maxima of 427 and 437 nm, respectively (Fig. 2e). Scheme 5 APEX reactions of N-methylindole (1a) and N-methylpyrrole (4c) with 3,3 0 -diiodo-2,2 0 -bibenzothiophene (7) for the synthesis of N-S-PACs. ## Conclusions In summary, we have developed a novel palladium-catalyzed APEX reaction to enable the annulative p-extension of indoles/pyrroles with diiodobiaryls. Use of the Pd(OPiv) 2 / Ag 2 CO 3 catalytic system in a mixed DMF/DMSO solvent allows the preparation of a diverse range of N-PACs in a single step, including several previously unsynthesized structures. Rapid access to exotic scaffolds such as complex, unsymmetrically substituted tetrabenzocarbazoles and extended N-heteroarenes featuring multiple helicene moieties is a particular highlight of the present APEX protocol. Developed APEX methodology also has great potential for the efficient and rapid synthesis of planar and non-planar p-extended N-PACs such as p-extended azacorannulenes, aza-buckybowls and pyrrolopyrroles which are regarded as one of promising materials for optoelectronics. 22 Further investigations into the reaction mechanism and applications of this APEX method towards the synthesis of larger pextended heteroaromatics are currently underway. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Annulative \u03c0-extension of indoles and pyrroles with diiodobiaryls by Pd catalysis: rapid synthesis of nitrogen-containing polycyclic aromatic compounds", "journal": "Royal Society of Chemistry (RSC)"}

uv-irradiation_of_self-assembled_triphenylamines_affords_persistent_and_regenerable_radicals

## Abstract: UV-irradiation of assembled urea-tethered triphenylamine dimers results in the formation of persistent radicals, whereas radicals generated in solution are reactive and quickly degrade. In the solid-state, high quantities of radicals (approximately 1 in 150 molecules) are formed with a half-life of one week with no significant change in the single crystal X-ray diffraction. Remarkably, after decay, re-irradiation of the solid sample regenerates the radicals to their original concentration. The photophysics upon radical generation are also altered. Both the absorption and emission are significantly quenched without external oxidation likely due to the delocalization of the radicals within the crystals. The factors that influence radical stability and generation are correlated to the rigid supramolecular framework formed by the urea tether of the triphenylamine dimer. Electrochemical evidence demonstrates that these compounds can be oxidized in solution at 1.0 V vs. SCE to generate radical cations, whose EPR spectra were compared with spectra of the solid-state photogenerated radicals. Additionally, these compounds display changes in emission due to solvent effects from fluorescence to phosphorescence.Understanding how solid-state assembly alters the photophysical properties of triphenylamines could lead to further applications of these compounds for magnetic and conductive materials. ## Introduction Intentional design of functional supramolecular assemblies requires precise control of intermolecular interactions as well as an understanding of how complex structures modulate chemical and physical properties to produce materials with emergent qualities. 1 This understanding is key for designing compounds used to probe the enhancement or quenching of luminescence of small molecules in the solid-state. 2 Controlled assembly of structures can also modulate conductivity 3 and dichroism in photoactive materials. 4 Here, we synthesize urea tethered triphenylamines (TPAs), and determine their photophysical properties in solution and in crystalline assemblies. Upon UVirradiation, in both solution and the solid-state, these materials displayed radical formation with solid-state samples proving to be quite stable (Fig. 1). Remarkably, solid-state samples yield high quantities of persistent radicals with $1 in 150 molecules containing a radical. Moreover, after decay, re-irradiation with UV light can regenerate the radicals in similar quantities. Thus, solid-state assembly alters the photophysical properties of TPAs and could prove helpful in the design of Fig. 1 (A) Self-assembly of a triphenylamine derivative affords persistent radicals upon irradiation with UV light. (B) UV-irradiation induces a noticeable change in color. (C) A significant radical signal is observed which corresponds to 1 in $600 molecules displaying a radical after 1 h of UV-irradiation and up to 1 in $150 after 8-11 h. conductive and magnetic materials that integrate TPA components. Para substituted TPAs are prevalent examples of molecules that exhibit persistent radicals. 5 For example, Magic Blue, an antimony salt of tribromo TPA, is a commercial one-electron oxidant employed for many chemical processes. 6 The stability of substituted TPA radicals, has led to their use as promising spin-containing units for organic polymer based magnets. 7 These organic magnets are designable offering moldability and tunability. The oxidation of the TPA also alters its photophysics, leading to quenching of its fluorescence. 8 Typically, TPA compounds require para substitution on all the phenyl rings to generate stable radical cations. 9,10 The extra substitution helps to slow down degradation reactions such as benzidine formation. 9 Usually, chemical or electrochemical oxidation is required to generate the radicals. 8 Even without the oxidation to a radical, TPAs still fnd many uses as twophoton absorbers, 11 organic light emitting diode materials, 12 solvatofluorochromatic intramolecular charge transfer (ICT) molecules, 13 and as aggregation induced emission (AIE) compounds. 14 The Shimizu group utilizes the three-centered urea interaction to drive assembly of linear and macrocyclic monomers into tapes, rods, and columns. 15 In the case of benzophenone containing monomers, assembly influences the photophysics and affords surprisingly stable radicals upon UV-irradiation. 16 For comparison, unassembled structures in solution show no radical formation upon UV-irradiation. Our hypothesis is that supramolecular assembly signifcantly enhances radical stability. Here, we test if urea-tethered triphenylamines will be affected in a similar manner. We synthesized and compared the structures and properties of a methylene urea bridged 4-bromo TPA dimer (3) against 4bromo TPA (1) and a protected urea analog (2). The structure of dimer 3 features one bromine on each TPA adduct to assist in intersystem crossing (ISC) from the excited singlet to the triplet state which can be aided by the heavy atom effect, and should help promote radical generation from UV-irradiation. The heavy atom effect increases ISC due to spin orbit coupling. 17 Additionally, one para position on each TPA unit of 3 was left intentionally unsubstituted. Typically, fully substituted TPA's are required for radical stability. 5 Here, we test if supramolecular assembly can provide stability to unsubstituted TPA radicals, which in turn, would allow for greater variability in TPA structures with stable radical characteristics. Radical formation was investigated by two methods: electrochemical oxidation and UV-irradiation. Both of these methods can generate radical cations in TPA compounds, with the former being well-known, 8 and the latter requiring a reducible agent in the molecule itself 18 or in molecules close-by (i.e. solvent). 19 Our goal is to characterize these systems by electron paramagnetic resonance (EPR) spectroscopy to understand how solid-state organization influences their ability to generate stable radicals versus dissolution. Specifcally, we are testing (1) if self-assembly can stabilize radicals and (2) if UV-irradiation is a useful tool to generate TPA radicals in reasonable quantities. Additionally, we will examine the solvent dependent photophysics of the triplet and singlet emissions of these molecules. ## Results and discussion The urea tethered triphenylamine 3 was synthesized in fve steps from commercial 4-bromotriphenylamine using a Vilsmeier-Haack reaction with phosphoryl chloride to yield the aldehyde, 20 which was subsequently converted to the alcohol via hydride reduction (Scheme 1). 21 After bromination of the alcohol, 22 two TPA units were tethered through triazinanone under basic conditions. Deprotection of the urea afforded 3 as a pale yellow powder. Colorless needles were regularly obtained by slow evaporation of ethyl acetate solutions ($20 mg mL 1 ) and were used for all solid-state measurements. The crystals were also subjected to X-ray diffraction analysis. Triphenylamine 3 crystallized in the orthorhombic system in the Pccn space group. The X-ray structure revealed the desired compound with a linear trans-trans arrangement of the ureas with the two TPA units outstretched on both sides of the methylene urea tether in an anti-parallel manner. Crystallographically, the structure is disordered with two molecular orientations present (Fig. S9 †) with the major component population of 91%. The urea carbonyl, which is located on a crystallographic C 2 axis, is common to both components. The individual molecules are organized into chains extending along the crystallographic c-axis through characteristic three-centered urea hydrogen bonds with a twisting angle of 51.6 (1) . The hydrogen-bonded urea groups (N(H)/O distances of 2.823(3) and 2.70(2) , (Fig. 2A)) generate an X-shaped chain when viewed down the c-axis (Fig. 2B). The twisting is likely caused by the extra steric bulk of the TPA since similar dibenzylic systems typically have straight urea chains according to a Cambridge Structural Database survey (CSD 5.39, September 28, 2018). 23 To examine how solvent and assembly affects the photophysics of the TPA compounds, the absorption and emission for 1, 2, and 3 were taken in six solvents and in the solid-state at room temperature. The studies in dichloromethane, dimethyl sulfoxide, ethyl acetate, ethanol, acetonitrile, and tetrahydrofuran are summarized in Table 1 and S2. † In all the tested solvents, the absorption spectra of 1-3 were nearly identical, with a strong pp* transition at approximately 300 nm Scheme 1 Synthetic scheme for 2 and 3. dominating the spectra with no other bands readily apparent (Fig. 3A). This suggests that in solution the proximity of the two TPA units have little effect on the absorption properties. On average, 2 and 3 were red-shifted by 2 nm compared to 1. As expected, the molar absorptivity for 2 and 3 were very similar and twice that of 1 with values ranging from 4.70-5.53 10 4 M 1 cm 1 and 2.12-2.68 10 4 M 1 cm 1 , respectively. For solid-state samples, crystals of 3 were frst examined by PXRD to probe if the bulk crystalline material was similar in structure to the single crystal of 3. Fig. S11 † compares the experimentally observed PXRD pattern to the predicted powder pattern simulated from the SC-XRD data. Seen here is an excellent correlation, suggesting that the bulk material is single phase and similar in structure to the solved crystal structure. This indicates that the photophysical measurements of the bulk material would be representative of the single crystals. Selfassembly of 3 resulted in a red shift of the absorbance of about 60 nm with slight broadening of the main peak (Fig. 3A). A similar red shift has been reported before for other triphenylamine derivatives on the basis of J-aggregates; 24 however, this is more common for planar dyes. 25 The emission spectra recorded in solution for 2 and 3 exhibited two main transitions either at approximately 370 nm (Band 1) and/or 450 nm (Band 2). As seen in Fig. 3B and C, the intensities of these bands varied widely on the basis of solvent with 2 exhibiting more Band 1 character and 3 more Band 2. Band 1 is generally considered the fluorescence band for TPA systems. 13 To identify Band 2, further experiments were carried out. First, the emission of 1 was taken in different solvents to probe if Band 2 was derived from an ICT process. Although ICT typically requires a donor-p-acceptor system, 26 the TPA units of 2 and 3 could rotate over to each other allowing the TPA units on either side of the urea tether to act as both the donor and acceptor in the ICT exchange without the need for a p-system intermediate. For this case, the TPAs would have to adopt acceptor characteristics since TPAs are typically only the donor in ICT systems. 27 As seen in Fig. S15, † 1, which is only a single TPA unit, also exhibited Band 2. This suggests that band 2 is not due to an ICT process. Second, DOSY NMR studies were carried out on 3 to probe if Band 2 originated from an AIE process. Since AIE has been known to create new emissive bands, 28 ureas are known aggregators, 29 and AIE has been observed in TPA systems before 14 it seems reasonable that Band 2 could be derived from this process. For DOSY NMR, aggregation is detected when the observed hydrodynamic radius is signifcantly higher than the radius of the monomer. DOSY studies were conducted on solutions of 3 in deuterated acetonitrile (1 mM and 100 mM). This solution was chosen since it displayed the most signifcant Band 2 character of all the trials. As seen in Fig. S18 and S19, † no aggregation was observed for 3 in acetonitrile since the observed hydrodynamic radius of approximately 8 for both solutions is only slightly higher than that calculated from the crystal structure monomer of 3 ($6 ). The slightly higher radius may be from solvation or a slight amount of dimerization, but defnitively no large-scale aggregation was observed. Considering that more dilute solutions (10 mM) were used for photophysical measurements and no aggregation was observed in more concentrated samples, this suggests that AIE is not responsible for Band 2. With ICT and AIE ruled out for the occurrence of Band 2, phosphorescence is suggested as the likely origin. Phosphorescence is common for structures containing TPAs that are employed in OLED materials. 30 Additionally, the peak position in the emission spectra is in good agreement for where phosphorescence is typically observed in TPA compounds. The bromine substituent can increase spin-orbital coupling via the heavy atom effect, which gives access to the triplet state thus enhancing phosphorescence. 17 This effect can occur with either the heavy atom being directly connected into in the p-system, 31 or in close proximity. 32 The former case could explain why all three compounds exhibit Band 2, while the latter case could explain the intensity of this band (3 > 2 > 1). 30 The emission spectra were also measured in the presence of a triplet quencher (triethylamine) and in an oxygen-saturated solution of dichloromethane. The emission was reduced in both cases (Fig. S31 and S32 †), further suggesting that this band arises from phosphorescence. For 1-3, increasing solvent polarity resulted in increased phosphorescence, except in the case of polar protic solvents (ethanol) which showed little to no phosphorescence (Fig. 3B and C). Solvent dependent phosphorescence has been observed before when S 1 and T n were similar in energy. 33 In this situation, different solvents stabilized either state in varying degrees resulting in different phosphorescent quantum yields for each solvent, which may be the case here as well. To further investigate the solvent dependence, we examined if phosphoresce could be 'turned on' by the addition of a polar solvent to a nonpolar system (Fig. 3D). Starting with a fluorescent non-polar system (2 in EtOAc), the addition of acetonitrile slowly turned on phosphorescence until the system was mostly phosphorescence clearly showing the solvent dependent nature of these emissive bands. The luminescent lifetimes for 2 and 3 were found to be quite short for TPA derivatives with fluorescence and phosphorescent lifetimes estimated to be around 0.1 ns and 3 ns, respectively (Table 1, Fig. S20 and S21 †). This may be due to competing nonradiative decay processes introduced by the heavy atoms. Typically, fluorescent lifetimes for TPA containing compounds tend to hover around 2 ns, 34 but heavy atom containing TPA derivatives with small p-systems have be seen to exhibit shorter lifetimes (<0.1 ns). 22 For the phosphorescent lifetimes, while the heavy atom effect does increase phosphorescence intensity 17 it can shorten the lifetimes as well. 35 In the solid-state it was not indicatively clear if fluorescence or phosphorescence was occurring as both the Stoke's shift (89 nm) and lifetime (1.0 ns) were in-between the expected values for fluorescence (65 nm, 0.1 ns) and phosphorescence (150 nm, 2.0 ns) determined from solution studies. Additionally, attempts at measuring accurate quantum yields for these compounds were unsuccessful due to radical generation and its subsequent effects on the photophysical properties. The short observed lifetimes could be explained by the formation of stable radicals or by other non-radiative pathways. Thus, we turned to X-band EPR spectroscopy to probe radical formation within solution samples. First, a solution of 3 ($1 mM) was prepared in degassed DCM and was sealed under argon. While no EPR signal was observed pre UV, UV-irradiation (1 h) of the sample resulted in an EPR signal with a g-value of 2.005 (Fig. S22 †). However, this radical was unstable and found to rapidly undergo degradation reactions. Fig. S26 † compares the 1 H NMR spectra of 3 in solution before and after UVirradiation. The post UV sample shows nearly a complete loss of all of the parent resonances. Post-UV absorption and emission spectra also followed this trend (Fig. S14 and S17 †) clearly indicating that radicals of 3 generated in solution are not stable. This is not unexpected since electrochemical studies of control 1 indicate an unstable radical cation in solution 36 and radical cations generated from TPAs with unsubstituted para positions are known to be unstable in solution. 9,10 Next, EPR spectra were recorded on crystals of 3 in order to investigate how solid-state assembly influences the formation of radicals. First, EPR spectroscopy was performed on a triply recrystallized sample of 3 (3.9 mg) which was UV-irradiated for 6 h. Fig. S23 † shows the recorded EPR spectra, which displays a broad signal with an axial powder pattern shape. The observed g-value is 2.006, which is in the range of TPA radical cations in solution (2.002-2.005). 5 Singly recrystallized samples of 3 (10 mg) were also examined pre and post UV-irradiation (Fig. 1C). As expected, no signal was seen pre irradiation; however, after 1 hour of UV-irradiation a broad EPR signal identical to the triply recrystallized sample was observed. The persistence of the photogenerated radicals was examined using dark decay studies in which the recrystallized sample was irradiated for one hour and then stored in the dark at room temperature under argon. The EPR spectrum was monitored over a month to estimate its stability (Fig. 4A). EPR signals were doubly integrated to obtain the area, which was plotted versus time after UV-irradiation (Fig. 4A, inset). A reliable radical signal persists up to a month with a half-life of approximately one week. After two months, when no radical signal was observable, we took a 1 H NMR of the sample to see if the sample had degraded similarly to the solution study. Remarkably, the NMRs were identical to the initially synthesized materials indicating that 3 is photostable in its crystal form (Fig. S27 †). Next, we estimated the maximum concentration of radicals that could be generated through UV-irradiation by plotting the area of the EPR signal versus time exposed to UV light. The amount of radicals increases steadily with irradiation time (1 to 6 h) as seen in Fig. 4B part I. The plot of the double integration of the EPR signal versus time starts to plateau after 7-11 h of UVirradiation with the crystals of 3 turning deep brown in color during the process (Fig. 1B). The concentration of radicals was approximated using a calibration with standard solutions of Magic Blue in DCM. Comparing the area of the EPR spectra of the solid-sample versus the Magic Blue calibration can give an approximate concentration of radicals generated in the solidstate. After 11 hours of UV-exposure, 9.0 mg of 3 generated the same amount of radicals as 100 mL of a 0.82 mM solution of Magic Blue, suggesting that 1 in 150 molecules of 3 have a radical (or 1 in 300 TPA units, Fig. S25 †). Similar calculations for the 3.9 mg of the triply recrystallized sample were of a similar magnitude with $1 in 250 molecules exhibiting a radical after only 6 h of irradiation. With no noticeable degradation of 3 occurring after radical formation in crystalline samples (Fig. S27 †), we investigated if the radicals could be 'regenerated' after decay with repeated UV exposure. Typically, with chemical or electrochemical oxidations of TPAs to their corresponding radical cations, loss of the radical signal likely means the sample has degraded, and samples must be resynthesized. Remarkably, once the signal of 3 decays to half signal, irradiation with UV-light restores the radical concentration back to its maximum value (Fig. 4B, Part II). The samples were re-irradiated for 6 hours at the start of weeks 2 and 3 to regenerate the signal. As seen in Fig. 4B part II, the radicals decay at approximately the same rate over the three week long cycles. Also notable is that similar quantities of radicals are generated each time the crystals of 3 are UVirradiated, demonstrating the exceptional stability and reproducible nature of the assembled structure versus in solution. Next, we probed how the photogenerated radicals influenced the properties of the crystals as a whole. First, we compared the photophysics of the crystals before and after UV-irradiation (4 h). Both the absorption and emission were signifcantly quenched upon radical formation (Fig. S13 and S16 †). This was quite striking considering the radical concentration was relatively low compared to the bulk sample. Oxidation of TPAs to radical cations is known to quench the photophysical properties, 8 but typically it is quantitative in nature, at least in solution. This indicates that the generated radical is strongly delocalized to effect the whole system and behaves similarly to a radical cation. To further probe the nature of the radical, irradiated (4 h) brown crystals of 3 were subjected to SC-XRD and IR spectroscopy. No change was observed in the overall single crystal structure; however, we noticed a minor change in the amount of disorder in the crystal going from 91% major conformer to 95%. This is likely correlated to the specifc single crystal chosen for SC-XRD and is probably unrelated to radical formation. We expect that the TPA units are too bulky to self-correct during crystal formation leading to an array of different conformer percentages depending on the single crystal. For the IR studies, pre and post UV spectra were found to be nearly identical with no visible changes (Fig. S28 †). The combination of SC-XRD and IR suggest that either the radical is highly delocalized and/or is not concentrated enough to be characterized by these methods. Cyclic voltammetry (CV) was used to further characterize the electronic structure of 3 in solution. Since the radicals in the solid-state acted similarly to TPA radical cations, this method could further characterize the types of radicals generated in this system. The oxidation of 3 in DCM shows two pseudo-reversible oxidation waves at 1.0 V and 1.2 V vs. SCE (Fig. 5A). In comparison, the oxidation of parent compound 1 leads to a degradation that was immediately visible in the CV, 36 which suggests that 3 is more stable towards oxidation than the parent compound. Controlled potential electrolysis performed on 3 at +1.2 V passed at total of 3.98 electrons per molecule (Fig. S30 †). Thus, each oxidative wave in Fig. 5A is attributed to a 2e oxidation of the symmetric compound 3. Each electron is expected to come from the TPA units of the molecule, which would be consistent with previously reported results that have only one TPA unit per molecule. 36 For comparison, an electrochemical study on compound 2 showed that slower scan rates (40 mV s 1 ) were required to obtain pseudo-reversible oxidation waves also near 1.0 V and 1.2 V vs. SCE (Fig. S29 †). This indicates the rigidity of backbone of 2 has a clear impact on electron transfer kinetics. Bulk electrolysis on a $1 mM solution of 3 in DCM was performed at the frst oxidation peak to generate a radical cation in solution. The frst peak was chosen for the oxidation since it likely affords an overall dication with each TPA unit being oxidized once. Electrolysis was performed for $5 hours to afford a bright yellow solution; however, once electrolysis was completed the resulting sample was unstable at room temperature and turned teal within 15 min. EPR analysis showed no signal. Thus, the electrolysis was performed on a new sample that was immediately immersed in liquid nitrogen for transportation and the EPR recorded at 10 K. As seen in Fig. 5B, two peaks for the electrolytic sample can be seen at g-values of 2.031 and 2.002. The UV-generated radicals of crystalline 3 were also recorded at 10 K for comparison and showed no change in the line width and a slight shift in g-value to 2.002 compared to its room temperature spectrum, consistent with population of lower energy states. The peak at g-value ¼ 2.002 was consistent in the electrolytic solution sample and the UV-irradiated crystals 3. This suggests that crystalline 3 may form a similar radical species to electrolytic sample. Additionally, this is in good agreement with where TPA radical cations typically appear in EPR spectra. 5 Thus, it is likely that the photogenerated radicals in assembled 3 are similar to radical triphenylamine cations formed by electrolysis (at g-value ¼ 2.002), although it is not clear what anion is being formed in the crystalline sample for this process to occur. We are currently examining macrocyclic derivatives and are planning high feld EPR studies that could help probe this question. The second isotropic signal at g-value ¼ 2.031 was exclusive to the electrolyte sample and is attributed to degradation products. Self-assembly of TPA urea dimers can stabilize the UV generated organic radicals in stark contrast to their solution counterparts. The concentration of the radicals is readily controlled by irradiation time up to a maximum of 1 in $150 molecules. The presence of the radicals can be visualized simply through their photophysical quenching behavior. Advantageous to this system is that the radicals can be generated in the solidstate without noticeable degradation to the starting materials and display a half-life up to a week. Additionally, these radicals can be regenerated upon re-irradiation without any loss in radical concentration. A comprehensive study on how different halogen substituents on these TPA compounds influences the radical generation, stability, and concentrations may be invaluable in revealing the factors that govern the photophysics of these compounds. ## Conclusions In summary, a TPA methylene urea-tethered dimer was synthesized and readily afforded single crystals that organized the TPA through urea hydrogen bonding interactions. This solid-state assembly signifcantly stabilizes UV-generated radicals. Radicals formed in solution were unstable, as expected for incomplete para substituted TPA systems. In the solid-state, high quantities of radicals were formed, up to 1 in $150 molecules, which were persistent at room temperature with no observable degradation or signifcant changes in the single crystal X-ray diffraction. Further, radicals generated within the assembled framework have been shown to last up to a month with a half-life around a week. Most remarkably, after radical decay, radicals can be regenerated to their original maximum concentration with re-exposure to UV light. The photophysics of these materials were signifcantly quenched likely due to TPA hole transport properties even with relatively low radical concentration. Electrochemical evidence demonstrates that these compounds can be oxidized in solution at 1.0 V vs. SCE to generate radical cations, whose EPR spectra are similar to the UV-generated radicals in the solid-state. This suggests that the TPA radical cation is being formed in the solid-state and this electron transfer is reversible and reforms the parent compound over time. We are currently planning to carry out high-feld EPR experiments as well as Dynamic Nuclear Polarization Magic Angle Spinning solid-state C13-NMR to further examine this process. Future work includes the synthesis of additional halogenated on the TPA analogs to elucidate the factors that govern radical formation, persistence, and quantity. Understanding how assembly enhances the stability of radicals would be exceedingly helpful in the end goal of making better conductive and magnetic materials that incorporate TPA scaffolds. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "UV-irradiation of self-assembled triphenylamines affords persistent and regenerable radicals", "journal": "Royal Society of Chemistry (RSC)"}

high_resolution_31p_nmr_spectroscopy_generates_a_quantitative_evolution_profile_of_phosphorous_trans

## Abstract: Phosphorus metabolism and circulation are essential bio-physicochemical processes during development of a plant and have been extensively studied and known to be affected by temperature, humidity, lighting, hormones etc. However, a quantitative description of how various phosphorous species evolve over time has not been reported. In this work, a combined 31 P liquid and solid state NMR spectroscopic methodology is employed, supported by a new extraction scheme and data analysis method, to carry out a quantitative investigation of phosphorous circulation in germinating sesame seeds in dark and under illumination with and without adding a growth hormone. The spectra show that only slight changes occur for phosphorous metabolism at the initial stage but a rapid change takes place between 48-96 hours after germination is started. The metabolism is found to be temperature dependent and affected by illumination and hormone. However, neither illumination nor hormone affects the final residual concentration of phytin. Moreover, phytin does not flow out of cotyledon and the phosphorous flowing to other parts of the plant is always in the inorganic form. The overall evolution profile of phytate consumption is found to be a Gaussian decaying function. These findings can be explained with a dynamic model on phytin conversion. Phosphorous is an essential element for organisms, involving many important physiological functions such as energy storage and transfer, cellular membrane, skeletal support etc. . In a plant, phosphorous accounts for about 0.2-0.5% of total mass and is the most important element only next to nitrogen . Therefore, for over one century, the research on phosphorous metabolism and translocation of phosphorous has been a vitally important subject and much advance has been achieved on phosphorous biochemistry, biophysics and physiology . However, there are still many open questions unanswered yet, e.g. the spatiotemporal details of a specific metabolism process and interactions between phosphorous-containing molecules/ions and other biomolecules . Indeed, for plants, how phosphorous is taken up from soil and how it is mobilized and translocated during germination of a seed are not fully understood. These questions have fundamental significance for plant biochemistry and physiology. They are also keenly related to the yield and nutrition composition of crops, genetic modification of crops, environmental ecology, and phosphorous conservation on land etc . A central and widely studied subject in phosphorous metabolism is related to the destiny of phytate (phytin) . While various experimental techniques have been used to study the biophysical chemistry of key biomolecules such as phytate, NMR spectroscopy has been shown a unique tool , mainly because it can obtain structural and dynamics information at the atomic level simultaneously. When combined with prudential isotope labelling, solid state NMR spectroscopy has been proven a powerful method for studying the biochemistry of plants in vitro or in vivo, as beautifully illustrated by Schaefer group . In the past years, NMR spectroscopy has been employed to investigate the phosphorous fate and phytate dynamics in a variety of systems from soil, manure, poultry litter, foods and fruits to living organisms . Shand et al. 17 applied 31 P solid state NMR to study phosphorous distribution in different types of soils. Cade-Menun et al. 18,19 investigated the destiny of phosphorous in soil and its affection on environment and agriculture. Hunger et al. 20 analyzed the phosphorous in poultry litter using solid state NMR spectroscopy. Jayasundera et al. 21 and He et al. 22 measured different forms of phosphorous in dairy manures. He et al. 23 studied the structural characteristics of a series of metal species of phytate using solid state NMR and x-ray absorption near edge structure (XANES). They showed that intensive sidebands in solid state NMR spectra, indicating highly anisotropic microenvironment of metal species of phytate. The phytate degradations by lactic acid bacteria in yeast and dough fermentation were measured by Reale et al. 24 . The phosphorous distribution and evolution in wetland was studied by Cheesman et al. 25 . More advanced 31 P NMR methods such as heteronuclear NMR experiments involving 31 P for studying the large-scale structure 26 or complex biochemical reactions 27 in biological systems were also demonstrated to be feasible and powerful. Gambhir et al. 28 studied the phosphorous metabolites of the seeds of wheat, soybean and mustard using liquid state NMR spectroscopy. The cytoplasmic 31 P signals over a range of moisture contents were recorded and examined. This enabled them to measure the metabolically important cytoplasmic pH value to provide clear evidence for the existence of a hypoxic state in developing seeds. However, to the best of our knowledge, although 31 P NMR research of living plants has been conducted in previous works , monitoring the circulation of phosphorus over an entire course of the development of a plant seed has not been conducted. The 31 P NMR spectroscopic results has not been analyzed in perspective of the biophysical mechanism of phosphorous metabolism and circulation. In addition, the resolution of the previous studies was not optimized which limited the quantitativeness of the spectra. In this work, therefore, we explore the application of 31 P NMR spectroscopy to monitor the germination of sesame seed. To achieve high resolution in the NMR spectra, possible paramagnetic species were carefully removed. It is worth mentioning that quantitativeness has been a challenge in applying 31 P NMR spectroscopy for either in vivo or in vitro research because of the diverse microenvironments for phosphorous in an organism. Insufficient quantitativeness hinders quality of research and causes misleading conclusions. In present work, quantitativeness is achieved via a novel sample preparation scheme and a modified method for data analysis, in addition to high spectral resolution. The high quality data enable us to obtain quantitative time evolution of phosphorous metabolism during the entire course of seed germination. The evolution curves can be rationalized by a practical model on the enzyme dynamics during seed germination. ## Results and Discussion Sesame seed (Sesamum indicum L.) was purchased from a local vendor. Gibberellin GA 3 (gibberellic acid, C 19 H 22 O 6 ) of purity >99% was purchased from Aldrich and used without further purification. Ethylenediaminetetraacetic acid (EDTA, C 10 H 16 N 2 O 8 ) for removing metal elements from the sample by chelation was purchased from Aldrich. HClO 4 and K 2 CO 3 used for sample processing were purchased from Aldrich. Deionized water was home made with a resistivity larger than 0.5 MΩcm. The sketchy diagram of sample preparation and experimental procedure is shown in Fig. 1. About 1 g of sesame seed was taken and placed in pure water in a culture dish with controlled lighting conditions (sunlight, dark), humidity (20%) and temperature (30 °C). Timing is recorded with a computer. At the selected time point, the seed was taken out, pulverized into powder and mixed with 3 ml of 1 M HClO 4 aqueous solution in a test tube for 10 mins. The mixture then was placed in a centrifuge (3000 rpm) for 10 mins. After centrifugation, the supernatant was taken and added with 500 μl of 0.1 M EDTA to form a homogeneous mixture so that the metal elements were chelated by EDTA. The mixture then was titrated with 3 M K 2 CO 3 to a pH value of 7.6. The solution was again put in a centrifuge (3000 rpm) for 10 mins to separate the metal-chelating EDTA and the precipitates (e.g. of KClO 4 ) from the solution. The supernatant was then used for analysis with high resolution liquid state NMR spectroscopy. The 31 P MAS solid state NMR spectra of sesame seed at two different magnetic fields are shown in Fig. 2. The spectra of both dry and germinating seeds show severely broadened peaks even under high speed sample spinning. There are two major origins responsible for this large broadening. The first is the strong dipolar coupling between phosphorous and paramagnetic ions such as iron and manganese, leading to paramagnetic broadening. The second is because sesame seed is highly inhomogeneous and phosphorous nuclei are located in a large number of different microenvironments, leading to a broad isotropic chemical shift dispersion. Comparing the linewidths of the central peaks of the dry seed at 200 MHz and 500 MHz, it is found that the linewidth cannot be explained with isotropic chemical shift dispersion only. With the two linewidths (17.7 ppm at 200 MHz and 9.85 ppm at 500 MHz), we can estimate that the isotropic chemical shift dispersion is 4.9 ppm and the paramagnetic broadening is about 1050 Hz. The presence of sidebands even at 10 kHz suggests large anisotropy of chemical shift of phosphorous in the seed and underscores that phosphorous spins in a sesame seed are largely immobile. However, for a germinating seed, the intensities of the sidebands are much weaker and the linewidth is much narrower, indicating that the phosphorous in a germinating seed is much more mobile than in a dry seed. The 4.38 ppm linewidth is largely from isotropic chemical shift dispersion, very close to the estimated values from the dry seeds. Although the solid state 31 P NMR spectra provide interesting information on seed germination, the isotropic chemical shift dispersion cannot be resolved, therefore, we cannot use solid state NMR spectra to investigate the change of individual phosphorous components during germination. Evolution profile under illumination. With liquid state 31 P NMR spectroscopy, the major components are well resolved. Shown in Fig. 3 are the 31 P spectra of sesame seed at different germinating times under normal sunlight illumination. The peak at 3.27 ppm corresponds to the inorganic phosphorous (Pi) and the remaining peaks are from organic phosphorous Po (phytin) 24 . The intensity ratio Pi/Po at each germination time is shown in each spectrum in Fig. 3. As anticipated, with germination proceeding, this ratio increases. Initially, the increase is Evolution profile in dark. Figure 4 shows the 31 P spectra of sesame seed at different germinating times in dark. Both the spectra and visual inspection (the photos) indicate that sesame seed can germinate in dark. Analogous to germination in light, with germination proceeding, Pi/Po ratio increases, initially, the increase being very slow and becoming rapid after 48 hours of germination. However, the increase of the ratio is much slower than that for seed germinating under illumination. After about 72 hours, Pi/Po is only about 1/2. It takes another 24 hours for the ratio to reach about 3 (supporting information Figure S1). Comparing Figs 3 and 4, the effect of illumination is obvious. Although sesame seed can germinate in dark, the growing speed is slower. Evolution profile in dark with GA 3 . When a sesame seed germinates in dark but the culture is added with GA 3 , the growth trend is similar to that under illumination, as shown in Fig. 5. However, some appreciable differences are worth mentioning. The initial Pi/Po ratio for germination in dark with GA 3 is bigger than that for germination under illumination, but after about 48 hours, the former is smaller than the latter. This suggests that the effect of GA 3 in the initial stage of germination is bigger than that of illumination, but smaller in the latter stage. This is probably because in the initial stage, GA 3 which is imbibed into seed can affect phytase but light cannot get inside the sesame seed to affect phytase. After about 48 hours when the seed buds and can absorb light, the illumination effect is more significant than GA 3 . From above results, it is clear that the germination speed is significantly affected by light and hormone. However, after a sufficient long time, virtually all phytin is consumed, i.e., neither illumination nor hormone affects the final residual concentration of phytin. By analyzing the spectra of seedlings (Figures S2 and S3), it is found that phytin does not flow out of cotyledon and the phosphorous flowing to other parts of the plant is always in the form of inorganic phosphorous. ## From phytin evolution profile to phytase kinetics and dynamics. To quantitatively analyze the phosphorous evolution during germination, we plot the phytin degradation with respect to germination time for above three cases as shown in Fig. 6(A). We find that all the three curves can be fitted with a Gaussian decay function with different decaying factors. Fitting with other functions such as exponential or power law leads to much larger errors. The phytin degradation (or phytate consumption) not only quantifies the kinetics of the chemical reaction from phytin to inositol but also provides important information on the action of phytase. This Gaussian behavior is different from the consumption curves of fermentation of yeast or dough 24 , or in soil samples 25 . This suggests that the enzyme dynamics of phytase in sesame germination is very different from that in fermentation and non-Michaelis-Menten behavior must be taken into consideration in order to quantitatively explain our data and the data from previous studies. An obvious fact about seed germination is that the seed must be sufficiently hydrated. This provides, among other conditions, an environment that the enzymes and reactants possess high mobility. This means, during the enzyme-substrate binding, diffusion of both the substrate and enzyme must affect the overall efficiency of catalysis, hence consumption of phytin. Therefore, we use a dynamic model for phytin consumption as shown at the bottom of Fig. 7. In this model, the conversion of phytin to inositol (and other products) and the activity of phytase assume a first-order reaction-diffusion mechanism, leading to an overall Gaussian type turnout trend of inositol. We denote the concentration of inositol at any given location r and time t as [I](r, t) and the relative diffusion tensor of phytin with respect to phytase D, then the overall reaction-diffusion equation is expressed as where k 0 is the reaction rate without considering relative diffusion. There is no general analytical solution to Eq. (1) but some approximate solutions for special boundary conditions have been proposed 35,36 and more recently some exact solutions for special cases have been reported 37 . Numerical solutions either based on solving coupled differential euqaitons 38 or based on molecular dynamics simulation 39,40 have also been obtained for a number of systems. We find that in our case, however, a simplified model is sufficient to provide us quantitative description of our experimental data. In this model, the diffusion tensor is assumed to be isotropic so that the problem can be converted into a one-dimensional case. For a germinating seed where the molecules have large mobility, this is an acceptable approximation. Under this assumption, the solution of Eq. ( 1) is given by. x Dt k t /4 1/2 2 0 The total consumption of phytin at any given time t is an integration over all possible phytase that can access phytin. Suppose all phytase molecules within a fixed distance (−b, b) from phytin will promote the reaction. Then phytin consumption follows where erf stands for error function. Some representative solutions are plotted in Fig. 6(B) which satisfactorily reproduce the trends shown in Fig. 6(A). From Eq. (2), it is obvious that the consumption curve of phytin would follow a pure exponential trend if the relative diffusion is neglected. This result clearly shows the importance of relative diffusion between phytin and phytase in the quantitative description of phosphorous translocation in a germinating seed. It is also noteworthy that when the diffusion is too fast, as shown in the cyan and blue lines in Fig. 6(B), the Gaussian type behavior is lost and the overall trend looks more like an exponential decay. We notice that an earlier study by Frias et al. 41 on legume seeds did show non-exponential behavior, but the data are only available for three time points, making quantitative comparison infeasible. A more recent study on legume seeds by Abdel-Gawad et al. 42 shows a clear Gaussian type activity for both phytase, consistent with our above experimental results on sesame seed. Therefore, Gaussian kinetics may be a universal characteristic for phytase action for all seeds. Although studies on more seeds are required to confirm this hypothesis, the above dynamic model of phytase action seems rather reasonable. In summary, the 31 P solid state MAS NMR spectra show that phosphorous in a dry seed and that in a germinating seed have different microenvironments and dynamics. However, solid state spectra cannot be resolved to show different types of phosphorous. With the sample preparation protocol that provides homogeneous solutions and removes the possible paramagnetic ions, high resolution 31 P NMR spectra of sesame seed extracts could be obtained with all major phosphorous species clearly resolved, that makes it possible to conduct quantitative analysis of phytin degradation. This quantitative study of phosphorous translocation during sesame seed germination in various conditions (in light, in dark and in GA 3 ) with 31 P NMR spectroscopy enables us to monitor the time evolution of phosphorous conversion and to obtain a quantitative evolution profile over the entire course of germination. The effects of light and GA 3 are clearly demonstrated and compared. In all cases, the evolution profile shows clear deviation from exponential function and displays a Gaussian decaying trend. These results led us to establish that the dynamic model on phytin conversion under phytase activation must take into account the relative diffusion the reactant and enzyme. We believe this model may be applicable to other seeds although experimental data from more seeds are required. ## Methods All liquid state NMR experiments were performed on a Bruker Avance 300 MHz liquid-state NMR spectrometer. All solid state NMR experiments were performed on a Varian UNITY INOVA 500 MHz solid-state NMR spectrometer and a Varian UNITY INOVA 200 MHz solid-state NMR spectrometer. For all experiments, a single 90° pulse excitation was used. For liquid-state NMR measurements, the major experimental parameters were: the 90° pulse width 10.8 μs, recycle delay 2 s, acquisition time 0.66 s and number of transients 12500. On the solid-state NMR 500 MHz spectrometer, the major experimental parameters were: the 90° pulse width 3.5 μs, recycle delay 2 s, acquisition time 0.02 s and number of transients 4000. The sample was spun at the magic angle (MAS) with a speed from 2 kHz to 10 kHz to improve spectral resolution by removing the anisotropies from chemical shift and heteronuclear dipolar couplings between phosphorous and hydrogen. To further improve spectral resolution, a continuous heteronuclear decoupling ( 31 P-1 H) with a power of 80 kHz on 1 H channel was also used during acquisition. On the solid-state NMR 200 MHz spectrometer, the major experimental parameters were: the 90° pulse width 3 μs, recycle delay 1 s, acquisition time 0.03 s and number of transients 40000. The sample spinning speeds were from 2 kHz to 6 kHz. A continuous heteronuclear decoupling ( 31 P-1 H) with a power of 60 kHz on 1 H channel was used during acquisition. The experimental FID data were zero-filled to 4096 data points and filtered with a Lorentzian window function with a line broadening factor of 2 Hz before Fourier transform. The spectral data were then transferred to a desktop computer for further analysis. To prevent bias in selecting sesame seeds and guarantee the representativeness and reproducibility of NMR spectra, we used tens of seeds for each germination experiment and chose the typical seeds for NMR experiments. The NMR measurement was repeated for at least two times for each spectrum. The good reproducibility was ensured. The overall errors in the germinating curves shown in Fig. 6 were less than 3%. The satisfactory agreement between experimental and theoretical data is also a good indication that the experimental data we obtained are of high quality.

chemsum

{"title": "High Resolution 31P NMR Spectroscopy Generates a Quantitative Evolution Profile of Phosphorous Translocation in Germinating Sesame Seed", "journal": "Scientific Reports - Nature"}

carbon_dioxide_hydrogenation_catalysed_by_well-defined_mn(<scp>i</scp>)_pnp_pincer_hydride_complexes

## Abstract: The catalytic reduction of carbon dioxide is of great interest for its potential as a hydrogen storage method and to use carbon dioxide as C-1 feedstock. In an effort to replace expensive noble metal-based catalysts with efficient and cheap earth-abundant counterparts, we report the first example of Mn(I)-catalysed hydrogenation of CO 2 to HCOOH. The hydride Mn(I) catalyst [Mn(PNP NH -iPr)(H)(CO) 2 ] showed higher stability and activity than its Fe(II) analogue. TONs up to 10 000 and quantitative yields were obtained after 24 h using DBU as the base at 80 C and 80 bar total pressure. At catalyst loadings as low as 0.002 mol%, TONs greater than 30 000 could be achieved in the presence of LiOTf as the co-catalyst, which are among the highest activities reported for base-metal catalysed CO 2 hydrogenations to date. ## Introduction Carbon dioxide (CO 2 ) is the end product of combustion of organic matter and its increasing concentration in the atmosphere is causing great concerns due to its negative effects on the environment. On the other hand, it can represent a cheap, readily available and abundant carbon feedstock, and its utilisation for the production of value-added chemicals and fuels is a hot topic of research today. 1 Among the possible target reactions, the catalytic hydrogenation of CO 2 to formic acid (FA) and its derivatives is a subject of increasing interest, as FA is a widely employed commodity chemical and can be used as a hydrogen storage material. 2 In recent years, a number of efficient homogeneous catalysts for the hydrogenation of CO 2 to FA under mild reaction conditions have been developed, most of them based on expensive and rare noble metals such as iridium and ruthenium. 3 The replacement of the scarce and expensive noble metals with cheaper, earth abundant, and less toxic frst-row transition metals would enhance the sustainability and industrial applicability of these hydrogenation reactions. Recently, iron-and cobalt-based homogeneous catalysts have been reported for this reaction (Scheme 1), 4 and in some cases show a catalytic activity comparable to that observed for some noble metal catalysts. Beller and co-workers described Fe(II)and Co(II)-PP 3 catalysts (PP 3 ¼ tripodal tetraphosphines) that could reach signifcantly high turnover numbers (TONs) for CO 2 hydrogenation to FA and derivatives. 4g,i,j Milstein reported that the PNP pincer complex trans-[Fe(PNP-tBu)(H) 2 (CO)] promotes the reaction already at a low pressure with TONs up to 788. 4h In 2015, Hazari, Bernskoetter and co-workers described the catalytic activity of a library of Fe(PNP) pincer complexes exhibiting very high activities which could be even further enhanced using Lewis acid co-catalysts achieving TONs up to ca. 59 000. This sets the current state-of-the-art in CO 2 hydrogenation to FA with a non-precious, earth-abundant metal catalyst. 4b More recently, the benefcial effect of Lewis acid co-catalysts was demonstrated also for related cobalt pincer complexes. 4k Scheme 1 Examples of efficient base-metal catalysts for the hydrogenation of carbon dioxide to formic acid or formate. Recently, 4a we demonstrated the remarkable catalytic activity of the Fe(II) PNP pincer complexes [Fe(PNP NH -iPr)(H) 2 (CO)] (Fe1) and [Fe(PNP NMe -iPr)(H) 2 (CO)] (Fe2) for CO 2 hydrogenation reaching TONs of up to ca. 10 000 in the presence of a base with quantitative yields of formate at 80 C (Scheme 1). Complex Fe2 also exhibited a signifcant activity even at room temperature affording FA in quantitative yield (TON ca. 1000) in 24 h. Based on these results, we decided to explore the activities of other earthabundant metal complexes starting with low valent manganese compounds. As yet, in contrast to iron, manganese plays a minor role, despite the fact that manganese, after iron and titanium, is the third most abundant transition metal in the earth's crust, and ubiquitously available and essentially non-toxic. Reports on homogeneous Mn(I) catalysed reactions only appeared very recently in the literature from the groups of Milstein, 5 Beller, 6 Kirchner, 7 Kempe, 8 and Boncella 9 and their collaborators. Here, we report the frst efficient hydrogenation of CO 2 to FA catalysed by the well-defned Mn(I) catalysts [Mn(PNP NH -iPr)(H)(CO) 2 ] (Mn1) and [Mn(PNP NMe -iPr)(H)(CO) 2 ] (Mn2). These complexes are isoelectronic with our previously tested Fe(II) catalysts (Fe1 and Fe2), and Mn1 was shown very recently to be highly active in the dehydrogenative coupling of alcohols and amines to give imines 7a as well as in the synthesis of substituted quinolines and pyrimidines. 7b ## Results and discussion The synthesis of the complexes [Mn(PNP NH -iPr)(H)(CO) 2 ] (Mn1) and [Mn(PNP NMe -iPr)(H)(CO) 2 ] (Mn2) was carried out as recently described by some of us. 7a In addition, X-ray quality crystals of Mn2 were grown from toluene/pentane and the corresponding solid state structure was obtained (Fig. 1 and ESI †). The structural data are comparable with those obtained for Mn1, 7a with slightly shorter bond distances for Mn1-C20 (1.700 vs. 1.747 ) and longer distances for Mn1-C21 (1.784 vs. 1.775 ) and Mn1-H1 (1.80 vs. 1.46 ). ## Catalytic CO 2 hydrogenation The catalytic activity of the Mn(I) pincer complexes Mn1 and Mn2 was initially tested in a THF/H 2 O (10 : 1) solvent mixture in the presence of DBU (DBU ¼ 1,8-diazabicycloundec-7-ene) as the base, at 80 C, under 80 bar total pressure (H 2 : CO 2 ¼ 1 : 1), for 24 h. 10 Selected results are reported in Table 1. Using a Mn1/DBU ratio of 1 : 1000, FA was obtained in quantitative yields with respect to DBU (TON ¼ 1000, entry 1). At lower catalyst loadings (Mn1/DBU ¼ 1 : 10 000; [Mn1] ¼ 0.18 mmol mL 1 ), FA was formed in a 55% yield after 24 h (TON ¼ 5520, entry 2). Notably, under such reaction conditions Mn1 had a superior performance compared to Fe1 and Fe2, which in turn gave the TONs of 2080 and 2750 respectively (entries 3 and 4). Furthermore, increasing the reaction time for the Mn1 tests to 48 h signifcantly improved the yield of formate (86%, TON ¼ 8600, entry 5). In contrast, Mn2 showed a lower catalytic activity, affording FA in only a 10% yield with a TON of 1010 after 24 h (entry 6). For 3 h runs under standard conditions, Mn1 gave a TON ¼ 475 (yield ¼ 48%), and Mn2 gave a TON ¼ 96 (yield ¼ 10%). Decreasing the catalyst concentration to 0.036 mmol mL 1 (Mn1/DBU ¼ 1 : 50 000) afforded FA in a lower yield (16%) but with a signifcantly increased TON (9100, entry 11), indicating that the catalyst activity strongly depends on the catalyst/DBU ratio. The promoting effect of adding H 2 O (10%) to THF may be attributed to a water-assisted dissociation of the formate ligand resulting from the CO 2 insertion into the Mn-H bond under catalytic conditions (see DFT calculations), as well as to an improved stabilization of the FA/DBU adduct due to hydrogen bonding. This is supported by the observation that the FA/DBU adduct is immiscible with THF, but cleanly dissolves in H 2 O. However, a striking difference between the Mn and Fe catalysts was observed when running the tests in anhydrous THF. Whereas with the Mn(I) complexes FA was formed even in the absence of water, albeit with lower TONs (4400 and 420 for Mn1 and Mn2, entries 7 and 8, respectively), and Fe1 and Fe2 were completely inactive under these conditions (entries 9 and 10). We thus reexamined the solvent effects for the Fe catalysts Fe1 and Fe2. 4a While their catalytic activity in the THF/H 2 O mixtures is comparable (entries 3 and 4), we observed a substantial difference running the tests in EtOH. In this solvent, the catalytic activity of Fe2 is substantially improved (TON ¼ 10 000; >99% yield, entry 15), while Fe1 resulted in being totally inactive (0% yield, entry 14), confrming the data previously reported. 4a The reason for the latter observation is that for Fe1 EtOH prevents the formation of the key catalytic intermediates [Fe(PNP NH -iPr)(H) 2 (CO)]. This was not the case for Fe2, which readily affords the corresponding dihydrides in the presence of the base and H 2 . 11b We then studied the effect of EtOH in the hydrogenation of CO 2 catalysed by Mn1 and Mn2. To our delight, using EtOH as the solvent improved the catalytic performance of Mn1 (entry 12) signifcantly, which afforded FA in an 80% yield (TON ¼ 8000). In turn, only a minor improvement was observed for Mn2 (entry 13). ## Effect of LiOTf In an effort to further improve the Mn-DBU catalytic system, and in line with literature data on other Fe-pincer systems, 4b we decided to test the effect of a Lewis acid additive in combination with Mn1, choosing LiOTf as it has previously shown to give the highest promoting effect. 4b The results are summarized in Table 2. As previously observed for Fe2, 4a the addition of LiOTf only had a minor effect on the catalysis in the presence of Mn1 when the reaction was run in EtOH (entries 1 and 2). In contrast to that, an enhancement effect was observed when using THF/H 2 O (10 : 1) as the solvent mixture. At a catalyst concentration of [Mn1] ¼ 0.18 mmol mL 1 and in the presence of 0.5 mmol of LiOTf (Mn1/DBU ¼ 1 : 10 000; Mn1/LiOTf ¼ 1 : 1000), FA was obtained in quantitative yields within the frst 24 h of the reaction (TON ¼ 10 000, entry 3). Remarkably, the reaction also proceeds at room temperature, albeit with a TON an order of magnitude lower (TON ¼ 1000, entry 4). Lower catalyst concentrations of 0.036 and 0.018 mmol mL 1 resulted in higher TONs but lower yields (entries 5 and 6). A longer catalytic run (entry 7) indicated that the catalyst remains active for up to 48 h, affording FA with a TON ¼ 26 600, a ca. two-fold increase compared to the 24 h catalytic run (entry 5), suggesting a constant reaction rate with an average TOF of ca. 550 h 1 . Increasing the LiOTf amount to 1.0 mmol (Mn1/LiOTf ¼ 1 : 5000) resulted in an increased TON ¼ 16 700 (entry 8), whereas a further increase to 1.5 mmol gave a slightly decreased TON ¼ 12 420 (entry 9). As previously suggested, 4b such an effect may be attributed to the limited LiOTf solubility in such a solvent mixture. The effect of the temperature on the catalysis was tested by increasing it from 80 to 100 C, which resulted in a ca. 2-fold increase in the TON to 31 600 after 24 h (63% yield, entry 10). However, increasing the temperature further to 115 C resulted in the formation of an unidentifed side product (4) in a ca. 25% yield with respect to DBU, along with a yield of formate of ca. 62% (entry 11). Trace amounts of the same by-product (<5%) were also observed at 100 C. Careful examination of the 1 H and 13 C{ 1 H} NMR spectra showed new signals (d H ¼ 8.00, d C ¼ 161.7) typical for N-formyl groups, suggesting that N-formylation of DBU occurs as a side reaction at high temperatures (see ESI †). This attribution was confrmed by obtaining the same product independently from the reaction of FA with DBU (1 : 1) at 120 C ## Mechanistic studies Details on the reaction mechanism were obtained for the most active catalytic system, complex Mn1, by a combination of experimental NMR techniques and DFT calculations. The reactivity of complexes Mn1 and Mn2 towards CO 2 was initially investigated in a stoichiometric reaction by NMR and IR techniques. When a solution of Mn1 in THF was stirred under an atmosphere of CO 2 (1 bar) for ca. 1 min, the immediate formation of the formate complex [Mn(PNP NH -iPr)(CO) 2 (k 1 -O-OC(O)H)] (Mn3) was observed (Scheme 2). The complex precipitated from the solution in essentially quantitative yield (see ESI †). This complex is characterised by a 1 H NMR singlet at d ¼ 8.21 ppm for the proton of the formyl ligand. In the IR spectrum, the characteristic n CO frequencies for the cis-CO and for the formyl carbonyl ligands were observed at 1923, 1842, and 1593 cm 1 , respectively. Mn3 could also be obtained by reacting Mn1 with 1 equiv. of HCOOH in THF at room temperature. In the case of Mn2, no reaction was observed with either CO 2 or HCOOH under the above conditions. It was possible to prove the reversibility of the CO 2 insertion in Mn1 by reacting Mn3 with H 2 (70 bar) in the presence of excess DBU at room temperature in THF (Scheme 2), whereas no reaction took place in the absence of base. Interestingly, isolated Mn3 was also active in the catalysis, giving comparable activity to Mn1 under the same test conditions (Table 1, entry 16 vs. 2), as expected for a reaction intermediate. The reactivity of Mn1 with CO 2 to give the formate complexes is somewhat remarkable, since the isostructural and isoelectronic cationic Fe analogues cis-[Fe(PNP NR -iPr)(CO) 2 H] + (Fe3, R ¼ H and Fe4, R ¼ Me) were found to be catalytically inactive for the hydrogenation of ketones and aldehydes, 11b as well as in stoichiometric reactions with CO 2 . We reasoned that this could be related to the electron density around the metal and in particular with the M-H bond. Indeed, the corresponding atomic charges (NPA, see ESI †) were found to be C Mn ¼ 0.92/C H ¼ 0.13 in Mn1 and C Fe ¼ 0.53/C H ¼ 0.04 in Fe3, showing an electron richer metallic centre and hydride ligand in the case of Mn1 (Scheme 3). Moreover, the M-H bond is weaker in the case of M ¼ Mn as shown by the corresponding Wiberg indices 12 (WI Mn-H ¼ 0.43, WI Fe-H ¼ 0.47), indicating a more reactive hydride for Mn1 than for Fe3. The free energy balances (gas phase) calculated for the reaction of the CO 2 insertion into the M-H bond (Scheme 3) corroborate the previous conclusions, including the absence of the reactivity of Mn2 toward CO 2 at room temperature. A deeper insight into the reaction mechanism was obtained by DFT calculations. 13 Free energy profles obtained for the entire reaction are presented in the ESI. † An explicit water molecule was considered in the model providing H-bond stabilisation for the intermediates (see ESI †). Two alternative mechanisms were investigated for the most active Mn catalyst Mn1. In the frst mechanism (Scheme 4, left and Fig. S9 in ESI †), a purely metal-centred mechanism is considered, i.e. without the participation of the N-H bond of the PNP ligand. This path starts with a nucleophilic attack of the hydride in intermediate A to the carbon dioxide C-atom, in an outer sphere reaction. This affords an intermediate (B) with a formate ligand bound to Mn in a C-H s-complex, with an accessible barrier of 9.6 kcal mol 1 . 14 B is less stable than the initial reactants (A) by DG ¼ 5.2 kcal mol 1 . From B, two pathways are possible. Formato ligand isomerisation can be achieved through a transition state with the energy of 13.2 kcal mol 1 , yielding a k 1 -O formate complex (C, corresponding to isolated Mn3), 9.7 kcal mol 1 more stable than the reagents, representing a resting state of the catalyst. Alternatively, the exchange of the formate in B by one H 2 molecule can give a dihydrogen complex (F 0 ), 7.0 kcal mol 1 less stable than the initial reagents (A). The H 2 coordination step has the highest energy transition state of the entire mechanism (DG s ¼ 17.9 kcal mol 1 ). Finally, the deprotonation of the dihydrogen intermediate F 0 (by either formate or DBU) regenerates the initial hydride complex Mn1 while the fnal product is obtained as [DBUH][HCOO]. Overall, the reaction has a barrier of 27.6 kcal mol 1 , considering C as the catalyst resting state, and the entire reaction is thermodynamically favorable with DG ¼ 11.2 kcal mol 1 . Scheme 2 Reaction of Mn1 with CO 2 (1 atm) and FA (1 equiv.) to give Mn3. For the former reaction, the reversibility from Mn3 to Mn1 was also demonstrated. [HCOO]. In turn, the so-obtained fve-coordinated Mn neutral complex, bearing a formally negative PNP ligand (Fig. S10 in ESI †), readily coordinates to H 2 and the resulting Mn(h 2 -H 2 ) complex (J) is 4.2 kcal mol 1 more stable than the starting species A. Water-(or protic solvent) assisted H-H bond splitting from complex J follows, yielding back the initial catalyst by forming a Mn-H hydride bond and reprotonating the PNP ligand. This last step has a transition as high as 20.0 kcal mol 1 , which is the highest of this path. The overall barrier for the reaction is 29.7 kcal mol 1 (with respect to C), which is within 2 kcal mol 1 from the frst pathway considered. We then turned our attention to the effects of the Li additive. We investigated the relative stabilities of the intermediates B and C in the presence of the Li(THF) 2 + adducts (Fig. 2 Based on the results of the catalytic experiments, however, we conclude that the bifunctional pathway must have a major role in the CO 2 hydrogenation in the presence of Mn1. In fact, the activity of Mn2, whose PNP structure does not allow for metalligand cooperation due to the NMe groups instead of NH groups, is much lower than that of Mn1. ## Conclusions In conclusion, this study has shown for the frst time that CO 2 catalytic hydrogenation to FA can be achieved with high TONs and yields using well-defned hydrido carbonyl Mn(I) PNP pincer complexes in the presence of an added base (DBU) and a Lewis acid (LiOTf), paving the way for the use of manganese as an earthabundant and cheap metal for efficient Carbon Dioxide Utilization (CDU). DFT calculations showed that, in the case of the most active catalyst Mn1, the reaction can follow two competing routes, involving either a metal-centred or a ligand-assisted mechanism, based on the possible bifunctional role of the PNPNR ligand when R ¼ H. Further studies are in progress to assess the effects of ligand structural modifcations on this reaction and to expand the scope of the catalysts to other challenging CDU processes.

chemsum

{"title": "Carbon dioxide hydrogenation catalysed by well-defined Mn(<scp>i</scp>) PNP pincer hydride complexes", "journal": "Royal Society of Chemistry (RSC)"}

reproduction_of_vesicles_coupled_with_a_vesicle_surface-confined_enzymatic_polymerisation

## Abstract: Molecular assembly systems that have autonomous reproduction and Darwinian evolution abilities can be considered as minimal cell-like systems. Here we demonstrate the reproduction of cell-sized vesicles composed of AOT, i.e., sodium bis-(2-ethylhexyl) sulfosuccinate, coupled with an enzymatic polymerisation reaction occurring on the surface of the vesicles. The particular reaction used is the horseradish peroxidase-catalysed polymerisation of aniline with hydrogen peroxide as oxidant, which yields polyaniline in its emeraldine salt form (PANI-ES). If AOT micelles are added during this polymerisation reaction, the AOT -PANI-ES vesicles interact with the AOT molecules in the external solution and selectively incorporate them in their membrane, which leads to a growth of the vesicles. If the AOT vesicles also contain cholesterol, the vesicles not only show growth, but also reproduction. An important characteristic of this reproduction system is that the AOT-based vesicles encourage the synthesis of PANI-ES and PANI-ES promotes the growth of AOT vesicles. O ne way to try to understand "What is Life?" is to attempt the synthesis of minimal cells which contain the essence of "life" 1 . Here, we consider "minimal cells" as molecular assembly systems that show autonomous reproduction and Darwinian evolution features. Thus, a minimal cell is characterised by three important properties; (i) metabolism that extracts usable energy and chemical resources from the environment, (ii) self-reproduction that is recursive growth and division of a compartment, and (iii) evolvability that requires the essential biological aspects of genetic variation and its phenotypic expression . Among several gateways towards the synthesis of minimal cells, one of the soft matter approaches focuses on constructing minimal cells using well-defined vesicle membrane-forming molecules to hopefully reveal some of the underlying physical chemical mechanisms of the growth and reproduction of living systems . In this approach, preformed vesicles take up precursors of the membrane molecules from the environment, which are then transformed by a chemical modification into the membrane molecules of pre-existing vesicles. The transformed membrane molecules incorporate into the vesicle membrane, which leads to vesicle growth or even division into independent vesicles having similar properties as the mother vesicles . Such simple reproducing vesicle systems were initially developed by Luisi's group 10,11 , whereby the growth and division of oleic acid vesicles was induced by the hydrolysis of oleic anhydride into oleic acid and oleate. The group of Szostak developed a cyclic reproduction of multilamellar fatty acid vesicles 13,14 . By addition of oleate micelles, spherical multilamellar oleic acid vesicles divided into multilamellar daughter vesicles through a growth to thread-like vesicles. Upon further micelle addition, the growth-division cycle continued in a cyclic manner. A unique feature of this mechanism is that growth followed by division is promoted by an affinity between specific chemicals (e.g. dipeptide) and fatty acids, which might bring Darwinian evolution 15 . A chemically completely different reproducing giant vesicle system was studied by Sugawara's group 16,17 . It is based on the design of engineered amphiphiles and their precursor molecules. The precursors are transformed to the amphiphiles by hydrolysis, which results in the growth and division of vesicles composed of the amphiphiles. In addition, the same group succeeded to couple this vesicle reproduction system with the amplification of encapsulated DNAs 18 . A similar protocol was developed by Devaraj's group 19,20 . In this study we demonstrate a specific example of the reproduction of giant unilamellar vesicles (GUVs) coupled with a vesicle surface-confined enzymatic polymerisation reaction. When aniline is polymerised with horseradish peroxidase (HRP) and hydrogen peroxide (H 2 O 2 ) in the presence of sodium bis-(2ethylhexyl) sulfosuccinate (AOT) vesicles, oligomeric and polymeric products are obtained which have characteristic properties of the emeraldine salt form of linear para-NC coupled polyaniline, abbreviated as PANI-ES 21 . Whenever the term "PANI-ES" is used we mean polyaniline products which consist of PANI-ES units. In this system, determining the molar mass of the formed products is not possible, since the final polymers become insoluble after AOT removal 22,23 . The reaction occurs localised on the surface of the vesicles (Supplementary Fig. 1 and Supplementary Note 1) through (PANI-ES) N-H⋯O-S (AOT) hydrogen bonding 24 , steric or electrostatic interactions 25 (Supplementary Fig. 2 and Supplementary Note 2). In this reaction, the vesicles serve as a kind of "template", which means that the outcome of this vesicle surface-confined reaction is influenced in a positive way by the vesicles 26 . In the work presented now, we find that the enzymatically formed PANI-ES on the AOT vesicles interact (through a kind of molecular recognition) with AOT molecules which are added to the external solution so that they are selectively incorporated into the AOT vesicle membrane. This results in the growth of the vesicles. The important observation is that the AOT vesicles not only promote the enzymatic synthesis of PANI-ES through the vesicle template effect, but that the obtained PANI-ES is also a kind of "information molecule" which promotes vesicle growth. This mutual promotion system fulfills the physical requirements of minimal cells 27,28 . In addition, by introducing to the AOT vesicle/PANI-ES system a second amphiphile having a negative molecular spontaneous curvature , we succeed in realising not only the growth of vesicles, but also the division of the grown vesicles (reproduction). On the basis of the observed mutual promotion mechanism, we develop a kinetic model, which predicts that this system has the potential for the experimental realisation of recursive reproduction and Darwinian evolution. Thus, the purpose of this study is to show that the "template"assisted enzymatic polymerisation occurring on the surface of the vesicles, known as PANI-ES synthesis on AOT vesicles, brings the reproduction of vesicles by feeding membrane molecules, which is a promising system for the development of minimal cells. ## Results Enzymatic polymerisation of aniline on AOT vesicles. An important feature of the enzymatic polymerisation of aniline in the presence of AOT vesicles 21,22 is the interaction between the vesicle-forming amphiphile and the formed products through specific PANI-ES-amphiphile contacts (Supplementary Fig. 2 and Supplementary Note 2). In agreement with previous studies using large unilamellar vesicle (LUV) suspensions 23 , we demonstrate that the outcome of this vesicle-assisted enzymatic polymerisation of aniline depends on the chemical structure of the vesicle template-forming molecules. Depending on vesicle type and experimental conditions, the obtained polyaniline products differ in their UV/vis/NIR absorption spectrum (Fig. 1). An important point is the fact that the reaction is localised on the surface of AOT vesicles and that the formed products (rich in PANI-ES structural units) are also localised on the vesicle surface 21 . For this to occur there must be attractive interactions between the vesicle surface and the polymerisation products (Supplementary Fig. 2 and Supplementary Note 2) 24,25 . To first re-examine and also extend earlier investigations on the effect of the polar head group structure of vesicle-forming amphiphiles on the PANI product formation 23 , we prepared three types of LUVs. They were composed of either (i) a zwitterionic phospholipid (DOPC), (ii) anionic phospholipids (DOPA, DOPG, or DOPS) or (iii) anionic amphiphiles with a sulfate or sulfonate head group (porcine brain sulfatides, which is a mixture of glycosphingolipid sulfates, AOT, or SDBS in a 1:1 (mol/mol) mixture with decanoic acid, abbreviated as SDBS/DA). DOPC stands for 1,2-dioleoyl-sn-glycero-3-phosphocholine, DOPA for 1,2-dioleoyl-sn-glycero-3-phosphate sodium salt, DOPG for 1,2dioleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) sodium salt, DOPS for 1,2-dioleoyl-sn-glycero-3-phospho-L-serine, and SDBS for sodium dodecylbenzenesulfonate. Since pure SDBS in dilute aqueous solution does not form vesicles 29 , LUVs composed of SDBS and DA were prepared 30 . The polymerisation reaction was triggered by adding H 2 O 2 (oxidant) to the LUV suspensions (20 mM NaH 2 PO 4 solution, pH = 4.3) containing aniline and HRP. HRP is the catalyst for initiating the reaction after being oxidised by H 2 O 2 21 . Under the optimal conditions elaborated for AOT LUVs, see Materials and Methods, the oxidation of aniline occurs very rapidly: 21 within less than one minute, 95% of the added H 2 O 2 is reduced and 75% of the aniline is oxidised (Fig. 3 of ref. 21 ). For all vesicle systems which we tested as potential templates for the reaction, the UV/vis/NIR absorption spectra of the reaction mixtures were measured after running the reaction for 24 h at room temperature (about 25 °C). The recorded spectra clearly depend on the type of LUVs used (Fig. 1). With zwitterionic-and overall neutral-DOPC LUVs (Fig. 1a), the spectrum shows strong absorption in the range of 280-320 nm (π → π * transition), a broad peak at 500 nm (indicative for extensive branching 31 and phenazine unit formation 25 ), and very low absorption at around 1000 nm (absence of π → polaron transition) as well as absence of an absorption peak at about 400 nm (no polaron → π* transition 32 ), i.e., absence of characteristic transitions for PANI-ES in polaron state with its unpaired electrons 21,22,33 . Therefore, the UV/vis/NIR absorption measurements indicate that highly branched PANI and/or phenazine units-rich polymers were obtained in the presence of DOPC vesicles. Thus, no significant "template" effect was observed for the polymerisation in the presence of DOPC LUVs. In a control experiment, the enzymatic polymerisation of aniline was also performed without vesicles. The obtained products showed a very similar absorption spectrum to that with the DOPC vesicles (Fig. 1a) with product precipitation upon storage. For the reaction in the presence of anionic phospholipid LUVs (DOPG, DOPA, or DOPS), new weak bands appear at 440 nm and 1000 nm (Fig. 1b, and for data obtained with vesicles from the related POPG 23 ). Thus, only a small amount of PANI-ES structural units in polaron state is synthesised in the presence of anionic phospholipid vesicles. For LUVs prepared from anionic amphiphiles having a sulfonate or sulfate head group (AOT, SDBS/DA (1:1), or porcine brain sulfatides), the spectra have high absorptions at 440 nm and 1000 nm and low absorption at 500 nm, indicating preferential synthesis of PANI-ES (Fig. 1c). All in all, the HRP-triggered polymerisation of aniline with H 2 O 2 as oxidant in the presence of vesicles produces different mixtures of reaction products, depending on the chemical structure of the vesicle membrane-forming amphiphile's head group. This is in full agreement with findings from similar experiments carried out with micelles or polyelectrolytes as templates 36 . Especially, the sulfonate or sulfate head group of the membrane-forming molecule is somewhat "encoded" in the chemical structure of PANI-ES through electrostatic, N-H⋯O-S hydrogen bonding, and steric interactions (Supplementary Fig. 1) 21,24,25 . Growth of vesicles coupled with "template" polymerisation. The first important result from our study is the demonstration that the growth of AOT vesicles can be coupled with the vesicleassisted enzymatic polymerisation of aniline. In other words vesicle growth is linked to a "template" polymerisation. If dissolved in pure water at a concentration which is above the critical micellization concentration (cmc ≈ 2.6 mM at room temperature 34 ), AOT molecules assemble into micelles; whereas in 100 mM NaH 2 PO 4 solution (pH = 4.3) and above the critical concentration for vesicle formation (cvc ≈ 0.4 mM 22 ) AOT forms bilayered vesicles. As a consequence, when a small amount of AOT micelles (in water) is added to an AOT GUV suspension (in NaH 2 PO 4 pH = 4.3 solution), the AOT molecules originally constituting the micelles will either form new vesicles or incorporate into the preformed GUV membranes, in analogy to what has been discussed before for fatty acid vesicles 37 . We then investigated whether AOT GUV growth is observed and whether the simultaneous enzymatic formation of PANI-ES on the surface of the vesicles has a significant influence on the process due to interactions between AOT and PANI-ES (see above). AOT GUVs were first prepared in 20 mM NaH 2 PO 4 solution (pH = 4.3) containing 4.0 mM aniline and 0.92 μM HRP. Afterwards, a solution of 2.0 M H 2 O 2 and 20 mM AOT micelles (prepared in pure water) was then micro-injected to a selected target GUV and its size change was analysed (entry #1 in Table 1). It should be noted that the micro-injection of the reactant solutions does not affect the reaction condition significantly (Supplementary Note 3 and Supplementary Fig. 3). To estimate vesicle growth quantitatively, the distance between the target GUV and the tip of the micropipette was fixed by using a double micro-injection technique (Supplementary Fig. 4 and Supplementary Note 4). When this micellar AOT/H 2 O 2 solution was micro-injected, the target GUV maintained its spherical shape during the initial 17 s (induction period) and then started to grow with deformation into a prolate shape (Fig. 2a and Supplementary Movie 1). During the growth period from 17 to 45 s, the prolate vesicle elongated with time due to a growth of the membrane area caused by the uptake of AOT molecules from the external solution. As seen in Fig. 2a, the image of the vesicle suspension became dark with time, which was due to the formation of polyaniline products. Since the reaction takes place on a growing GUV under micro-injection, it is difficult to analyse the polymer product directly. Instead, the PANI-ES synthesized on AOT GUVs with feeding AOT micelles in bulk solution was characterised using UV/vis/NIR absorption spectroscopy (Supplementary Note 5 and Supplementary Fig. 5). The PANI-ES was identical to that obtained in the previous studies. The synthesis of PANI-ES on the target AOT GUV in the centre of the image shown in Fig. 2a "encouraged", i.e., promoted, the growth of the vesicle. Time dependent changes of the vesicle surface area, A(t), were estimated by approximating the vesicle shape with an axisymmetric prolate shape using the Surface Evolver software package 6,35 (Supplementary Fig. 6). The obtained vesicle area is plotted as a function of time (Fig. 3a, entry #1), whereby the vesicle surface area is normalised by the surface area of the initial spherical vesicle, A(0); the error bars represent standard deviations obtained from three different experiments. In the growth stage (between 17 and 45 s), the vesicle area increased linearly with the growth rate of d(A(t)/A (0))/dt = 0.012 s −1 . The analysis of the vesicle growth kinetics based on the elasticity theory of vesicles (Supplementary Note 6 and Supplementary Fig. 7) is shown in Supplementary Note 7 and Supplementary Fig. 8. For better understanding the observed vesicle growth, several control experiments were performed. Importance of simultaneous "template" polymerisation for vesicle growth: AOT micelles were micro-injected to AOT GUVs under conditions where no polymerisation took place (Table 1). When aniline and HRP were not present in the external solution (entry #2), the growth of AOT GUVs induced by the microinjection of AOT micelles and H 2 O 2 was remarkably suppressed (growth rate of 0.0013 s −1 in the growth stage between 17 and 100 s) (Figs. 2b and 3a, and Supplementary Movie 2). For these experimental conditions, the observed slight AOT GUV growth indicates that some of the AOT molecules which were supplied through addition of the AOT micelle solution incorporated into the preformed GUV membrane, although the extent of uptake was comparatively low. When the preformed GUV suspension did not contain aniline (entry #3) or HRP (entry #4), microinjection of AOT micelles and H 2 O 2 did not result in a growth of the AOT GUV (Fig. 3a). In addition, when only AOT micelles (without H 2 O 2 ) were micro-injected to AOT GUVs in the presence of aniline and HRP (entry #5), there was no AOT GUV growth (Fig. 3a and Supplementary Movie 3). Thus, the synthesis 1. Length of the scale bars: 20 μm. a Entry #1 of Table 1: A 3.0 mM AOT GUV suspension consisting of 4.0 mM aniline and 0.92 μM HRP was first placed into the sample chamber, followed by microinjection of two different solutions from opposite sites to a selected GUV. With this double micro-injection (Supplementary Fig. 4), the target GUV can be kept in place. The upper image shows the double micro-injection setup with the target GUV in the centre and the two micropipettes on the left and right hand side, respectively. The dark particle with the white halo is an AOT aggregate which was not completely dispersed. of products consisting of PANI-ES units on the AOT GUV surface remarkably accelerates the growth of AOT GUVs in these AOT micelle feeding experiments. The importance of interactions between PANI-ES and AOT vesicle is also demonstrated by additional control experiments (Supplementary Fig. 9, Supplementary Note 8 and Supplementary Movies 4 and 5). Selectivity of vesicle growth coupled with the synthesis of PANI-ES: The enzymatic polymerisation of aniline in the presence of vesicles produces a mixture of PANI-ES products which consist of aniline repeating units that differ in the way the aniline units are connected (constitutional isomerism) and in their oxidation and protonation state. The PANI product mixture obtained depends on the chemical structure of the vesicleforming amphiphiles 23,36 , as is evident from the results shown in 3). The observed growth rates of the GUVs were normalised by that of the growth observed for the AOT GUV/AOT SUV system (Table 2). GUVs from neutral or anionic phospholipids did not show any significant growth. On the other hand, GUVs composed of amphiphiles with a sulfonate head group showed growth (addition of AOT SUVs to AOT GUVs or addition of SDBS micelles to SDBS/DA GUVs). Thus, a coupling between template polymerisation and vesicle growth is observed only for those vesicle systems, which yield products rich in PANI-ES (Fig. 1c) and to which amphiphiles with a sulfonate head group are added. It is worth noting that growth of AOT GUVs was also induced by injection of SDBS micelles (Supplementary Movie 6) and growth of SDBS/DA vesicles was also induced by addition of AOT SUVs (Fig. 3b). Thus, the sulfonate head group is responsible for the observed growth of the vesicles which consist of surface-localised PANI-ES products. This indicates that specific interactions between the polymeric products and the amphiphile head group play an important role in the vesicle growth process. PANI-ES localised on the vesicle surface 21 interacts with externally added amphiphiles bearing a sulfonate head group which results in a selective incorporation into the vesicle membrane composed of the sulfonate amphiphiles, which leads to a growth of the vesicle. We therefore call the PANI-ES which is produced by the template polymerisation "information polymer". Putting all experimental data presented in this section together, there is one key observation: AOT (or SDBS/DA) GUVs promote the synthesis of PANI-ES, and PANI-ES promotes the growth of the AOT (or SDBS/DA) GUVs. This mutual promotion mechanism is an essential feature of minimal cells 27,28 , see the kinetic model description in the "Discussion" section. Reproduction of binary AOT/cholesterol (9:1) GUVs. By feeding spherical AOT GUVs with AOT molecules from the external medium and by coupling this feeding process with a GUV surface-localised enzymatic synthesis of PANI-ES, the AOT GUVs grow to a prolate shape, but they never show any vesicle reproduction, i.e., they never show a deformation to the limiting shape consisting of a pair of spheres connected by a narrow neck and a breaking of the neck of the limiting shape. Thus, the shape and topology of the growing vesicles have to be controlled by other means. The deformation and division of vesicles is well described by the elastic theory of membranes 38,39 . Most importantly, the membrane elasticity model shows that vesicle division is hard to observe in one-component vesicles 40,41 , which agrees well with experimental observations 42 . Therefore, a second amphiphile having a negative molecular spontaneous curvature (amphiphile with small polar head group and bulky hydrocarbon part) has to be introduced 5,6 ; whereby the coupling between the membrane curvatures and the local amphiphile composition is responsible for the deformation and division of the vesicle 6,7,43 (Supplementary Note 7 and Supplementary Fig. 8). Cholesterol with its small polar head group (-OH) and bulky hydrocarbon part, i.e., with its negative spontaneous curvature property, may be suited as the second amphiphile. To test this, we prepared binary GUVs composed of AOT and cholesterol (at a molar ratio of 9:1, 2.7 mM AOT, 0.3 mM cholesterol) in 20 mM NaH 2 PO 4 solution (pH = 4.3) containing 4.0 mM aniline and 0.92 μM HRP, and micro-injected to individual binary GUVs a micellar solution consisting of 20 mM AOT and 2.0 M H 2 O 2 (prepared in pure water). Immediately after injection, the targeted GUVs started to deform to the limiting shape and then spontaneously divided into two daughter GUVs in about 10 s (Fig. 4 and Supplementary Movies 7 and 8). Although we have no experimental evidence about the compositions of the two daughter vesicles, it is likely that they were the same or very similar. This conclusion is based on a consideration of the characteristic diffusion time, τ, of AOT for a spherical vesicle. It is given by τ = R 2 /(4D L ), where R is the radius of the vesicle and D L (=2.7 × 10 −11 m 2 /s) is the lateral diffusion coefficient of AOT in the lamellar membrane 44 . For a giant vesicle with R = 10 μm τ is ~0.9 s, which is much smaller than the observed time range for vesicle division (~10 s). Thus, the two daughter GUVs probably have almost the same AOT/cholesterol membrane composition, although AOT molecules are supplied locally by micro-injection. The vesicle reproduction was confirmed by carrying out more than 10 identical experiments in which the binary AOT/ cholesterol GUVs always showed division, with occasional continuation of the division of the daughter vesicles to a third generation (Supplementary Movie 9). We analysed the vesicle division processes (n = 4) induced by the simultaneous microinjection of AOT micelles and H 2 O 2 . When the mother GUV with initial surface area A(0) and volume V(0) showed growth and division, the surface area and volume of the GUV increased to (1.25 ± 0.05)•A(0) and (1.05 ± 0.05)•V(0), respectively. The size ratio of daughter to mother vesicles is determined by the balance between the uptake rate of AOT molecules (reduced volume) and the flip-flop rate (normalised preferred area difference) as shown in Supplementary Fig. 7. In addition, the synthesis of PANI-ES was confirmed by recording the UV/vis/NIR absorption spectrum (Supplementary Fig. 10). Thus, we succeeded with the reproduction of AOT-based vesicles due to (i) the presence of a second amphiphile (cholesterol), (ii) a feeding of the vesicles with AOT molecules and (iii) a coupling of the feeding process with the synthesis of an "information polymer" (PANI-ES). A suggested molecular interpretation which is consistent with the microscopic observations is the following: AOT molecules present in the external solution bind to the formed PANI-ES through electrostatic interactions or hydrogen bonding. This binding decreases the hydrophilicity of the AOT molecules, which promotes incorporation of the bound AOT molecules into the outer monolayer of the AOT/cholesterol GUVs. The incorporated AOT molecules increase the area of the outer monolayer. Flipflop motions of the AOT molecules coupled with the negative molecular spontaneous curvature of cholesterol relaxes the initially spherical vesicle shape to a prolate shape and then to the limiting shape 6 (Supplementary Fig. 8). In the limiting shape, cholesterol molecules are excluded from the neck due to the coupling between the molecular shape and the membrane Gaussian curvature, which destabilises the neck and causes vesicle division 7 . By further injection of AOT micelles and H 2 O 2 , the offspring GUVs produced a cascade of multiple small vesicles, i.e., no recursive vesicle reproduction (Supplementary Movie 10). In this type of experiment, H 2 O 2 together with pure AOT micelles were micro-injected to binary AOT/cholesterol (9:1) GUVs, since it is difficult to incorporate mixed cholesterol-containing AOT micelles. Therefore, the concentration of cholesterol in the binary GUVs decreased with time, which appears to be responsible for the observed division mode transition (a cascade of small vesicles formation). In fact, when AOT micelles and H 2 O 2 were microinjected to binary AOT/cholesterol GUVs consisting of 5 instead of 10 mol% cholesterol, the target GUVs produced a similar cascade of small daughter vesicles from the beginning (Supplementary Movie 11). Kinetic model of vesicle growth coupled with polymerisation. In theoretical descriptions of minimal cells, such as the chemoton model 45 , a mutually catalytic reaction network is encapsulated in a compartment (vesicle), whereby for sustaining the compartment-confined reaction network, ingredients and waste have to be transported selectively through the compartment shell (vesicle membrane). To establish such traffic systems, the compartment boundary should be equipped with sophisticated "transport units", which is an obstacle for the synthesis of minimal cells 46,47 . In our vesicle model of a simple minimal cell system, the coupling of membrane growth and template polymerisation takes place at the outer surface of the vesicles, which circumvents transmembrane transport and makes the traffic of nutrients and waste easily possible 48,49 . Our model reaction system, however, does not provide a sustainable vesicle reproduction, as shown in Supplementary Movie 10. In the following, we outline how to develop our existing vesicle reaction system to a sustainable one on the basis of a kinetic model, which we call "template" polymerisation on vesicle (TPV) model (Supplementary Note 9). This TPV model is based on three stoichiometric equations describing three kinetic processes, (i) synthesis of the first PANI-ES segment, (ii) chain propagation, and (iii) incorporation of AOT molecules from the external micellar AOT solution into the membrane, as shown in Fig. 5 21 . The mutual promotion mechanism is captured by this TPV model, independent from the actual length of the formed products consisting of PANI-ES units. The kinetic equations based on this TPV model show synchronised sustainable vesicle growth and "information polymer" production, where the number of polymer segments on the vesicle surface is doubled when the vesicle area doubles (Supplementary Fig. 11a). In this model, the vesicle membrane exhibits exponential growth as shown in Supplementary Fig. 11b. In our vesicle growth experiments, however, the observed membrane growth profile shows linear growth after an induction period (Fig. 3a). This experimentally observed growth profile is explained by the TPV model, when we take into account the experimental constraints, such as polymerisation rate constants, geometry of micro-injection and surface coverage by the polymer product (Supplementary Note 10 and Supplementary Fig. 12). In addition, we experimentally confirmed the synchronisation between vesicle growth rate and polymer segment production rate, as predicted by the TPV model (Supplementary Note 11 and Supplementary Fig. 13). Another important prediction is that the vesicle growth rate is exponentially accelerated by the polymerisation (Supplementary Fig. 11b inset). This great enhancement indicates advantage of PANI-ES as "information molecule" for promoting vesicle growth. To achieve reproduction of vesicles, a second amphiphile having a negative spontaneous curvature, such as cholesterol, must participate in the vesicle reproduction process (see the discussion above and Fig. 4). For the binary AOT/cholesterol GUV system, however, cholesterol is not incorporated into the vesicle membrane because cholesterol is not present in the micelles which were added to the GUVs. Moreover, there are no specific interactions between cholesterol and PANI-ES segments. In this case, the concentration of cholesterol in the vesicle membrane monotonically decreases towards zero, i.e., the capability for vesicle division is lost with time (Supplementary Movie 10). The TPV model predicts that a synchronisation between vesicle growth rate and polymer segment production rate would be maintained if the second amphiphile and the information polymer would have mutual catalytic properties like AOT and PANI-ES (Supplementary Fig. 11c). Thus, a key point to attain sustainable vesicle reproduction is to replace cholesterol as second amphiphile by an amphiphile which is directly linked to the enzymatic synthesis of PANI-ES, whereby this second amphiphile must modify the Gaussian curvature modulus through its molecular shape. We have shown that the enzymatic polymerisation of aniline in the presence of three different types of vesicles composed of either zwitterionic phospholipids (DOPC), anionic phospholipids (DOPA), or synthetic anionic amphiphiles having a sulfonate head group (AOT or SDBS/DA (1:1)), yields products which have different spectroscopic properties, and with this different chemical constitutions and/or oxidation and protonation states, depending on the type of amphiphiles (Fig. 1). Among them, the AOT/PANI-ES pair behaves according to a mutual promotion system, resulting in a sustainable synchronized AOT vesicle growth and the production of PANI-ES. If we regard the PANI-ES/AOT pair as a "species", and if another polymer/amphiphile pair would also behave in the same way as second "species", then our TPV model predicts that there will be a competition between two "species". This would represent a chemical compartment system which is characterised by a natural selection feature, whereby the selection would depend on the rate constants for the species-specific polymerisation and vesicle membrane growth (fitness) (Supplementary Note 12 and Supplementary Fig. 14). Therefore, the "information polymer"/amphiphile system investigated-although being composed of molecules which are not considered prebiotic at all-has the potential for the development of simple chemical systems which have properties that are characteristic of biological systems undergoing Darwinian evolution. ## Discussion We have developed a model for a minimal cell system which consists of the reproduction of vesicles coupled with a "template"assisted enzymatic polymerisation occurring on the surface of the vesicles. GUVs composed of AOT and cholesterol molecules were first prepared and products consisting of PANI-ES units are obtained from aniline on the vesicle surface, where PANI-ES "interacts with" AOT molecules added to the external solution, and incorporates them into the vesicle membrane, which leads to a growth of the vesicles. Thus, in this model system, PANI-ES plays a role as "information polymer". An important property of (3) seg represents a segment of PANI-ES (Supplementary Fig. 2) the system is that the AOT vesicles promote the synthesis of PANI-ES and that PANI-ES promotes the growth of the AOT vesicles. Conceptually, such mutual promotion property can be seen as an essential feature of minimal cells. In addition to AOT, the second amphiphile, cholesterol, plays an important role. It has a different molecular shape than AOT and induces a negative spontaneous membrane curvature. With this, cholesterol is responsible for vesicle division to occur once the vesicle is grown, through a coupling between vesicle membrane curvatures (mean and Gaussian curvature) and local molecular composition. A general scheme of the system investigated is summarized in Fig. 6. Since PANI-ES is not coupled with the uptake of cholesterol by the vesicle membrane, the vesicle reproduction is limited and will only last for a few generations. On the basis of our experimental observations we developed a kinetic model of the particular vesicle-assisted polymerisation (called "TPV model"). This TPV model predicts that the reproduction of vesicles becomes recursive when the second amphiphile incorporates into the vesicles through interactions with the formed "information polymer". In addition, a chemical version of the Darwinian evolution mechanism is expected to emerge when a second "information polymer" formed on the surface of the vesicles constituted by the second amphiphile couples with the growth of the membrane formed by the second amphiphile (emergence of new species). Thus, the reproduction of vesicles coupled with the vesicle-assisted polymerisation system contains (i) a kind of metabolism which extracts usable molecules from the environment, (ii) reproduction, i.e., growth and division of vesicles, and (iii) chemical evolvability, i.e., a "struggle" for the existence of either of two "amphiphile/information polymer" pairs (species). In other words, PANI-ES and AOT vesicles play roles as information tape and general constructor, respectively, if analogy is made to the self-reproducing automaton proposed by von Neumann 50 . Of course this model system is very different from any known biological system and further investigations are necessary. However, our work may be of interest from a conceptual point of view, not only for researchers dealing with prebiotic compartmentalisation, i.e., protocells, or minimal cells, but also for those investigating fundamental aspects of soft, dynamic compartment systems. ## Methods Amphiphiles used for vesicle and micelle preparation. AOT (sodium bis-(2ethylhexyl) sulfosuccinate, purity > 99%, Catalogue No. 86139) and SDBS (sodium dodecylbenzenesulfonate, hard type, >95%, D0990) were purchased from Sigma-Aldrich Japan (Tokyo, Japan) and Tokyo Chemical Industry (Tokyo, Japan), respectively. DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine, >99%, 850375), DOPA (1,2-dioleoyl-sn-glycero-3-phosphate sodium salt, >99%, 840875), DOPG (1,2-dioleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) sodium salt, >99%, 840475), DOPS (1,2-dioleoyl-sn-glycero-3-phospho-L-serine, >99%, 840035), porcine brain sulfatides (as ammonium salt, >99%, 131305) and cholesterol (ovine wool, >98%, 700000) were purchased from Avanti Polar Lipids, Inc. (AL, USA), and DA (decanoic acid, >99%, 041-23256) was from Wako Pure Chemical Industries (Osaka, Japan). The amphiphiles were used without further purification and dissolved in chloroform at 10 mM and stored at −20 °C as stock solutions. Reagents for the polymerisation of aniline. Aniline (>99%), hydrogen peroxide (guaranteed grade, 30% in water, =9.8 M), sodium dihydrogenphosphate (NaH 2 PO 4 ) dihydrate (>99.0%), phosphoric acid (H 3 PO 4 , guaranteed grade, 85% in water) and chloroform (>99%) were purchased from Wako Pure Chemical Industries (Osaka, Japan). Horseradish peroxidase isoenzyme C (HRP, Grade I, 285 U/mg, RZ > 3, Lot No. 73120) was from Toyobo Enzymes (Osaka, Japan). The peroxidase concentration was determined spectrophotometrically by using ε 403 = 1.02 × 10 5 M −1 cm −1 as molar absorbance 51 . Ultrapure water purified with a Direct-Q 3 UV apparatus (Millipore, USA) was used for the preparation of all aqueous solutions and suspensions. Preparation of GUV suspensions. Most of the investigations presented were carried out with GUVs with a size range of 1-200 µm or even larger. They allow a direct quantitative analysis of changes in size and shape of individual vesicles by optical microscopy. GUVs composed of AOT were prepared by using the gentle hydration method 52 . First, AOT (17.8 mg) was dissolved in 1 mL chloroform in a glass vial. Then, a thin AOT film was prepared on the inner surface of the vial by using a nitrogen gas stream under rotation of the vial by hand. To remove the organic solvent completely, the AOT film was put under vacuum overnight, while the vial was kept wrapped with aluminium foil. The pre-warmed, dried AOT film was hydrated with 2.0 mL of a NaH 2 PO 4 solution (20 mM, pH = 4.3) at 60 °C for 1 h, which resulted in the formation of GUVs with radii of 5-30 μm. The obtained 20 mM AOT GUV suspension was stored at 25 °C. GUVs composed of phospholipids (DOPC, DOPA, DOPG, or DOPS) or sulfatide were prepared by using the same protocol. The final lipid concentrations in these suspensions were 5 mM. Binary GUVs composed of either AOT and cholesterol (9:1 or 95:5, molar ratio, 5 mM in total), or SDBS and DA (1:1 molar ratio, 20 mM in total) were prepared in the same way as described above from the corresponding chloroform solutions containing the appropriate mixture of the two amphiphiles. Please note that stable AOT/cholesterol GUVs could not be prepared for cholesterol contents above 10 mol%. Preparation of LUV suspensions. To characterise the PANI products which were obtained in the presence of vesicles by UV/vis/NIR absorption spectroscopy, 20 mM AOT, 20 mM SDBS/DA (1:1), 5 mM phospholipids (DOPC, DOPA, DOPG, and DOPS) or 5 mM sulfatides, LUVs with a size range of ~100-1000 nm were prepared by using the freezing-thawing extrusion method in a similar way as described before 21 . First, GUV suspensions were prepared as described above. Then, the obtained GUV suspensions were frozen in liquid nitrogen and thawed in a water bath heated to 60 °C. This procedure was repeated ten times. After these freezing-thawing cycles, the phospholipid or sulfatide suspensions were extruded 21 times through a 100 nm pore size Nucleopore polycarbonate membrane, using a Mini Extruder set from Avanti Polar Lipids, Inc., USA. For the preparation of AOT and SDBS/DA (1:1) LUVs, the GUV suspensions were extruded ten times through a 200 nm pore size Nucleopore polycarbonate membrane, and then, ten times through a 100 nm pore size membrane, using a LIPEX TM Extruder from Northern Lipids Inc., Canada. The obtained LUVs were characterised previously by dynamic light scattering (DLS) and cryo transmission electron microscopy 21 . The LUVs had an average hydrodynamic diameter of about 80 nm with a polydispersity of about 0.1. The cvc value for AOT in 20 mM NaH 2 PO 4 solution (pH = 4.3) was determined by turbidity measurements (1.5 mM), as described before for AOT in 100 mM NaH 2 PO 4 solution (pH = 4.3) (0.4 mM) 22 . Preparation of SUV suspensions and micellar solutions. SUV (sonicated, small unilamellar vesicles with a size range of ~20-100 nm) suspensions for the microinjection experiments were obtained by sonication of the prepared GUV suspensions (see above) for 5 min at room temperature using a Branson Sonifier model 150 (Emerson, USA). AOT and SDBS micellar solutions were also prepared for the use in the micro-injection experiments by simply dissolving the two amphiphiles in deionized water. The SUV suspensions and the micellar solutions were mixed with an aqueous hydrogen peroxide solution just before use and then pressed through a 0. Enzymatic polymerisation of aniline on GUVs. Changes in the size and morphology of GUVs in response to the enzymatic polymerisation of aniline coupled with feeding of the amphiphiles were analysed by phase contrast light microscopy (see below). The experiments were performed in the microscope sample chamber, which is a hole in a silicone rubber sheet placed onto a glass slide. The hole had a diameter of 12 mm and a depth of 1 mm. The GUV suspension mixed with the polymerisation components, except H 2 O 2 , was carefully transferred at room temperature from the glass vial into the sample chamber. The initial concentrations of the different components of the reaction mixture were the same as the ones for the polymerisation experiments with LUVs: 20 mM NaH 2 PO 4 solution (pH = 4.3) containing 4.0 mM aniline and 0.92 μM HRP. The polymerisation was triggered by micro-injecting a 2.0 M H 2 O 2 solution containing either SUVs (20 mM DOPC, DOPA or AOT), or micelles (20 mM AOT or 100 mM SDBS). The micro-pipette used for the micro-injection was a Femtotip II with an inner diameter of 0.5 ± 0.2 μm (Eppendorf, Germany). The position of the micro-pipette was controlled using a hydraulic micro-manipulator MMO-202ND (Narishige, Japan), and the microinjection was performed using a Femtojet system (Eppendorf, Germany). To minimize GUV drifts caused by the injection flow, we adopted a double micropipette injection technique with a symmetric configuration, where the target GUV was located at the bottom of the chamber (Supplementary Fig. 4). The two injection flows trap the GUV at a fixed point, which makes it possible to measure the growth of the GUV quantitatively with high reproducibility. The distance from the tip of the pipettes to the target GUV was about 100 μm and most of the microinjection experiments were performed with an injection pressure of 180 hPa. The volume injected from the pipette was 0.14 nL/s. In additional experiments, single micro-injections were also performed, and the growth of the vesicles coupled to the aniline polymerisation was analysed. Microscope observation of GUVs. The GUV growth in response to the microinjection was followed by using an Axio Vert. A1 FL-LED inverted fluorescence microscope in phase contrast mode (Carl Zeiss, Germany) with a 40× objective (LD A-Plan 40× N.A. = 0.55) and an Axiocam 506mono (Carl Zeiss, Germany) for recording the images. To estimate the vesicle surface area (A) and volume (V), a 3D image of the GUV was reconstructed from the 2D microscope image by using the Surface Evolver software package 6,35 . UV/vis/NIR absorption measurements. Absorption measurements in the UV/ vis/NIR region of the spectrum were recorded with a V-730 spectrometer (JASCO, Japan) at 25 °C, using quartz cuvettes S15-UV-1 (from GL Sciences Inc. Japan) with path lengths of 1 mm. DLS measurements. Dynamic light scattering measurements were performed using an ALV-5000 goniometer system (ALV, Langen, Germany) with an ALV-6000 multi-bit multi-tau correlator and a diode-pumped laser Verdi V-2 (Coherent Inc. Santa Clara, USA) operating with vertically polarised light at λ = 532 nm. The sample was filtered before the measurement with a 0.22 μm membrane filter (from Millipore).

chemsum

{"title": "Reproduction of vesicles coupled with a vesicle surface-confined enzymatic polymerisation", "journal": "Nature Communications Chemistry"}

unravelling_molecular_interactions_in_uracil_clusters_by_xps_measurements_assisted_by_ab_initio_and_

## Abstract: The C, N and O 1s XPS spectra of uracil clusters in the gas phase have been measured. A new bottom-up approach, which relies on computational simulations starting from the crystallographic structure of uracil, has been adopted to interpret the measured spectra. this approach sheds light on the different molecular interactions (H-bond, π-stacking, dispersion interactions) at work in the cluster and provides a good understanding of the observed XPS chemical shifts with respect to the isolated molecule in terms of intramolecular and intermolecular screening occurring after the corehole ionization. the proposed bottom-up approach, reasonably expensive in terms of computational resources, has been validated by finite-temperature molecular dynamics simulations of clusters composed of up to fifty molecules. H-bonds and van der Waals interactions are ubiquitous in nature and influence the structure, stability, dynamics, and function of molecules and materials throughout chemistry, biology, physics, and material science. Molecular clusters are weakly bonded systems 1 with properties different from those of a single molecule or a condensed molecular film. However, the study of the weak interactions in gas-phase clusters of increasing size can give information on structures and mechanisms at work in both the liquid and condensed phases. For instance, gas-phase nucleobase pairs may also follow the Watson-Crick pairing mechanisms 2 . X-ray photoemission spectroscopy (XPS) provides detailed information on the environment of an emitting atom in a sample 3 . In a molecular cluster, the weak, non-covalent interactions modulate the position of nominally equivalent XPS peaks. The measurement and interpretation of such fine variations have provided a better understanding of the chemical equilibrium in ionized aggregates, as well as of their stability and reactivity 4,5 . In the present work the C, N and O 1s XPS spectra of uracil clusters have been measured with synchrotron radiation. The spectra have been interpreted using a simplified bottom-up procedure to describe the interaction patterns within the cluster and a systematic calculation of the ionization energy of all atoms based on ab initio simulations. Different studies on core spectroscopy and fragmentation processes of the isolated uracil molecule have been previously reported 6,7 , and their results provide the reference for the present work. ## Results The XPS spectra of uracil clusters, measured at an oven temperature of 183 • C, are reported in Fig. 1a-c together with those previously measured for the isolated molecule 6 . The uracil ring is formed by two -aza-nitrogen atoms in positions 1 and 3 and four carbon atoms. The C2 and C4 atoms form carbonyl groups with O8 and O7 atoms, respectively. In the isolated molecule (Fig. 1a to the different chemical connections, with the two at higher binding energy (BE) slightly overlapping. The C2 atom has the highest BE because the larger electronegativity of the nearby O and N atoms induces the strongest charge depletion. Vice-versa the peak at the lowest BE has been assigned 6 to C5, situated in the middle of the conjugated C4-C5-C6 moiety. The two central peaks have been assigned 6 to C4 and C6, whose shifts depend on the neighboring O and N or only N atom, respectively. All of these assignments have been confirmed by present ab initio calculations (see Table SI in Supplementary Information). Each one of the N and O 1s spectra (Fig. 1b,c top), is expected to contain two non-equivalent contributions, with predicted separation of 0.43 eV and 0.37 eV, respectively. In the cluster case, the whole C 1s spectrum is broadened and shifted by about 0.9 eV to lower BE, with the C4 and C2 features clearly overlapping (Fig. 1a bottom). Also the N and O 1s features (Fig. 1b,c bottom) are broadened and shifted to lower BE by a similar amount, within the experimental uncertainty. ## Discussion To understand how the different interactions in the cluster affect the observed shift a simplified bottom-up theoretical approach has been developed. This approach is motivated by the un-sustainable computational cost of the calculation of XPS lines in large (>10 molecules) clusters, performed on several configurations sampled along equilibrated molecular dynamics trajectories. A posteriori we will demonstrate that such approach well reproduces all the short-range structural motifs and intermolecular patterns of realistic clusters. In detail, several different cluster sizes (from dimer to dodecamer) have been cut out from the uracil crystal structure 8,9 in order to sample all kinds of local connectivity present in the periodic structure, and fully optimized in gas phase, with the BE of each atom calculated, as reported in Table SI of Supplementary Information. As for the connectivity, three different patterns of intermolecular interactions, leading to three possible neutral dimer configurations shown in the top panel of Fig. 1, have been identified: a bidirectional and symmetric H-bond (formation energy per molecule 0.30 eV, "dimer1" in Table SII of Supplementary Information), a monodirectional and asymmetric H-bond (0.30 eV, "dimer2") and a stacked dimer (0.06 eV, "dimer3"). The formation of large clusters is driven by the anisotropic distribution of H-bond donor and acceptor sites, which can be also modulated by weaker dispersion forces between the π-conjugate charge distributions. Calculated average values of the C, N and O1s BEs are shown in Fig. 1a-c, respectively. An overall very good agreement, with slight discrepancies in the case of O1s, is already found for a cluster of 12 molecules, which is a sensible limit for this kind of calculation. The two different dimer configurations having the same formation energy per molecule, have been simulated first. In both dimers and for all atoms, the BEs shift towards lower values, with a maximum shift of about 1.1 eV for N3 ("dimer1"). This immediately remarks a characteristic and expected property of NH groups H-bonded to electron-rich C=O groups, where the core-hole in the N atom is very efficiently screened by the carbonyl group. This is not the case for N1, which experiences instead a slight increase with respect to the monomer. As for "dimer3" the absence of H-bonds leads to BE shifts ≤ 50 meV (see Table SI of Supplementary). Knowing that the H bond can be partially seen as a resonance of N-H⋯O=C with N − ⋯H-O + -C, we have also checked if it did not lead to an increase of charge on N3 already in the neutral dimer. The results of such investigation, described in detail in the Supplementary Information S2, show that the formation of C=O⋯H-N "prepares" the system, inducing the polarization of the involved atoms without displacing significant amounts of charge between atomic sites. The formation of the core hole takes advantage of the prepared path to provide a more efficient screening of the core holes with respect to the isolated uracil molecule, with the strong polarization of the bond towards the O side playing a key role. Tetramer and hexamer structures have been obtained by piling up two or three "dimer1" structures, respectively. In this arrangement the noncovalent, attractive π-stacking interactions between the aromatic rings introduce additional collective shifts towards lower BE with respect to the dimer, with an average of 140 meV for the tetramer and a further 100 meV for the hexamer. Such shifts are due to the isotropic participation of neighboring molecules to the screening of the core hole. A tetramer in planar configuration, where the molecules are only connected through H-bonds, has been considered too. This structure implies a higher formation energy per molecule (0.48 eV) with respect to the stacked configuration (0.37 eV). Such a stable planar motif, likely ubiquitous in realistic aggregates, will be discussed below for the interpretation of the intermolecular screening. Finally, the additional shift in the dodecamer with respect to the hexamer varies between 400 meV (C6) and 70 meV (O7), with an average of 220 meV. All calculated C1s BEs are in excellent agreement with the experiments. The shifts in this case depend on the size of the cluster, but not on the connectivity between molecules, because C 1s atoms are not involved in the formation of H-bonds. We note that our "crystallographic structure" model likely tends to overestimate the energy separation between N3 (always involved in bidirectional H-bonds) and N1 (partially involved in monodirectional H-bonds) values and to underestimate the one between O7 and O8. On the basis of the overall satisfactory agreement between calculated and measured BEs, and considering that all the intermolecular and side interactions among the three stacks of uracil molecules (Fig. 2) are active, the dodecamer can be considered as a "representative" model of our cluster distribution. To shed light on the intraand intermolecular contributions to the observed shifts, calculated charge difference-density maps (Fig. 2), which quantify the displacement of the valence electronic charge density from blue to red zones of the cluster to screen the core hole, have been used. As an example the ionization of the C2 atom in each one of the four molecules (A to D) of the tetramer in planar configuration has been chosen, due to the abundance of the different intermolecular configurations and a relatively simple graphical representation of this cluster. Two kinds of map have been calculated: in the first case (Fig. 2a) the charge density of the whole cluster containing the core-ionized molecule is subtracted from the charge density of the neutral system. One can clearly see that: (i) the major contribution to the screening of the core hole comes from an intramolecular contraction of the charge density (blue arrow); (ii) the surrounding molecules participate to the screening with an isotropic polarization (magenta arrows) around the hole, with the first neighbors (full arrows) more involved than the second neighboring molecules (dashed arrows). This type of plot clearly shows a cooperative charge redistribution in the cluster, but does not explain the reason why the four C2 atoms in the cluster have different BEs, ranging from 294.35 to 295.37 eV for A and D molecules, respectively. This effect is attributed to intermolecular interactions. To illustrate this and to filter out the intramolecular contribution to the screening, the map in Fig. 2b shows the difference between the electron density of the core-ionized tetramer (in molecule A) and the electron density of the core-ionized molecule (A), alone, plus the charge density of the other three molecules (B, C, D) calculated all together in their neutral ground state. In this way the intramolecular contribution to the screening of molecule A as well as the charge displacement due to the intermolecular interactions between molecules B, C and D in their neutral ground state are removed and the map shows the charge displacement induced in molecules B, C and D by the introduction of the core www.nature.com/scientificreports/ hole in A. The map clearly shows that the screening charge is effectively moved from the peripheral regions of molecules B, C and D. The blue arrows, labeled 1 in Fig. 2b , identify the peripheral regions of the system where the non-ionized molecules suffer a charge depletion, when the core hole is "turned on". The charge moves towards the regions adjacent to the ionized molecule in an anisotropic manner, mainly populating the regions of diffuse charge around the O atoms of carbonyl groups (red arrows labeled 1). The two carbonyl groups of molecule A (red arrows labeled 2) are directly involved in the accumulation of the charge density to screen more effectively the core hole. The overall result is a red shift of about 1 eV with respect to the isolated molecule. The similar, but slightly more complex evaluation of the intermolecular screening in the case of molecules B, C and D and the extension to the dodecamer are reported in Supplementary Information S3. These maps graphically demonstrate how an atom of the same species in a molecule located in different positions in the cluster is affected by different forces and interactions, and therefore due to the different screening has a different BE. Finally, we have used finite-temperature molecular dynamics simulations to investigate the connections that can be found in a cluster formed in gas phase by randomly assembling n molecules. The parameter used to evaluate the connectivity in the cluster is the distribution of (N(H)⋯O) intermolecular distances, whose first maximum around 2.8 (Fig. 3) describes the H bond. After a benchmark calculation to verify the consistency of the ab initio 10 and tight-binding 11 molecular dynamics simulations in the case of the dodecamer, simulations starting from both the crystallographic structure and a random distribution of molecules have been performed for clusters of 12, 24 and 50 molecules. The results are compared in Fig. 3a. In the simulations starting from the crystallographic structure the distribution of the H-bond connectivity is already well defined in a cluster of 12 molecules and, as shown by the intensity of the peak at 2.8 , reaches its convergence at 24 molecules. In the case of a random distribution a similar situation is observed, with the convergence already reached for a cluster of 12 molecules. In the case of the dodecamer a metadynamics simulation 12 for about 200 ps to identify the global energy minimum, has been also performed, following a computational protocol described in detail elsewhere 12 for the exploration of potential energy surfaces. The most stable structure has been further optimized with the same method (B3LYP) used for the crystallographic structure. The comparison of the connectivity in the crystallographic and randomly oriented clusters (Fig. 3b) shows that the characteristic features of the distribution are clearly present in both simulations, validating the proposed extraction of the cluster structure from the crystallographic one. Then the calculated BEs for the randomly oriented cluster (Fig. 1a-c and Table SI1 in Supplementary Information) are the same of the ones of the crystallographic cluster in the case of C 1s (maximum difference is 50 meV for C6). The splits of the N1s (O1s) lines in randomly oriented clusters reduces (increases) due to the reduction of the site specificity for these atoms, while the disagreement in the absolute value of the O1s BE remains. This latter observation proves that the difference is not due to the definition of the cluster structure. In summary a theoretical modeling of uracil clusters with a bottom-up approach based on the crystallographic structure has been proposed. Molecular dynamics simulations at finite temperature have confirmed that the model used is substantially able to reproduce connections and structures found in realistic gas phase clusters. Hence this approach can be proposed as a computationally sustainable method to study the chemical physical properties of weakly bonded clusters, provided their crystal structure is known. It can be applied to a large variety of systems from the bases of nucleic acids up to large proteins and therefore find a broad use. The modeling of the XPS spectra has allowed to disentangle the contribution of the different intermolecular interactions for increasing cluster size and shown the relevance of the H-bonds in the different positions of the cluster to determine the screening of the original core hole. The ability to display the contribution of intra-and ## Methods experiment. The XPS measurements have been performed at the PLEIADES beamline of Soleil synchrotron. A gas aggregation source was used to produce neutral clusters of uracil molecules with a log-normal size distribution centered about 30-50 molecules, depending on the source parameters 15,16 . The details of the experimental procedure are provided in Supplementary Information S4. ## theory Ab initio simulations of the properties of uracil clusters, including the calculations of core-ionization energies, have been performed by using the Quantum ESPRESSO suite of programs 17 in a plane-wave/pseudopotential framework . In addition, molecular dynamics simulations to determine the structural properties of clusters at finite temperature have been performed using both ab initio 10 and tight-binding 11 methods. As for the metadynamics (MTD), the computational protocol described by Grimme 12 in his introductory article assessing the functionalities of the xTB program, based on the GFN2-xTB Hamiltonian 11, for the exploration of potential energy surfaces has been followed. A complete account of all the employed computational protocols is provided in Supplementary Information S1.

chemsum

{"title": "Unravelling molecular interactions in uracil clusters by XPS measurements assisted by ab initio and tight-binding simulations", "journal": "Scientific Reports - Nature"}

investigating_pyridazine_and_phthalazine_exchange_in_a_series_of_iridium_complexes_in_order_to_defin

## Abstract: The reaction of [Ir(IMes)(COD)Cl], [IMes ¼ 1,3-bis(2,4,6-trimethylphenyl)imidazol-2-ylidene, COD ¼ 1,5cyclooctadiene] with pyridazine (pdz) and phthalazine (phth) results in the formation of [Ir(COD)(IMes)(pdz)]Cl and [Ir(COD)(IMes)(phth)]Cl. These two complexes are shown by nuclear magnetic resonance (NMR) studies to undergo a haptotropic shift which interchanges pairs of protons within the bound ligands. When these complexes are exposed to hydrogen, they react to form [Ir(H) 2 (COD)(IMes)(pdz)]Cl and [Ir(H) 2 (COD)(IMes)(phth)]Cl, respectively, which ultimately convert to [Ir(H) 2 (IMes)(pdz) 3 ]Cl and [Ir(H) 2 (IMes)(phth) 3 ]Cl, as the COD is hydrogenated to form cyclooctane. These two dihydride complexes are shown, by NMR, to undergo both full N-heterocycle dissociation and a haptotropic shift, the rates of which are affected by both steric interactions and free ligand pK a values. The use of these complexes as catalysts in the transfer of polarisation from para-hydrogen to pyridazine and phthalazine via signal amplification by reversible exchange (SABRE) is explored. The possible future use of drugs which contain pyridazine and phthalazine motifs as in vivo or clinical magnetic resonance imaging probes is demonstrated; a range of NMR and phantom-based MRI measurements are reported. ## Introduction Pyridazine and phthalazine are nitrogen-containing heterocycles with derivatives that have proven to be biologically active as anti-asthmatic, 1 antifungal, 2 anti-HIV-1, 3 anti-cancer 4,5 and anti-inflammatory drugs in addition to acting as treatments for Alzheimer's disease. 9,10 A wide range of pyridazine and phthalazine coordination complexes are known where ligation takes place through one or both of the nitrogen centres. When these ligands bind through a single nitrogen centre, the resulting complexes undergo both dissociative ligand loss and 1,2-metallotropic shifts through which the contiguous nitrogen donor atoms interchange positions. 12 Such shifts were frst described in 1972 for a series of ruthenium porphine complexes that contained an axially-bound methylated pyridazine ligand. 12 Rates for ligand shift and dissociation were obtained by nuclear magnetic resonance spectroscopy via line-shape analysis and subsequently the activation parameters for these processes were determined. The authors suggested that the sizeable, and positive, activation entropies that were observed are consistent with a dissociative pathway, although the precise mechanism for the 1,2-metallotropic shift was not discussed. 12 More recent studies have focused on related octahedral systems where full ligand dissociation was more difficult and the 1,2-metallotropic shift proceeded readily. Examples of such complexes can be found for ruthenium, 13 chromium, 14 tungsten, platinum 11,15,16 and rhenium 17 metals. In the more recent reports, the 1,2-metallotropic shift has been referred to as a 'haptotropic shift' because of the potential to involve a pbonding interaction between the two nitrogen centres during exchange. A theoretical study on an octahedral chromium carbonyl complex suggests that this shift can proceed within the solvent cage that surrounds a 16-electron intermediate. 18 These processes are illustrated in Scheme 1. The tendency for ligand dissociation has been predicted to increase as the metal becomes less electron rich or the ligand trans to the exchanging site becomes more basic. 19 Ligand exchange processes are readily probed by Nuclear Magnetic Resonance (NMR) spectroscopy, 20 despite its low sensitivity. 21 When slow, such exchange pathways are often followed by 1-D and 2-D nuclear Overhauser effect spectroscopy (NOESY). 20 When fast, line-shape analysis allows exchange to be probed. 22,23 It has been shown that magnetisation transfer from para-hydrogen (p-H 2 ) can help to overcome the sensitivity problem associated with NMR. 24,25 This is because, whilst p-H 2 itself has no net spin angular momentum and is NMR silent, when used as a reagent in a reaction, products can be formed that possess non-Boltzmann nuclear spin distributions. The resulting signals that are detected in such NMR experiments can be many thousands of times stronger than normal. Parahydrogen induced polarisation (PHIP) 30 that is created in this way has been shown to enable the study of reaction intermediates that cannot otherwise be detected by NMR spectroscopy because they exist in low concentration. 31 In this traditional approach, the reactants must accept H 2 and thus be both unsaturated and highly reactive. 31 p-H 2 has also been used in a non-hydrogenative manner, 32 where molecules are hyperpolarised through the establishment of J-coupling with p-H 2 , whilst they are simultaneously bound to a metal centre in low feld. This has also been achieved by radio frequency excitation. The non-hydrogenative process requires the substrate and p-H 2 to exchange freely with their ligated forms in order to build-up a concentration of hyperpolarised substrate in solution. This effect has been termed signal amplifcation by reversible exchange (SABRE). 36 Polarisation transfer now proceeds over the short time period that the sample is stored in a small pre-determined magnetic feld. 25 This process has attracted interest because, unlike the traditional PHIP approach, it does not result in any chemical change to the detected species and reflects a novel catalytic process. A growing range of materials, including pyridine, nicotinamide, isoniazid, pyrazinamide and acetonitrile have been polarised in this manner. Here, we examine the substrates pyridazine (pdz) and phthalazine (phth) using the SABRE approach, whilst taking the opportunity to simultaneously explore their coordination chemistry. [Ir(IMes)(COD)Cl] (1) (where IMes ¼ 1,3-bis(mesityl) imidazol-2-ylidene and COD ¼ 1,5-cyclooctadiene) is used as the synthetic precursor in this study. 44 ## Experimental procedures All the experimental procedures associated with this work were carried out under nitrogen using standard Schlenk techniques or in an MBraun Unilab glovebox. The solvents used in the synthetic chemistry were dried using an Innovative Technology anhydrous solvent system, or distilled from an appropriate drying agent under nitrogen. Methanol-d 4 , pyridazine and phthalazine were obtained from Sigma-Aldrich and used as supplied. [Ir(IMes)(COD)Cl] (1) was prepared according to literature methods. 45 ## Instrumentation and procedures All NMR measurements were recorded on Bruker Avance III series 400 MHz or 500 MHz systems. NMR samples were prepared in 5 mm NMR tubes ftted with Young's valves. Samples were degassed prior to p-H 2 (3 bar) addition. Typical procedures for reactions with pyridazine and phthalazine are described. NMR characterisation data was collected using a range of 1 and 2 D methods that include nOe, COSY and HMQC procedures. The slow dynamic processes exhibited by the complexes studied here were monitored by EXSY methods. 51 [Ir(H) 2 (IMes)(pdz) 3 ]Cl (4a). In a typical experiment, 1 (2 mg, 3.12 mmol) and fve equivalents of pyridazine (1.13 mL, 15.6 mmol) were dissolved in methanol-d 4 (0.5 mL) in a 5 mm NMR tube and the solution was degassed. H 2 , at a pressure of 3 bar, was added to the sample prior to it being heated to 348 K for 15 minutes to fully activate it thereby forming 4a in solution. [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b). 1 (2 mg, 3.12 mmol) and fve equivalents of phthalazine (2.03 mg, 15.6 mmol) were dissolved in methanol-d 4 (0.5 mL) in a 5 mm NMR tube and the solution was degassed. H 2 (3 bar) was added, and the sample was heated to 348 K for 15 minutes to form 4b. ## SABRE analysis NMR samples were prepared as above. Arrays of NMR measurements were collected using a range of substrate concentrations, as revealed in the ESI. † 1 H NMR spectra were recorded with either p/4 or p/2 excitation pulses in order to differentiate PHIP and SABRE effects. Each single-scan NMR spectrum was collected immediately after shaking the sample in a magnetic feld of 65 G (unless otherwise specifed), whilst under 3 bar of p-H 2 . Enhancement factors were calculated by comparing the peak areas in the resulting NMR spectra with those of analogous traces collected under normal H 2 and Boltzmann equilibrium conditions. In order to do this, the peak areas of individual resonances in the corresponding normal Scheme 1 Proposed ligand exchange pathways of pyridazine, the blue nitrogen label is a visualization aid. NMR spectra were calibrated to 1 and enhancements calculated by comparing these values to those in the polarised spectra. Total signal enhancements were calculated by multiplying this enhancement value by the number of protons in the group and summing across the molecule. ## Polarisation transfer eld measurements Polarisation transfer feld measurements were made using a dedicated hyperpolarisation apparatus. 44 Samples were frst prepared in 3 mL of methanol-d 4 in an ampoule containing 10 mg of 1 and 5 equivalents of pyridazine or phthalazine. The solution was degassed, and the ampoule was flled with 3 bar pressure of H 2 and heated to 348 K overnight to form 4a or 4b. At this point, the resulting solutions were introduced into the flow system and the SABRE effect monitored for polarisation transfer felds that ranged from 0.5 to 140.5 G, in steps of 10 G. This involved the collection of a series of single-scan spectra using either p/2 pulses or shaped pulses 50,52 for the multiple quantum fltered experiments. The peak integrals in the resulting NMR spectra were compared to those of a thermally polarised spectrum to determine signal enhancement factors. ## EXSY measurements A series of exchange spectroscopy (EXSY) measurements were made to probe the dynamic behaviour of these systems. 53 This process involved the selective excitation of a single resonance and the subsequent measurement of a 1 H NMR spectrum at time, t, after the initial pulse. The resulting measurements consisted of a series of data arrays such that t was varied over 10 values, typically from 0.1 to 1.0 s. The precise values were varied with temperature to suit the speed of the process. Data was collected for a range of temperatures and sample concentrations as tabulated in the ESI. † ## Kinetic analysis Integrals for the interchanging peaks in the associated 1 H EXSY spectra were obtained and converted into a percentage of the total detected signal. These data were then analysed as a function of the mixing time according to a differential kinetic model. 54 Rates of exchange were determined by employing the solver add-on in Microsoft EXCEL to minimize the sum of the residuals in the associated least mean squares analysis. The rate constants obtained in this way were doubled when calculating barrier heights to account for step reversibility. 55 Errors associated with the calculated rate constant were determined by the Jackknife procedure of Harris. 56 ## Image acquisition and processing For the imaging experiments, samples containing 4a and 4b were prepared as described earlier. All data was then acquired on a 400 MHz Bruker spectrometer equipped with a micro imaging gradient system (1 T m 1 ) and a standard 30 mm 1 H- 13 C double resonance birdcage. For the SABRE experiments the sample tubes were flled with 3 bar of p-H 2 and shaken in the stray feld of the MRI magnet (ca. 70 G). As a reference, images of the samples with Boltzmann-equilibrated magnetisations were recorded in conjunction with a recycle delay of 100 s and 512 averages. A standard rapid acquisition with refocused echoes (RARE) 57 sequence composed of a train of 64 p pulses separated by an echo time (TE) of 8.15 ms (effective echo time, TEeff ¼ 114 ms) was used to collect the images. According to Hennig et al., by using a relatively small number of echoes, combined with a short echo time, the effect of transverse relaxation is minimised and the image contrast is dictated by spin density. 57 All images used a feld of view (FOV) of 30 30 mm, a slice thickness of 5 mm and a 64 64 matrix size. This led to a raw resolution of 0.47 0.47 5 mm 3 . Data were processed by FT and zero flled to 128 128 (resulting digital resolution: 0.23 0.23 5 mm 3 ) using a sine bell noise flter. Signal-to-noise (SNR) values were calculated using an algorithm that assumed a Rice-type noise distribution. ## Reaction of [Ir(IMes)(COD)Cl] (1) with pyridazine and phthalazine It has already been reported that when 1 reacts with pyridine in methanol-d 4 it forms [Ir(COD)(IMes)(py)]Cl. 44 When pyridine is replaced with either pyridazine or phthalazine the corresponding solutions immediately change colour from yellow to orange due to the formation of the analogous complexes [Ir(COD)(IMes)(pdz)]Cl (2a) and [Ir(COD)(IMes)(phth)]Cl (2b). Both of these complexes have been characterized by NMR spectroscopy and mass spectrometry. Attempts to follow this process by stop-flow UV-Vis spectroscopy, with a dead-time of 0.6 ms, failed because the reaction was too rapid. ## Ligand exchange exhibited by [Ir(COD)(IMes)(pdz)]Cl (2a) and [Ir(COD)(IMes)(phth)]Cl (2b) When these products were analysed by 1 H NMR spectroscopy, signals for the bound N-heterocycles could not be resolved at room temperature. To overcome this, a series of NMR spectra were collected at 253 K. The ESI † presents the corresponding 1 H, 13 C and 15 N resonance assignments for these species. A series of 2-D nOe spectra were also collected at 253 K. These confrmed that the NCHCH and NCH protons of the pyridazine ligand in 2a and the phthalazine ligand in 2b switch their nitrogen binding sites on a timescale of 0.5 seconds (see Scheme 2). The broadening of the NMR signals for these ligands, seen at room temperature, is due to this rapid exchange. NMR signals for the free ligand were also present in these spectra, although no exchange into them was observed at this temperature, or upon warming to 298 K. This is therefore an intramolecular process. The intramolecular nature of this process means that the 1,2metallotropic shift is haptotropic. It occurs via an intermediate with an h 2 -N]N coordination mode according to Eaton et al., 12 rather than dissociation which would lead to 14-electron [Ir(COD)(IMes)] + and the observation of exchange with the free ligand pool. The rate of the haptotropic shift was determined over the temperature range 240 to 260 K by EXSY spectroscopy. Exchange in 2a, between the proton signal at d 7.69 (corresponding to IrNCHCH) and that at d 7.51 (corresponding to IrNNCHCH), was probed. The analogous process in 2b was followed between 235 and 255 K by selecting the proton signal at d 9.55 (corresponding to IrNCH) and observing its exchange into the signal at d 8.37 (corresponding to IrNNCH). The associated exchange rates were quantifed by data ftting (see ESI †). Fig. 1 shows a typical exchange profle, while Table 1 lists the corresponding rate data. The activation parameters for these processes were calculated by Eyring analysis and are given in Table 2. They are similar to those reported. 11, The enthalpy of activation for the haptotropic shift of 2a is larger than that of 2b by 5 kJ mol 1 . This difference suggests that the Ir-pdz bond is stronger than the Ir-phth bond and yet the reported pK a values of pyridazine and phthalazine are 2.33 and 3.50, respectively. 58 We suggest that steric interactions result in the smaller activation entropy gain for binding site interchange of the larger and more basic phthalazine ligand. ## Addition of H 2 to [Ir(COD)(IMes)(pdz)]Cl (2a) and [Ir(COD)(IMes)(phth)]Cl (2b) Two separate methanol-d 4 solutions of 2a and 2b were exposed to a 3 bar pressure of hydrogen at 240 K. In both cases the solutions changed colour from orange to colourless due to the formation of [Ir(H) 2 (COD)(IMes)(pdz)]Cl (3a) and [Ir(H) 2 -(COD)(IMes)(phth)]Cl (3b), respectively. These complexes are easily identifed by 1 H NMR spectroscopy as they yield two coupled hydride resonances, which in the case of 3a lie at d 13.84 and 17.69, and for 3b lie at d 13.87 and 17.55. As the hydride ligands of both these products are chemically inequivalent, they must adopt one of two possible product geometries, assuming that hydrogen addition follows the expected concerted route, as shown in Scheme 3. 59 In one of these product geometries, the hydride ligands lie trans to the Nheterocycle and COD (3a) while in the second they lie trans to IMes and COD (3a 0 ). 60 One route to differentiate these two orientations is to recognise that a trans H-Ir-15 N coupling is signifcantly larger than a cis coupling. When a 15 N-optimized HMQC NMR spectrum was recorded to probe this, strong cross-peaks were found between the hydride signals at d 17.55 (of 3b) and 17.69 (of 3a) and the nitrogen signals of the bound N-heterocycle at d 294.17 and 266.53 respectively. This confrms their mutual trans orientation and that 3a and 3b form by H 2 addition over the COD-Ir-N axis of 2a and 2b respectively. This is consistent with the nature of the ligands determining the direction of addition, which in this case favours addition over the axis containing the weak nitrogen donor rather than the strong IMes ligand. 61 Furthermore, the corresponding 2D-1 H COSY NMR spectrum contains cross-peaks between the hydride ligand signals at d 13.84 (of 3a) and 13.87 (of 3b) and the inequivalent proton signals that arise from the two CH environments in the bound COD that are trans to hydride. In the case of 3a these COD signals appear at d 5.09 and 4.45. Both species 3a and 3b have been fully characterised by NMR spectroscopy. We note that when solutions of 3a and 3b are degassed, reductive elimination of H 2 occurs and 2a and 2b reform. H 2 addition is therefore a reversible process. These complexes are examples of a relatively rare class of alkene dihydride species that are typically active in alkene hydrogenation. A related iridium pyridylpyrrolide COD complex has been shown by Searles et al. to reversibly bind hydrogen in a similar way. One route to probe the concerted nature of this process is to employ p-H 2 . This is because PHIP requires the pair-wise In order to compare the reactivity of these complexes we prepared a sample that contained approximately equal amounts of pyridazine and phthalazine (6.73 equivalents of phth and 6.16 equivalents of pdz relative to 1 as measured by NMR). Upon NMR examination, the binding of phthalazine proved to dominate over that of pyridazine, with the ratio of 2b : 2a being 3.7 : 1. When corrected for these slightly unequal starting ligand concentrations, the equilibrium ratio is 3.40 : 1 and DDG 2a-2b 0 at 240 K is approximately 2.44 kJ mol 1 with 2b being more stable than 2a. The normalized PHIP enhancements that are exhibited by the hydride signals of species 3a and 3b upon p-H 2 addition at 240 K are 56.8-fold and 38.2-fold, respectively. As the p-H 2 concentration in solution is constant and its purity identical at the point of measurement, these numbers can be compared. If we assume that both complexes retain the same level of p-H 2 derived magnetisation at the point of their formation then their relative signal intensities will be proportional to the product of their rates of reaction and concentration. The greater abundance of 2b (3.7), in combination with the hydride signal enhancement levels, suggests that the rate of H 2 addition to the more stable 2b is a factor of ca. 4.7 slower than that to 2a. The PHIP effect that is seen in these hydride signals vanishes rapidly upon the introduction of the sample into the NMR probe at 240 K. This is because the hydrides do not undergo H 2 elimination on the NMR timescale at this temperature. However, if the sample is removed from the magnet and shaken to dissolve fresh p-H 2 , the PHIP effect is re-established in the associated hydride ligand signals. This is due to the transiently higher temperatures that the sample experiences whilst shaken. Under these conditions, however, 3a and 3b are present and it is their exchange rates that control the level of PHIP. In subsequent NMR spectra the levels of observed signal enhancement are almost equal, at 27.6-fold and 20.1-fold. However, in the corresponding thermally equilibrated NMR spectra, at 240 K, the phthalazine dihydride complex 3b still dominates, now by a factor of 3.2. In fact, these data suggest that DG 3a-3b 0 is 2.33 kJ mol 1 at 240 K. Based on these data we can conclude that less stable 3a undergoes more rapid H 2 loss than 3b. Limited evidence for the formation of the isomers 3a 0 and 3b 0 of Scheme 3 was observed. This is in agreement with the IMes-Ir-COD axis being the most electron rich and the most sterically hindered, and therefore the least able to promote H 2 addition across it. 59,63 Hydride signals for these species were seen at d 9.12 and 13.48 for 3a 0 and at d 9.11 and 13.48 for 3b 0 , with very low intensity. The values of the hydride chemical shifts in this pair of complexes reflect their trans-orientations to soft carbene and COD ligands. 64 The signals did not exhibit PHIP, despite being coupled. They are also not visible in the low temperature 1 H NMR spectra but can be observed at 298 K. Upon cooling from this temperature, in an effort to trap the species, the corresponding hydride signals vanish and only those for 3a and 3b remain. We conclude from these data that stability increases in the order 2a < 2b < 3a < 3b and that the addition of H 2 proceeds more rapidly to 3a. These data confrm that 3a undergoes pyridazine loss on a shorter timescale than NMR relaxation. One surprising feature of these chemical systems is the polarisation of bound CH proton environments of the h 4 -COD ligand, at d 5.09 and 4.45 in the case of 3a, although no free COD, COE or COA signals are evident. We note that there is no evidence in the corresponding EXSY spectra to suggest that the bound COD reversibly accepts a hydride ligand. 65 SABRE magnetisation transfer proceeds readily from the hydride ligands of 3a and 3b into the ligands that are trans to them. Both 3a and 3b were stable at temperatures up to 250 K, with their hyperpolarisation effects disappearing in solution after several seconds. Above this temperature, however, the COD ligand hydrogenates and forms COA; there is no NMR evidence for COE. 65 The vacant sites that are created by the hydrogenation of COD then become occupied by two further pyridazine or phthalazine ligands. This leads to the formation of [Ir(H) 2 -(IMes)(pdz) 3 ]Cl (4a) and [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b) respectively. These reactions take approximately one hour to go to completion at 298 K. Both 4a and 4b have been characterised by NMR spectroscopy and mass spectrometry. The reaction pathway for the conversion of 1 into 4a is illustrated in Scheme 4. A weak unassigned hydride signal is seen in these 1 H NMR spectra at d 22.87. This may correspond to an intermediate with a similar structure to that of [IrH(1-k-C 8 H 13 )(PMe 3 )(NCCH 3 ) 2 ]BF 4 , which provides a diagnostic hydride resonance at d 20.35. 65 The organic region of a 1 H NMR spectrum of 4a is shown in Fig. 3. The corresponding hydride region, which is not displayed, contains a single peak, at d 21.47, due to the two chemically equivalent, but magnetically distinct, hydride ligands of this complex. The organic region of this NMR spectrum is complicated as the symmetry of pyridazine is broken upon binding to the metal centre. Seven new aromatic 1 H resonances are visible, four of which appear as simple doublets, at d 9.62, 9.31, 8.75 and 8.29 with integral ratios of 1 : 2 : 2 : 1 respectively, relative to the 2-proton hydride ligand signal. Their doublet nature means that they arise from pyridazine NCH environments, and their relative areas confrm the presence of a single axial ligand (d 9.62 and 8.29) and two equivalent equatorial ligands (d 9.31 and 8.75). The remaining three multiplets, at d 7.62, 7.48 and 7.40, have integral ratios of 2 : 1 : 3 respectively and arise from the remaining NCHCH environments. The arrangement of these groups around the metal centre was confrmed by nOe measurements at 300 K with mixing times of $0.6 s. These data revealed that both the equatorial IrNCH and IrNNCH environments interact with methyl-IMes protons that resonate at d 2.22 and 2.04. These connections confrm that the equatorial pyridazine ligands lie in close proximity to the IMes ligand. Complex 4a was further characterised by extensive COSY, 13 C-optimised HMQC and 15 N-optimised HMQC measurements. The phthalazine analogue of 4a, [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b) was characterised in a similar way. Part of its 1 H NMR spectrum is shown in Fig. 3b. A single hydride peak is also observed for this product, at d 21.0. Peaks at d 9.63, 8.26 and 8.15 correspond to the NCH, NCHCCH and NCHCCHCH environments of free phthalazine respectively. Four singlets are observed at d 10.28, 10.16, 9.27 and 8.60 with integral ratios of 1 : 2 : 2 : 1. These resonances correspond to the bound NCH environments of 4b and can be broken down into axial and equatorial contributions. NMR data for 4a and 4b is given in the ESI. † Modern NMR methods, specifcally inverse sequences, allow insensitive nuclei, such as 15 N, to be detected when they are present at natural abundance. 66 15 N NMR data, collected by inverse HMQC methods for 4a and 4b are detailed in Table 3. In the case of 4a, pyridazine binding to iridium moves the 15 N Scheme 4 Products of H 2 and pyridazine addition to 1. chemical shift of the free ligand from d 383.2, to d 326.2, when bound trans to hydride, and to d 305.8 when bound trans to carbene. The value of the change in chemical shift on binding Dd (free-bound) has been suggested to relate to the extent of interaction with the metal. 67 As revealed in Table 3, the largest Dd value is observed for the axial N-heterocycle arrangement with the inplane change being 30% smaller. This suggests that there is a weaker in-plane ligand interaction as a consequence of the greater trans-influence of hydride relative to IMes. 67 The corresponding Dd values for the remote nitrogen centres in these ligands are much smaller and positive in sign thereby illustrating that electron density surrounding them is signifcantly less perturbed by the introduction of a metal-ligand interaction. ## Hydride and N-heterocycle ligand exchange dynamics of [Ir(H) 2 (IMes)(pdz) 3 ]Cl (4a) and [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b) The ligand exchange pathways of 4a and 4b have been explored by NMR spectroscopy. The samples examined here employ d 4methanol as the solvent. A series of 1D EXSY measurements were undertaken, at 10 different temperatures, and both ligand exchange and H 2 loss pathways were revealed. These processes were too slow to be detected at temperatures below 280 K and the upper limit of our measurements was 325 K, due to the boiling point of methanol. At this temperature, the exchange rates were too slow to observe any line-shape changes. In the frst of our EXSY experiments, peaks corresponding to the NCH environment of the axial ligands in 4a and 4b were probed. No magnetisation transfer was observed from these sites, at 320 K, and we conclude that the axial ligands in both 4a and 4b remain bound to the metal on this timescale and that their involvement in exchange can be neglected. This observation matches the higher strength of interaction predicted through the larger 15 N-Dd value detailed earlier. In contrast, when the equatorial ligands of 4a and 4b were probed, magnetisation transfer indicative of ligand exchange and the 'shifting' of the equatorial ligand's nitrogen binding centre was observed by a 1,2-metallotropic shift. A typical EXSY trace is shown in Fig. 4. Both of these processes are common in such systems. Shifting coordination from one nitrogen centre to the other is likely to proceed via a trigonal bipyramidal transition state, rather than a Berry-pseudorotation, in accordance with the molecular orbital studies by Alvarez et al. 19 When the hydride ligand signals of 4a and 4b were examined, in a similar way, exchange into the signal for H 2 was observed. The rate constants for these processes were calculated by analysing an array of EXSY spectra that were recorded with a series of magnetisation transfer times. ESI Scheme 2 † illustrates pictorially the exchange processes that occur with pyridazine, and the extracted rate constants as a function of temperature are also available. Eyring analysis of these data provided the activation parameters that are shown in Table 4. The barrier to Ir-N bond rupture (dissociation) is twice the sum of the full ligand loss rate if upon reaching the transition state there is an equal probability of a null-reaction. The haptotropic shift was considered as a separate intramolecular pathway in this analysis. 3.9. Mechanism of pyridazine and phthalazine ligand dissociation and binding site shift in [Ir(H) 2 (IMes)(pdz) 3 ]Cl (4a) and [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b) Exchange behaviour was probed as a function of substrate concentration at 285 K (see ESI Tables 6-10 †). A plot of ligand exchange rate versus ligand concentration, for a constant concentration of 1, is presented in Fig. 5b for 4a and Fig. 5c for 4b. It can be seen from these data that the experimental rate constants for pyridazine and phthalazine ligand dissociation are unaffected by an increase in the free ligand concentration (Fig. 5b and c). This zero-order rate behaviour is consistent with a dissociative process and a haptotropic shift. 18 The DH ‡ values in Table 4 reflect the associated Ir-N bond energies and the barriers to ligand exchange. For each pair of complexes, the enthalpy change on reaching the transition states for the haptotropic shift is lower than that of dissociation. Bulkier phthalazine has a weaker Ir-N bond and hence smaller DH ‡ term. Consequently, the Ir-N bond is longer and the gain in entropic freedom on reaching the transition state for both phthalazine dissociation and shifting falls. The two effects act to compensate each other such that their DG s (300 K) values are similar. This is particularly noticeable in the haptotropic shift, where the DH ‡ and DS ‡ terms of phthalazine are lower by 16.2 kJ mol 1 and 55 J K 1 mol 1 respectively. As mentioned previously, the literature values for the pK a of pyridazine and phthalazine are 2.33 and 3.50 respectively. Our observed values of 15 N-Dd upon ligand binding were 57.0 and 49.9 respectively. Less basic pyridazine exhibits a stronger bonding interaction than phthalazine; ground state steric effects must therefore control the Ir-N bond strength. ## Mechanism of H 2 loss from [Ir(H) 2 (IMes)(pdz) 3 ]Cl (4a) and [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b) The rate of H 2 loss from 4a depends on the pressure of H 2 for a constant substrate excess, as shown in Fig. 5a. However, when the H 2 loss pathway was monitored at constant H 2 pressure as a function of substrate concentration, inhibition was observed (Fig. 5b and c). The pseudo-rate constant for H 2 loss is inversely proportional to the substrate concentration (Fig. 5d). This behaviour matches with that predicted by DFT for the related pyridine system in which the dihydrogen-dihydride intermediate [Ir(H) 2 (H 2 )(IMes)(py)]Cl is involved in H 2 exchange. 44 The inhibition by substrate occurs as it reacts with 16-electron [Ir(H) 2 (IMes)(sub)]Cl rather than H 2 . Both 4a and 4b undergo ligand exchange on the NMR timescale. Such exchange is necessary for the SABRE effect to successfully hyperpolarise a material. We investigated the effect of shaking a sample of 1 (5.2 mM), containing 5 equivalents of pyridazine and 3 bar of p-H 2 , in the Earth's magnetic feld for 10 seconds at 298 K, prior to collecting a 1 H NMR spectrum. 1 H NMR signal enhancements are observed in all of the free and bound equatorial substrate resonances, and the hydride resonance of these complexes. No evidence for the formation of any dinuclear complexes was evident, and the phase of the observed hyperpolarised signals for the corresponding free and bound sites were identical. Upon integration, the free pyridazine NCH resonance was 180times larger than the corresponding signal recorded when the magnetisation was probed under thermal equilibrium conditions. The total signal enhancement of the two resonances in this substrate was 410-fold. When a similar phthalazine sample was probed in an identical way, the hydride signal for 4b, and all of the free and equatorially bound substrate resonances were again enhanced. In this case, the corresponding free NCH resonance was 85-fold enhanced and the total molecular enhancement was 195-fold. The net enhancement factor therefore drops signifcantly on moving from pyridazine to phthalazine, despite their comparable ligand loss rates. Furthermore, the remote ring of phthalazine, which is connected by a 5-bond coupling of 0.43 Hz into the frst ring is only weakly polarised by comparison to the frst ring (14%). The underlying difference of 2.1-fold between the polarisation levels observed for these two molecules cannot be accounted for by a spin dilution effect (the consequence of sharing the magnetisation of p-H 2 with 6 rather than 4 receptor protons). By implication, the J-coupling that exists between the hydride and Ir-NCH protons of pyridazine must be larger than that to phthalazine, hence better polarisation transfer results; this fts with the predicted shorter bond length. The concentration dependence on the level of SABRE seen in a series of 1 H NMR spectra recorded with p/2 pulses is illustrated in Fig. 6a for pyridazine and Fig. 6b for phthalazine. It is clear, in these data, that the proton resonance which corresponds to the site next to nitrogen is most strongly enhanced with 4a delivering a 380-fold NCH intensity gain when a 17 mM concentration of pyridazine is employed in conjunction with a metal concentration of 6.24 mM. The signal enhancement level falls as the substrate concentration increases, when a constant metal concentration is maintained. In addition to reflecting spin-dilution, this change also results in a progressive reduction in the rate of H 2 exchange. Both of these effects reduce the levels of signal gain seen for all of the protons of pyridazine and phthalazine. Surprisingly, when the inverse of the sum of the enhancement levels is plotted against equivalents of free substrate, a negative linear relationship can be seen (Fig. 6c). The signal intensity gains that were observed in these 1 H NMR spectra also varied with temperature as shown in Fig. 7. A continuous growth in the hyperpolarisation level with increasing temperature is observed between 273 K and 323 K, the highest temperature used. In the frst of these samples, the concentration of free pyridazine was 12.3 mM while in the second, the concentration of free phthalazine was 14.1 mM. The corresponding rates of H 2 loss at 323 K were estimated to be 29.7 s 1 and 13.49 s 1 respectively. This suggests that an optimal ligand exchange rate has not been reached at 323 K, despite the short lifetimes of these complexes. The gain in signal strength appears to be linear with temperature, although the gradient exhibited by the 5 resonance intensities differs substantially, falling with distance from the ligation site. This change mirrors the efficiency of their polarisation. The gradients shown for the pyridazine signals are larger than those for phthalazine in accordance with its greater exchange rate sensitivity to temperature. Furthermore, the sensitivity to temperature falls as the substrate excess increases (Fig. 7c). This observation can be rationalised by the fall in H 2 exchange rates with increase in ligand concentration. ## Probing the effect of sample dilution on the level of SABRE enhancement In the case of pyridazine, the effect of the concentration of 4a (with the excess of free pyridazine kept constant) on the level of SABRE enhancement has been rigorously studied. The smallest ligand excess delivered the largest level of polarisation, alongside a metal concentration of 2.8 mM. The maximum total signal enhancement observed in the 1 H NMR spectra collected using pyridazine was 1278-fold with a p/2 pulse. Normally, the S/N value associated with a given resonance is proportional to concentration. Here, however, the S/N value is linked to the product of the concentration and the hyperpolarisation level. An approximately linear dependence on the detected signal strength with concentration was observed experimentally for samples in which the substrate concentration lay between 2.8 mM and 6.2 mM, as shown in Fig. 8. We conclude from these observations that there is no beneft in working at higher concentration, even if there is an abundance of the available substrate, since the fall in SABRE level acts to offset the signal gain resulting from an increase in concentration. 68,69 3.13. Probing the effect of the polarisation transfer feld on the efficiency of SABRE The variation in the level of SABRE observed with the magnetic feld experienced by the sample at the point of polarisation transfer is substantial (see ESI †). Both substrates behave in a similar way, each yielding their largest free ligand single spin enhancement values when the polarisation transfer feld is 65 G. Consequently, this feld should be used for detection. SABRE, however, also creates higher order magnetic terms which are optimally enhanced at different feld values (see ESI †). Ultimately, the maximum polarisation level reaches a plateau as T 1(effective) of the probed state counteracts the rate of SABRE build-up. The polarisation builds up most quickly in the single spin NCH protons, and ultimately reaches a higher polarisation level than any other spin state. While the observed rate of twospin order build up is comparable to that of a single spin their ultimate amplitudes are lower as a consequence of their more rapid relaxation (se ESI †). ## Using SABRE to collect MRI data We recorded a series of SABRE images and compared them to their thermal counterparts (Fig. 10). As expected, these images exhibited better signal to noise, improved contrast, and signal intensity. Furthermore, the samples of pyridazine were superior to those of phthalazine in terms of signal intensity (one order of magnitude higher) for the same concentration of substrate. In order to more precisely quantify the improvement in image quality, fve one-shot images of the hyperpolarised samples were acquired and the resulting image SNR's calculated. The average result was divided by the SNR of a reference image that was recorded under Boltzmann equilibrium conditions using 512 averages. The underlying improvement in contrast and image quality is revealed in Table 5. A considerable increase in the SNR values was observed, and the acquisition times required for SABRE images were 6 orders of magnitude smaller than the scan times necessary under thermal equilibrium conditions (260 ms versus several hours). The optimum conditions used to hyperpolarise both pyridazine and phthalazine in methanol, at 298 K, have been applied to 5-aminophthalazine, which has antifungal properties. 73 Preliminary results are very promisingwith total signal enhancements of ca. 500-fold being observed. The applicability of SABRE, in hyperpolarising pyridazine and phthalazinederived drug molecules for use as contrast agents, is dependent ultimately on performance in biologically-compatible solvents. Although water would seem an obvious solvent choice, SABRE has not yet been shown to work efficiently in this medium. 37,39 An ethanol-water mixture is, however, biocompatible, hence the use of ethanol, followed by dilution has been predicted to be viable. 70,74 We have therefore examined pyridazine and phthalazine in ethanol-d 6 . Our preliminary results reveal superior SABRE in this solvent, with the pyridazine signal enhancement of ca. 400-fold (9.62-fold excess, 60 mM free pyridazine) being 33% better than in methanol. For phthalazine the total signal enhancement was 480-fold (2.12-fold excess, 13.2 mM free phthalazine) or 20% better. Our studies on 5-aminophthalazine confrm that further functionalisation of the remote ring will not prevent SABRE activity and an array of such materials could be probed. 75 ## Conclusions We have established that [Ir(IMes)(COD)Cl] (1) reacts rapidly with pyridazine and phthalazine to form the ligand substitution products [Ir(IMes)(COD)(pdz)]Cl, 2a and [Ir(IMes)(COD)(phth)] Cl 2b, respectively. Phthalazine is more basic than pyridazine, and 2b is more stable than 2a. 76 These complexes undergo a haptotropic shift on the NMR timescale which is more rapid for 2b. Complexes 2a and 2b add H 2 to form the corresponding Ir(III) complexes [Ir(H) 2 (COD)(IMes)(pdz)]Cl (3a) and [Ir(H) 2 (COD)-(IMes)(phth)]Cl (3b), where 3b is more stable than 3a. When the H 2 addition reaction was monitored with p-H 2 , PHIP was observed in the hydride NMR signals of these two products and 2a is more reactive to H 2 addition than 2b. 3a and 3b also exhibit SABRE in specifc ligand resonances that lie trans to the hydrides. This demonstrates that hyperpolarisation can be used to characterise such complexes, even if they have a transient existence. 3a and 3b are highly reactive and upon warming they rapidly convert to the dihydrides [Ir(H) 2 (IMes)(pdz) 3 ]Cl (4a) and [Ir(H) 2 (IMes)(phth) 3 ]Cl (4b). These two complexes also undergo a haptotropic ligand shift in addition to full ligand loss on the NMR timescale. The iridium-nitrogen bonds of 4a are stronger than those of 4b. The hyperpolarisation of both pyridazine and phthalazine by 4a and 4b respectively, under SABRE, proved to be facile. NMR enhancements increase with temperature, due to faster ligand exchange, and are optimised with 5 equivalents of substrate, and a polarisation transfer feld of 65 G. An optimised enhancement level, at 298 K, of 1278-fold was achieved for pyridazine. The total signal enhancement of pyridazine was consistently larger than that of phthalazine even though the ligand loss rates of 4a and 4b were comparable. This difference in behaviour cannot be accounted for by the sharing of magnetisation from p-H 2 with 4 rather than 6 receptor protons, but instead suggests that the J-coupling between the hydride and Ir-NCH protons of pyridazine is larger than that to phthalazine. This deduction agrees with the predicted shorter Ir-N bond length in 4a. A series of more complex magnetic states can also be created which allow the observation of these substrates in protio solvents through the OPSY protocol. Finally, we have shown that it is possible to collect optimised MRI images for these substrates on phantoms. Whilst neither pyridazine nor phthalazine are used clinically, their derivatives, in the form of drugs such as Hydralazine, 77 play an important role in medicine; we have polarised 5-aminophthalazine which has antifungal properties and achieved total enhancement levels of 500-fold. In the future, we plan to use the results of this study to aid in the optimisation of SABRE for the analysis of similar heterocycles that are used pharmaceutically.

chemsum

{"title": "Investigating pyridazine and phthalazine exchange in a series of iridium complexes in order to define their role in the catalytic transfer of magnetisation from para-hydrogen", "journal": "Royal Society of Chemistry (RSC)"}

synthetic_glycosidases_for_the_precise_hydrolysis_of_oligosaccharides_and_polysaccharides

## Abstract: Glycosidases are an important class of enzymes for performing the selective hydrolysis of glycans. Although glycans can be hydrolyzed in principle by acidic water, hydrolysis with high selectivity using nonenzymatic catalysts is an unachieved goal. Molecular imprinting in cross-linked micelles afforded water-soluble polymeric nanoparticles with a sugar-binding boroxole in the imprinted site. Post-modification installed an acidic group near the oxygen of the targeted glycosidic bond, with the acidity and distance of the acid varied systematically. The resulting synthetic glycosidase hydrolyzed oligosaccharides and polysaccharides in a highly controlled fashion simply in hot water. These catalysts not only broke down amylose with similar selectivities to those of natural enzymes, but they also could be designed to possess selectivity not available with biocatalysts. Substrate selectivity was mainly determined by the sugar residues bound within the active site, including their spatial orientations. Separation of the product was accomplished through in situ dialysis, and the catalysts left behind could be used multiple times with no signs of degradation. This work illustrates a general method to construct synthetic glycosidases from readily available building blocks via self-assembly, covalent capture, and post-modification. In addition, controlled, precise, one-step hydrolysis is an attractive way to prepare complex glycans from naturally available carbohydrate sources. ## Introduction Carbohydrates are the most abundant biomolecules on earth. Cellulose makes up 35-50% of lignocellulosic biomass that is produced at an annual scale of 170-200 billion tons. 1,2 Starch, consisting of linear and branched polymers of glucose, is the dominant energy-storage material in plants and the predominant carbohydrate in the human diet. With glycosylation being the most common post-translational modifcation of proteins, carbohydrates (glycans) mediate important biological events including cell adhesion, bacterial and viral infection, inflammation, and cancer development. To process carbohydrates, most organisms use 1-3% of their genome to encode enzymes for glycosylation and hydrolysis of glycans. 9 Many of these enzymes, however, cannot be obtained easily and new catalysts with controlled glycosidic selectivity are in great need for glycomics. 3 Moreover, enzymes tend to work only under a narrow set of conditions in aqueous solution. Synthetic catalysts with stronger tolerance to adverse temperatures and solvent conditions are highly desirable for challenging operations such as biomass conversion. 10 Chemists have long been interested in creating synthetic glycosidases to hydrolyze glycosides/glycans. In their pioneering work, Bols and co-workers used cyclodextrin to bind p-nitrophenyl b-D-glucopyranoside and acidic groups installed on the macrocycle for hydrolysis. 11,12 Striegler et al. developed binuclear copper catalysts to hydrolyze glycosides under basic conditions. 13,14 The group of Bandyopadhyay reported azobenzene-3,3 0 -dicarboxylic acid as a simple glycosidase mimic with photoresponsive properties. 15 Although these catalysts only worked on activated glycosides carrying good aryl leaving groups, they proved important design principles with readily available scaffolds. Instead of relying on hydrolysis, Yu and Cowan combined a metal-binding ligand and the sugar-binding domain of odorranalectin (a natural lectin-like peptide) to remove L-fucose selectively through oxidative cleavage. 16 The fundamental challenge in building a synthetic glycosidase is two-fold, whether on a purely synthetic platform or a hybrid one as above. First, the catalyst needs to recognize the targeted glycan in an aqueous solution. Molecular recognition of carbohydrates is a long-standing challenge in supramolecular and bioorganic chemistry, due to strong solvation of these molecules in water. 17 Also, inversion of a single hydroxyl in a glycan, connecting the monosaccharide building blocks by different hydroxyls, or altering the (a/b) glycosidic linkages could all change the biological properties of a glycan profoundly. Distinguishing these subtle structural changes requires a great level of selectivity in the recognition. Second, catalytic groups need to be installed precisely at the correct glycosidic bond for selective hydrolysis. Although many platforms are available for building artifcial enzymes, 18,19 this type of precision is still very difficult to achieve for a complex substrate. In this work, we describe a synthetic glycosidase designed and synthesized rationally through molecular imprinting within a cross-linked micelle. 23 Having an acidic group near the exocyclic oxygen of a particular glycosidic bond while recognizing the adjacent sugar residues with reversible boronate ester and hydrogen bonds, these biomimetic catalysts hydrolyzed oligosaccharides and polysaccharides in a precise manner in 60 C water, to afford specifc glycans in a single step. The catalysts also distinguished sugar building blocks and the glycosidic linkages, especially those bound within the active site. Because synthesis of complex glycans requires extensive protection/deprotection chemistry and is often very challenging, selective hydrolysis of naturally available glycan sources can be a highly attractive alternative. ## Construction of the active site for glycan binding To build an active site for a glycan, we employed molecular imprinting that can quickly produce binding sites within a cross-linked polymer complementary to the template molecules being used. The templated polymerization was particularly effective in a nanosized environment such as a surfactant micelle. 24 Scheme 1 shows the general method of micellar imprinting. The template molecule is frst solubilized in the mixed micelle of 1a (or 1b) and 2. These surfactants are functionalized with either terminal alkyne or azide on the headgroup. Micellization brings these functional groups into close proximity, and they react readily by Cu(I)-catalyzed cycloaddition to cross-link the surface of the micelle. The mixed micelle also contains divinylbenzene (DVB) and 2,2-dimethoxy-2-phenylacetophenone (DMPA, a photoinitiator). UV irradiation triggers free-radical polymerization/cross-linking in the core of the surface-cross-linked micelle, among DVB and the methacrylate of the surfactants around the template molecule. The doubly cross-linked micelle is then decorated with a layer of hydrophilic ligand (3) by a second round of click reaction. Precipitation into acetone and washing with organic solvents removes the template and other impurities to yield water-soluble molecularly imprinted nanoparticles (MINPs) with templatecomplementary binding sites. 23 Boronic acid and benzoboroxole derivatives are known to bind specifc 1,2-or 1,3-diols of sugars through reversible boronate bonds. Vinylbenzoboroxle 4 is particularly useful as a functional monomer (FM) for covalent imprinting 36 of sugars. 35 Aryl glycosides 5a-c have two important partsa glycan and an aglycon containing a hydrolyzable imine bond (Scheme 2). These template molecules reacted with FM 4 readily to form amphiphilic, anionic boronate esters such as p-6a, stabilized by the cationic micelle. 35 The imine bond in MINP(p-G1) was hydrolyzed in 6 M HCl at 95 C, and the aldehyde group in the imprinted site of MINP-CHO(p-G1) derivatized through reductive amination with 7a-f in DMF. 37 The resulting MINP(p-G1 + 7a-f), i.e., the MINP prepared with template 5a and postfunctionalized with amino acid 7a-f, was purifed by precipitation into acetone and washing with organic solvents. The MINP is expected to bind the terminal glucose of a glucoseterminated oligo-or polysaccharide at the nonreducing end, with its carboxylic acid near the exocyclic oxygen of the glycosidic bond. The intermediate MINP-CHO(p-G1) indeed was found to bind not only glucose but also maltose and other glucoseterminated oligosaccharides strongly in aqueous buffer (Table 1, entries 1-4). One might consider it strange for a binding site imprinted from a monosaccharide template to bind a larger oligosaccharide, as several works of ours indicate that ftting a smaller guest in a larger imprinted pocket affords a reduced binding, but ftting a larger guest in a small pocket is impossible. 35,38,39 However, the amphiphilicity of the template-FM complex (p-6a) demands it stay near the surface of the micelle, a feature helpful to not only the removal of the template after imprinting but also anchoring the binding site near the surface of the micelle. In this way, a glucose-terminated oligosaccharide could use its terminal glucose to interact with the boroxole in the MINP binding pocket, with the rest of the structure residing in water, as shown in Scheme 2. Ability to bind a longer sugar is a prerequisite to the catalysis. Had a longer sugar been excluded, the acid-functionalized MINP(p-G1) would not be able to recognize its substrate (G2 and above). Table 1 also shows that the binding of MINP-CHO(p-G1) weakened steadily with an increase in the chain length of the sugar guest. We attributed the trend to the 1,4-a-glycosidic linkage that creates a signifcant curvature to the oligosaccharide backbone. 40 The longer sugar, with its folded backbone, likely experienced some steric repulsion with the MINP receptor, which was covered with a layer of hydrophilic ligands and averaged $5 nm in size according to dynamic light scattering (Fig. S8-10 †) and transmission electron microscopy (Fig. S4 †). Templates p-5b and p-5c were similar to p-5a, except that their glycan was maltose and maltotriose, respectively. Similar hydrolysis of the MINPs in 6 M HCl afforded MINP-CHO(p-G2) and MINP-CHO(p-G2). All three MINP-CHO's could bind glucose (G1), maltose (G2), and maltotriose (G3). Notably, among the three sugar guests, the MINP always bound its templating sugar most strongly (Table 1), consistent with successful imprinting. The nonimprinted nanoparticles (NINPs) showed very weak binding, not measurable by ITC. The imprinting factor, defned as the MINP/NINP binding ratio, was >170 for glucose. It is good that the binding of MINP-CHO(p-G1) for glucose in aqueous buffer (K a ¼ 8.85 10 3 M 1 ) already approached those for monosaccharides by natural lectins (K a ¼ 10 3 to 10 4 M 1 ). 5,41 The strong binding suggests that the polyhydroxylated surface ligand 3 could not fold back to interact with the boroxole group in the imprinted site (to interfere with the guest binding). On the other hand, the stronger binding of MINP-CHO(p-G1) for glucose than the longer sugars is a potential problem in catalytic hydrolysis, as glucose is the expected product from the hydrolysis and product inhibition would be inevitable (vide infra). ## Installation of acidic functionality for catalytic hydrolysis In the initial stage of catalysis, we used maltose (G2) as a model oligosaccharide and studied its hydrolysis by the acidfunctionalized MINP(5a + 7a-f). It is extremely encouraging that these MINPs could hydrolyze maltose simply in hot water, without any additives (Table 2). Control experiments (entries 14-16) indicated that neither the nanoparticle itself (i.e., NINP) Scheme 2 The preparation of the artificial glycosidase MINP(p-G1 + 7d) and its binding of maltose. a The FM/template ratio in MINP synthesis was 1 : 1. The cross-linkable surfactants were a 3 : 2 mixture of 1b and 2. The titrations were performed in 10 mM HEPES buffer at pH 7.4 at 298 K. b N is the average number of binding sites per nanoparticle measured via ITC curve ftting. c Nonimprinted nanoparticles (NINPs) were prepared with the same amount of FM 4 as all the MINPs but without any template. d Binding was extremely weak; because the binding constant was estimated from ITC, DG and N are not listed. nor the amino acid used for post-modifcation (e.g., 7d) was able to catalyze the hydrolysis. To optimize the catalytic design, we varied the position of the imine bond on the phenyl ring in the template. By using the ortho, meta, or para derivative, we could change the position of the to-be-installed acid group relative to the glycan. Our initial hypothesis was that the ortho template might allow the acid group to be particularly close to the exocyclic glycosidic oxygen of the glycan substrate to be bound (Scheme 2). Hydrolysis of maltose, however, increased steadily from the ortho-to the meta-and then to the para-derived MINP, from 18 to 32% yield (Table 2, entries 1-3). Molecular imprinting and guest-binding in MINP have a strong driving force in water from hydrophobic interactions. 23 During imprinting, the free hydroxyls of o-, m-, or p-6a need to stay close to the surface of the micelle (to be solvated by water) but the aglycon and the aryl group of the boroxole prefer to stay inside the micelle due to their hydrophobicity. It is possible that these requirements were best met in the para derivative, given the geometrical constraints set by the nanodimensioned micelle. When the ortho and meta amino acid were used in the post-modifcation of MINP(p-G1), hydrolysis of maltose became less efficient (Table 2, entries 4 & 5). The reductive amination protocol was established previously 37 and imine formation was found to correlate with the binding of the amine in the imprinted site. 42 The lower yields in entries 4 & 5 could come from a less complete post-modifcation (due to a mismatch in the dimensions of the reactants with the imprinted sites) and/or poor positioning of the acid in the active site. Surfactant 1b equips the MINP with a layer of hydrogenbonding amides near the surface, within the hydrophobic core of the micelle. 43 Hydrogen bonds are weakened by competition from water in aqueous solution. They are known, however, to be much stronger inside the hydrophobic microenvironment of a micelle. 35,44 Indeed, switching the surfactant from 1a to 1b in the MINP preparation increased the yield of maltose hydrolysis from 32% to 54%, suggesting that hydrogen bonds played important roles in the binding of the substrate. We also evaluated linear amino acids 7b-f in the post-modifcation, reasoning that flexibility of the acidic group might be benefcial to the catalysis. The hypothesis was confrmed by the catalysis. Although MINP(p-G1 + 7b) was very inactive, MINP(p-G1 + 7c/d) gave a much higher yield (70%) in maltose hydrolysis. Accurate positioning of the acidic group was clearly key to the hydrolysis, as too short or too long a spacer in the amino acid diminished the yield (Table 2, entries 7, 10, and 11). Note that, although the MINP active site contained both an acid and a basic amino group, the acid remained active. Similar situations are frequently found in enzyme active sites, where acidic and basic groups co-exist but do not "self-destruct" through intramolecular acid-base reaction, because the resulting ionic species are poorly solvated and thus unstable in a hydrophobic microenvironment. Precise hydrolysis of oligosaccharides With the structure of the template and the distance between the acid and glycan optimized, we prepared a series of other MINPs using p-5b-c as the templates, following similar procedures illustrated in Scheme 2. Since the corresponding MINPs were expected to bind glucose, maltose, and maltotriose, respectively, our goal was to hydrolyze maltohexaose (G6) and even a glucose-based polymer in a controlled fashion (Scheme 3). Table 3 shows the hydrolysis of G6 in water under our standard conditions (60 C for 24 h). We also studied the hydrolysis in buffer at pH 5.5-7.5 and found that the reaction yield in water was close to the highest yield obtained at pH 6 (Table S2 †). In this set of experiments, we varied the acidity of the acid catalyst, using 7d, 7g, and 7h (Scheme 3), respectively, in the post-modifcation. To our delight, the yield of the intended products-i.e., glucose (G1) from MINP(G1), maltose (G2) from MINP(G2), and maltotriose (G3) from MINP(G3)-correlated positively with the increased acidity. Meanwhile, the yields of the unintended hydrolytic products were quite random (and low), suggesting that these products probably came from a Reactions were performed in duplicate with 0.2 mM maltose and 20 mM MINP in 1.0 mL of water. Yields were determined via LC-MS using calibration curves generated from authentic samples (Fig. S32). b NINP is a nonimprinted nanoparticle prepared without any template or post-modifcation. ## Scheme 3 The selective hydrolysis of maltohexaose (G6) by MINPs. uncontrolled hydrolysis. Positive correlation between the acidity and the hydrolysis was also observed for maltose in Table 2 (entries 9, 12, and 13). Fig. 1a shows the full characterization of the hydrolyzed products by MINP(p-G1 + 7h), including the starting material (G6) and all the possible hydrolyzed fragments (G1-G5). The theoretical yield of glucose was 600 mM from 100 mM G6. The yield of glucose increased from 45% (24 h at 60 C) to 77% (48 h at 90 C) and fnally to 86% (96 h at 90 C). Most importantly, when the three MINPs were used in the hydrolysis, the dominant product was always the intended one, in a yield of 77%, 82%, and 88% for glucose, maltose, and maltotriose, respectively (Fig. 1b). Another interesting observation was the noticeable "absence" of intermediate products (G2-G5) when G6 was hydrolyzed by MINP(p-G1). 48 Whether at moderate ($50%) or higher (>80%) conversion, these intermediate products were insignifcant in comparison to G1. These results were in line with the observation that the binding of MINP-CHO(p-G1) decreased with increasing chain length of the sugar (i.e., glucose > maltose > maltotriose > maltohexaose, Table 1). Since hydrolysis requires the binding of the sugar by the MINP, the shorter fragments, with a stronger binding for the MINP catalyst, would be hydrolyzed preferentially over the starting material. Essentially, once the hydrolysis starts on a long sugar, it prefers to go all the way to break down the sugar (although not necessarily of the same molecule). An inevitable result from the stronger binding of the shorter sugars, unfortunately, was product inhibition. Fig. 1a shows that, even at 90 C for 96 h, MINP(p-G1 + 7h) could not hydrolyze maltohexaose completely. To better understand this behavior, we frst measured the Michaelis-Menten parameters for all three MINPs (Table 4). For MINP(p-G1 + 7h), we also performed the study with both maltose and maltohexaose as the substrate. The K m value of MINP(p-G1 + 7h) for maltose was about half of that for maltohexaose, indicating that the catalyst bound maltose more strongly. Interestingly, the catalytic turnover (k cat ) for maltose was also double that for maltohexaose. The catalytic efficiency (k cat /K m ) of the MINP for maltose was thus more than 4 times higher than that for maltohexaose. These results supported our above explanation for the "absence" of intermediate products in the hydrolysis of G6 by MINP(p-G1 + 7h). For the hydrolysis of G6, the three MINP catalysts showed similar k cat but stronger binding for the substrate as the active site was designed to bind a longer sugar-i.e., K m decreased in the order of MINP(p-G1 + 7h) > MINP(p-G2 + 7h) > MINP(p-G3 + 7h). The trend was similar to what was observed in the ITCdetermined binding constants for the corresponding sugars (Table 1, entries 1, 6, and 10). Both should be derived from a larger binding interface of a longer sugar with its complementary imprinted binding site. Not only could the substrate a MINPs were prepared with surfactants 1b and 2. Reactions were performed with 0.1 mM maltohexaose (G6) and 20 mM MINP in 1.0 mL of water. Yields were determined via LC-MS using calibration curves generated from authentic samples (Fig. S32). S4 † for the exact numbers. form more hydrogen bonds with the amides of (cross-linked) 1b in the MINP, but more water molecules in the active site would also be released during binding. 49 We then performed the Michaelis-Menten study for the hydrolysis of maltose by MINP(p-G1 + 7h), with different amounts of glucose added to the reaction mixture. The inhibition constant (K i ) was found to be $68 mM (Fig. S27 and S28 †), which was 5-10 times smaller than the K m value for maltose and maltohexaose. Thus, strong product inhibition indeed was present. Fortunately, with MINP being much larger than the sugars and the staring material (G6) also larger than the desired products (G1, G2, and G3), we could overcome product inhibition simply by performing the hydrolysis inside a dialysis membrane that was permeable to the desired product but impermeable to the starting material and the catalyst. In this way, the starting material and the MINP catalyst would stay inside the membrane during hydrolysis, and the product would escape into the bulk solution. This simple change in reaction setup then could turn the adversity into an advantage because product inhibition would no longer be a problem when the concentration of the product inside the membrane was diluted $40 times under our dialysis condition. Not only that, the product would be isolated in situ from the starting material and the catalyst, greatly simplifying the purifcation of the product and reuse of the catalyst (vide infra). To test this hypothesis, we chose dialysis tubing with a MWcutoff (MWCO) of 500 Da, which should let G1 (MW 180) and G2 (MW 342) easily escape but might have difficulty with G3 (MW 504). Indeed, hydrolysis of maltohexaose into glucose, maltose, and even maltotriose all improved signifcantly with the catalysis performed inside the dialysis membrane. The improvements can be seen by comparing the solid and dashed lines in Fig. 2a-c. At the end of 24 h, the yields of the desired product went from 43% to 87% for glucose (G1), 39% to 89% for maltose (G2), and 49% to 72% for maltotriose (G3). The stronger bene-fts of dialysis on G1 and G2 over G3 supported our experimental hypothesis, since G3 (MW 504) was very close to the MWCO of the membrane. Fig. 2d compares the formation of the desired products with and without dialysis. Fig. 2e shows the LC-MS analyses of the reaction mixtures with the three MINP catalysts. The data indicate that the desired sugar could always be produced as the major product, with the yield increased substantially with dialysis. ## Controlled hydrolysis of polysaccharides and recyclability of the catalysts A rich source of polysaccharides is found in nature. Their precise cleavage based on our selective one-step hydrolysis can be a convenient and economical way to produce glycans that otherwise require complex multistep synthesis and extensive protective/deprotective chemistry. 5 Gratifyingly, not only could these MINP glycosidases hydrolyze maltohexaose in a highly controlled fashion, but they could also hydrolyze amylose, a polysaccharide of glucose connected by the same 1,4-aglycosidic linkage, with equally good selectivity (Fig. 3a). The hydrolysis once again happened inside the dialysis membrane, with the polysaccharide and MINP catalysts trapped inside and the product released into the bulk solution. Another beneft of performing the hydrolysis inside a dialysis membrane was the facile recycling of the catalyst. As highly cross-linked polymeric nanoparticles, our MINP-based artifcial glycosidase could be reused many times without any concerns regarding loss of activity when maltohexaose was repeatedly added into the dialysis tubing that contained MINP(p-G2 + 7h) (Fig. 3b). ## Substrate selectivity Substrate selectivity is one of the most important performance criteria for a synthetic glycosidase, since different building blocks, connection sites, and spatial orientations of the glycosidic linkage can influence the property of a glycan profoundly. Table 5 shows the hydrolysis of a number of oligosaccharides by our MINP catalysts. The yields were for hydrolysis at 60 C after 24 h and the binding constants were for the same MINP determined by ITC at 25 C. Consistent with the binding-derived catalysis, there was an overall correlation between the hydrolytic yields and the K a values. For example, among the disaccharides, maltose gave the best yield with MINP(p-G1 + 7h) and its binding was also the strongest. Xylobiose was completely inactive and its binding was also the weakest. For the sugars with intermediate binding constants (cellobiose, sucrose, and maltulose), the correlation was weak. Another conclusion from the hydrolyses was the importance of the boronate ester formation to the substrate selectivity. MINP(p-G1 + 7h) was designed to bind the terminal glucose of a suitable oligo-or polysaccharide at the non-reducing end (Scheme 2). Thus, it was not a surprise that cellobiose, sucrose, and maltulose could all be hydrolyzed by this MINP. The reducing sugar residue on these molecules is expected to reside in water, outside the active site. For the same reason, MINP(p-G1 + 7h) should NOT be particularly selective for the reducing sugar, whether in its chemical structure or spatial orientation. Meanwhile, MINP(p-G1 + 7h) should be much more selective toward the sugar at the nonreducing end, especially if the hydroxyl involved in boronate formation is altered. Lactose, with a galactose at the non-reducing end, indeed gave a very Table 5 The hydrolysis of oligosaccharides by MINP catalysts a a The hydrolysis experiments were performed at 60 C in water for 24 h, with [oligosaccharide] ¼ 0.2 mM and [MINP] ¼ 20 mM. Yields were determined by LC-MS using calibration curves generated from authentic samples (Fig. S32). Isothermal titration calorimetry (ITC) data are reported in Table S1. poor hydrolytic yield and binding constant, because the C4 hydroxyl was involved in boronate formation (Scheme 2). It is quite impressive that inversion of a single hydroxyl decreased the yield of hydrolysis from 67% for cellobiose to 17% for lactose. Xylobiose is missing the hydroxymethyl from cellobiose. Its inactivity indicates that the C6 hydroxyl was also essential to the binding. For a monosaccharide-derived catalyst such as MINP(p-G1 + 7h), its only selectivity was in the terminal sugar at the non-reducing end and the a/b selectivity was low. For a disaccharide-derived catalyst, the situation was different because the a/b linkage between the frst two sugar residues would affect the binding of the substrate. MINP(p-G1+7h) can hydrolyze maltotriose and cellotriose into glucose. Table 5 shows that the yield of glucose was 71% and 54% from the two trisaccharides, respectively. The a/ b selectivity (1.3 : 1) was slightly higher than that observed in maltose/cellobiose (1.2 : 1), possibly because two hydrolyses were needed to hydrolyze the trisaccharides, but only one for the disaccharides, which magnifed the a/b selectivity. When MINP(p-G2 + 7h) was used, however, the yield for the desired (disaccharide) product was 85% from maltotriose and only 24% from cellotriose. This was because the imprinted site was designed to bind maltose in this catalyst (Scheme 2). Thus, the b glycosidic bond in between the frst and second sugar from the nonreducing end of cellotriose would weaken the binding of this substrate. ## Conclusions Micellar imprinting provided a rational method for constructing robust synthetic glycosidases from readily synthesized small-molecule templates. The natural glucan 1,4-alphaglucosidase removes one glucose residue at a time from the nonreducing end of amylose, 50 and beta-amylase removes two glucose residues (i.e., maltose) at a time. 51 Our synthetic glycosidases not only duplicated the selectivities of these enzymes but also had selectivity not available from natural biocatalystsi.e., the selective formation of maltotriose from maltohexaose or amylose. Substrate selectivity was mainly determined by the sugar residues bound within the active site, including their spatial orientations. As cross-linked polymeric nanoparticles, the MINPs tolerate high temperature, 23,42 organic solvents, 42 and extreme pH, 37 outperforming natural enzymes completely in these aspects. Importantly, the design of our synthetic glycosidase is general, using molecular imprinting to create a glycan-specifc active site, followed by post-modifcation to install an acidic group right next to the glycosidic bond to be cleaved. Similar designs should be applicable to complex glycans. 52 The total synthesis of carbohydrates is often extremely challenging. 5 Selective one-step hydrolysis using a rationally designed synthetic glycosidase potentially can be a powerful method to produce complex glycans from precursor oligosaccharides or polysaccharides that are either naturally available or prepared through enzymatic synthesis. The facile separation of the products by dialysis demonstrated in this work, the excellent reusability of the MINP catalysts, and the simplicity of the hydrolysis requiring only hot water are attractive features for such purposes, and can open up new avenues in glycoscience and technology. ## Hydrolysis of Oligosaccharides and Polysaccharides Hydrolysis experiments without dialysis were carried out as follows. In general, a 200 mL aliquot of a 100 mM MINP stock solution in Millipore water was diluted by water or a 10 mM MES buffer (pH 6.0) to 990 mL and sonicated for 0.5 min. To this solution, a 10 mL aliquot of a 10/20 mM oligosaccharide stock solution was added. The reaction mixture was allowed to react in a Benchmark heating block at 60 or 90 C for the indicated time. The reaction mixture was centrifuged (20 000 RPM for 10 min) to remove the MINP catalyst before LC-MS analysis using calibration curves generated from authentic samples (Fig. S32 †). Hydrolysis experiments with dialysis were carried out as follows. In general, a 200 mL aliquot of a 100 mM MINP catalyst in Millipore water was diluted with Millipore water to 990 mL and sonicated for 0.5 min, and then the solution was added to a dialysis tubing (MWCO 500), followed by the addition of a 10 mL aliquot of a 10 mM maltohexaose stock solution (or 1 mg of amylose). The reaction mixture was dialyzed against 40 mL of Millipore water at 60 C. The hydrolysis was monitored by LC-MS analysis of the external solution using calibration curves generated from authentic samples (Fig. S32 †). ## Conflicts of interest Iowa State University Research Foundation has fled a patent application relating to the technology.

chemsum

{"title": "Synthetic glycosidases for the precise hydrolysis of oligosaccharides and polysaccharides", "journal": "Royal Society of Chemistry (RSC)"}

solid-state_¹⁷o_nmr_study_of_α-<scp>d</scp>-glucose:_exploring_new_frontiers_in_isotopic_

## Abstract: Solid-state 17 O NMR study of α-D-glucose: exploring new frontiers in isotopic labeling, sensitivity enhancement, and NMR crystallography Synthesis of site-specifi cally 17 O-labeled α-D-glucose was reported. Complete solid-state 17 O NMR characterization was achieved with advanced NMR technologies such as an ultrahigh magnetic fi eld of 35.2 T and CPMAS CryoProbe, which made it possible to obtain solid-state 17 O NMR data with unprecedented sensitivity and resolution. Quantum chemical computation was used to aid the interpretation of experimental data. The results revealed remarkable sensitivity of 17 O NMR parameters to hydrogen bonding. rsc.li/chemical-science Solid-state 17 O NMR study of a-D-glucose: exploring new frontiers in isotopic labeling, sensitivity enhancement, and NMR crystallography † ## Introduction The element of oxygen is a key constituent of organic and biological molecules. Oxygen-containing functional groups are often directly involved in chemical reactions including biological transformation such as enzyme catalysis. While NMR spectroscopy is a powerful technique for structural elucidation of organic and biological molecules, most NMR studies are based on detection of signals from hydrogen, carbon, nitrogen, and phosphorus atoms. While it is highly desirable to add oxygen to the list of nuclear probes available for NMR studies, two major obstacles have made it difficult to characterize NMR signals from oxygen atoms. First, the NMR-active oxygen isotope, 17 O, has an exceedingly low natural abundance (0.037%). Thus, it is usually necessary to prepare 17 O-enriched molecular systems in order to boost NMR detectability. This 17 O-labeling process can be a difficult task. Second, 17 O has a quadrupolar nucleus (I ¼ 5/2), which often gives rise to signifcantly broader NMR signals than those commonly encountered from other more NMR-friendly spin-1/2 nuclei such as 1 H, 13 C and 15 N. This quadrupole line broadening is a major roadblock to 17 O NMR applications in terms of spectral resolution. Over the last two decades, however, signifcant progress has been made in demonstrating 17 O NMR as a viable tool to study organic and biological molecules in both solution and the solid state. For 17 O NMR studies of biological molecules, in particular, some important developments have occurred in recent years. Zhu et al. 8 showed that it is possible to obtain solid-state 17 O NMR spectra from protein-ligand complexes where the ligand molecules are site-specifcally 17 Olabeled. Tang et al. 9 applied this approach to study hydrogenbonding interactions around the "oxyanion hole" in several acyl-enzymes. Zhu et al. 10,11 demonstrated a technique known as quadrupole-central-transition (QCT) NMR in obtaining highresolution 17 O NMR spectra for biological macromolecules undergoing slow tumbling motion in aqueous solution. Young et al. 12 applied the 17 O QCT method to monitor the formation of enzymatic intermediates of tryptophan synthase under active catalysis. Recently, Paulino et al. 13 reported a comprehensive 17 O solid-state NMR study of the water-carbonyl interactions in gramicidin A ion channel. The latest advancement in the feld was the work by Lin et al. 14 where they demonstrated a general approach to incorporate 17 O isotopes into recombinant proteins and reported solid-state 17 O NMR spectra for yeast ubiquitin. In addition to the abovementioned new applications, there have also been recent developments in solid-state 17 O NMR methodology. One particular area of interest is concerned with heteronuclear correlation solid-state NMR spectroscopy between 17 O and other nuclei such as 1 H, 13 C, and 15 N. For example, Hung et al. 19 reported a new 3D D-RINEPT/DARR OCC experiment where overlapping 17 O NMR signals can be completely separated in the 13 C dimension. Another highly promising direction is to use dynamic nuclear polarization (DNP) to enhance 17 O NMR signals for organic and biological molecules. Currently, most DNP-enhanced 17 O NMR studies were performed at low or moderate magnetic felds (#14.1 T) to study inorganic materials; it would be highly benefcial for the study of organic and biological molecules if DNP for 17 O becomes feasible at higher magnetic felds. 24 While fundamental 17 O NMR data on chemical shift (CS) and electric-feld-gradient (EFG) tensors have been reported for many oxygen-containing organic functional groups, there are still many unexplored classes of organic compounds for which little is known about their 17 O NMR tensor properties. One notable example is concerned with carbohydrates. Carbohydrates are an important class of oxygen-rich organic molecules of biological signifcance. However, solid-state 17 O NMR studies dealing with carbohydrate molecules are very rare in the literature. Sefzik et al. 25 reported the frst solid-state 17 O NMR study of several protected carbohydrate compounds. Yamada et al. 26 obtained the solid-state 17 O NMR signal for the O6 atom of Dglucosamine. More recently, Hung et al. 19 reported 2D and 3D 13 C- 17 O heteronuclear correlation solid-state NMR spectra of [1-13 C, 17 O]-a/b-D-glucose. Also relevant are two 17 O QCT NMR studies by Shen et al. 27 and by Gan et al. 28 where 17 O-labeled Dglucose samples were examined with the aid of high magnetic felds. One major challenge in solid-state 17 O NMR studies of carbohydrates is the synthesis of 17 O-labeled target compounds. To further explore synthetic procedures and solid-state 17 O NMR for unprotected carbohydrate compounds, we selected Dglucose as an initial target (Scheme 1). In this work, we report synthesis of a total of six site-specifcally 17 O-labeled D-glucose compounds and their full solid-state 17 O NMR characterization. For the latter part, because crystallization of D-glucose into a pure anomeric form (a or b) often encounters low yields, we decided to prepare all solid samples of D-glucose in the form of a D-glucose/NaCl/H 2 O (2/1/1) cocrystal. This cocrystal is known to contain exclusively a-D-glucose and can be readily prepared in crystalline form with near 100% yields. Throughout this work, we will use "a-D-glucose" as a shorthand name for the a-Dglucose/NaCl/H 2 O (2/1/1) cocrystal. Another objective of the present work is to demonstrate utilization of the current state-of-the-art solid-state 17 O NMR technologies achieving unprecedented sensitivity and spectral resolution for organic and biological molecules. To this end, we explore the following three areas. First, we perform solid-state 17 O NMR at multiple magnetic felds including an ultrahigh magnetic feld of 35.2 T. 28 Second, we investigate the effect of paramagnetic doping in shortening spin-relaxation times for 17 O nuclei so that fast data acquisition might be possible. Third, we test the sensitivity enhancement for solid-state 17 O NMR applications using a new CPMAS CryoProbe. 32 ## Synthesis of sitespecically 17 O-labeled D-glucose compounds In this work, we employed three strategies to synthesize site-specifcally 17 O-labeled D-glucose compounds; see Scheme 2. First, the anomeric O1 atom in D-glucose can be readily 17 Olabeled by a simple exchange with 17 O-water. This exchange occurs through the hydration/dehydration process of the aldehyde functional group in the open chain glucose tautomer. Second, for both primary and secondary hydroxyl groups (O2, O3, O4, O6), 17 O isotopes can be incorporated into glucose by S N 2 nucleophilic substitution (via either triflate displacement route or Mitsunobu reaction) from appropriate starting epimers. 36 In this case, either sodium [1, labeling of the O5 atom, we utilize a combined oxidation/ exchange/reduction method starting from 1,2-O-isopropylidene-D-glucofuranurono-6,3-lactone as illustrated in Scheme 2. After oxidation of the OH group by chromium trioxide, 37 17 O-labels are introduced onto the keto group from 17 O-water via an acid-catalyzed hydration/dehydration process (or keto/gem-diol exchange). Then, reduction with NaBH 4 converts the keto group back to the hydroxyl group. 38 Finally, removal of protecting groups allows the furanose/pyranose equilibrium to occur, producing [5-17 O]-D-glucose. 39 Full details of the synthetic procedures and compound characterization are provided in the ESI. † ## Preparation of solid samples As mentioned above, because crystallization of D-glucose into the pure a (or b) form is often associated with low yields, we prepared all solid samples of 17 O-labeled D-glucose as a Dglucose/NaCl/H 2 O (2/1/1) cocrystal where all D-glucose molecules are in the a-form. 29 The D-glucose/NaCl/H 2 O (2/1/1) cocrystal was readily prepared by adding solid NaCl to aqueous solution of D-glucose followed by lyophilization. A solid sample was prepared as an equal molar mixture of 17 O-enrichment, the mixing process was monitored by solution 17 O NMR; see ESI. † As a result, the level 17 O enrichment in this [3/5/6-17 O]-a-Dglucose sample was only about 10%. The integrity of all solid samples was checked by acquiring solid-state 13 C CPMAS NMR spectra; all spectra are provided in ESI. Solid samples with paramagnetic Cu(II) dopants were prepared in the following fashion. To 2 mL of H 2 O was frst added 15 mg of solid Na 2 [-Cu(EDTA) 2 ] to give a clear blue solution, followed by addition of 150 mg D-glucose/NaCl/H 2 O (2/1/1) cocrystal. The solution turned greenish when solids were fully dissolved. The solution was then dried under a stream of N 2 until it became a syrup. Addition of 2 mL of absolute ethanol induced crystallization. After removal of the supernatant, solids were dried in air. Cu(II)doped solid samples displayed the same solid-state 13 eliminate probe ringing artifacts. For MAS experiments, a 3.2 mm Bruker HX MAS probe was used where the effective 90 pulse width for the 17 O CT was 1.0 ms. For static experiments, a homebuilt 5 mm solenoid probe was used with powder samples packed into 5 mm Teflon tubes to reduce background signals. On this solenoid probe, the 90 pulse width for the 17 O CT was 2.0 ms. 1 H decoupling with 75 kHz rf feld was applied during data acquisition. A liquid H 2 O sample was used for both rf power calibration and 17 O chemical shift referencing (d ¼ 0 ppm). All spectral simulations were performed with DMft. 40 Solid-state 17 O NMR experiments at 35.2 T were carried out on the 36 T series-connected hybrid (SCH) magnet 28 at the National High Magnetic Field Laboratory (NHMFL, Tallahassee, Florida, USA) with a Bruker Avance NEO console. A singleresonance 3.2 mm MAS probe with an external feld regulation circuit designed and constructed at the NHMFL was used with pencil-type ZrO 2 rotors spinning at a MAS frequency of 16 kHz. A Hahn-echo sequence was used with 5 and 10 ms pulses (with 16.7 kHz rf feld) and a recycle delay of 0.1 s. Solid-state 17 O and 13 C NMR experiments were also performed on a Bruker NEO-800 (18.8 T) at the Bruker application lab (Fällanden, Switzerland) with a broadband 3.2 mm CPMAS CryoProbe. The sample spinning was 15 kHz. The 17 O rf feld was about 64 kHz, which gave an effective 90 pulse of 1.3 ms for the CT. The 1 H decoupling feld was 83 kHz. An apodization weighted sampling (AWS) scheme 41 was used for collecting 2D 17 O shifted-echo 3QMAS data. For the 13 C refocused INADE-QUATE experiment, the 13 C 90 pulse was 5.0 ms. The spectral width in the F 1 dimension was 7.5 kHz. A frequency swept TPPM 1 H decoupling (83 kHz) scheme was applied during data acquisition. ## Computational details All quantum chemical calculations were performed using the CASTEP code 42 (version 2019) together with BIOVIA's Materials Studio. CASTEP employs DFT using the plane-wave pseudopotential approach. The generalized gradient approximation with either the Perdew-Burke-Ernzerhof 43 or revised Perdew-Burke-Ernzerhof (rPBE) 44 exchange correlation functionals was chosen. First, geometry optimization was performed employing the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm together with OTFG on-the-fly ultrasoft pseudopotentials (version 2017R2), a cut-off energy of 598.7 eV and a k-point grid with a maximum separation of 0.071 1 . We also tested the treatment of dispersion interactions by using the two-body force-feld method of Grimme (D2) (ref. 45) with a reparameterized damping function (s6 ¼ 1.0; d ¼ 3.25 or d ¼ 5.0) 46,47 in geometry optimizations. Subsequently, the NMR parameters were calculated using the Gauge Including Projector Augmented Waves (GIPAW) method implemented in the NMR module of CASTEP. In this work, a total of four sets of GIPAW DFT computations were performed and they are denoted as: (1) PBE, (2) rPBE, (3) rPBE-D2 (d ¼ 3.25), and (4) rPBE-D2 (d ¼ 5.0). However, because these four methods produced essentially the same results, we will focus on the results obtained with the PBE method and report the complete results from all four methods in the ESI. † ## Results and discussion Determination of 17 O NMR tensors in a-D-glucose Fig. 1 shows the solid-state 17 O NMR spectra obtained for all six site-specifcally 17 O-labeled D-glucose compounds. In each 17 O MAS NMR spectrum, a well-defned powder line shape was observed, which is known to arise from the second-order quadrupole interaction. In general, second-order quadrupole interactions are inversely proportional to the applied magnetic feld. However, as seen from Fig. 1, even at 21.1 T, second-order quadrupole interactions cause a line broadening on the order of 100 ppm. This is because the oxygen-containing functional groups in D-glucose (hydroxyl and ether groups) are known to experience rather large 17 O nuclear quadrupole interactions. It is also immediately clear that the relatively small 17 O chemical shift variations among the six oxygen-containing groups in Dglucose can be easily obscured by such second-order quadrupole broadenings (vide infra). In each case, an analysis of the observed powder line shape obtained under MAS conditions allowed us to obtain three 17 O NMR parameters: d iso , C Q , and h Q . Complete experimental results are listed in Table 1. When the solid-state 17 O NMR experiments are performed for stationary (non-spinning) powder samples, even broader powder line shapes are observed, as also seen from Fig. 1. At 14.1 T, each static powder line shape spans about 700 ppm, which is reduced to roughly 300 ppm at 21.1 T. This is because now both 17 O CS and QC tensors contribute to the static powder line shape. The interplay between the two NMR tensors is responsible for the observed feld dependence of the static 17 O NMR spectra. From an analysis of these static powder line shapes, we were able to obtain the principal components of the 17 O CS tensor and their relative orientations with respect to the 17 O QC tensor. All experimental 17 O NMR tensor parameters determined for the six oxygen atoms in a-D-glucose are summarized in Table 1. In general, the values of jC Q ( 17 O)j found in a-D-glucose are about 8-10 MHz with h Q close to 1. These parameters are similar to those previously reported for protected carbohydrate compounds, 25 D-glucosamine, 26 and several other related functional groups such as hemiacetal/hemiketal, 51,52 gem-diol, 53 hydroxyl, 54 and phenolic groups. Because the six oxygen-containing functional groups in a-Dglucose are very similar, their 17 O isotropic chemical shifts, d iso ( 17 O), are found within a small range of 60 ppm. Nonetheless, there is a general trend in the observed d iso ( 17 O) values: O5 (C-O-C part of a cyclic hemiacetal) > O1 (C-OH part of a cyclic hemiacetal) > O2, O3, O4 (secondary alcohol groups) > O6 (a primary alcohol group). These agree with previous solution 17 O NMR studies 58,59 as well as with our own measurements for the 17 O-labeled D-glucose compounds in aqueous solution (see ESI †). Compared with the 17 O NMR parameters found for crystalline hydrates, the values of jC Q ( 17 O)j for the alcohol and ether groups in a-D-glucose are somewhat larger, but the spans of the 17 O CS tensors are comparable. To aid the interpretation of experimentally determined 17 O NMR tensor parameters, we performed extensive GIPAW DFT computations. As mentioned earlier, the choice of making Dglucose/NaCl/H 2 O cocrystal for solid-state 17 O NMR experiments was based on the considerations for having a pure anomeric form and easy preparation of crystalline samples. Now this turns into a computational challenge, because the Dglucose/NaCl/H 2 O cocrystal has a very large unit cell (trigonal space group P3 1 , a ¼ 16.836 , c ¼ 17.013 , V ¼ 4176 3 , Z ¼ 9) that contains six crystallographically distinct glucose molecules in the asymmetric unit. 29 Careful examination of the crystal structure reveals that the six crystallographically unique Dglucose molecules form three "dimers" via Na + chelation with the O1 and O2 atoms, as depicted in Fig. 2. Furthermore, the asymmetric unit contains three water molecules, each involved in hydrogen bonding with both two symmetry-related D-glucose molecules and one Cl ion. In the original crystal structure, 29 one of the water molecules was missing a single hydrogen atom, which was added back into the structure before the geometry was optimized using DFT. As a result, all three water molecules have a similar hydrogen-bonding environment (see Fig. S6 in the ESI †). The GIPAW DFT results obtained with the PBE method for 17 O NMR parameters are listed in Table 2; complete GIPAW DFT results from all four methods are given in the ESI. † It can be seen from Table 2 that all six crystallographically independent D-glucose molecules have similar 17 O NMR parameters (vide infra). Thus, within the spectral resolution limit of the 1D 17 O MAS spectra, we can assume just one 17 O NMR signal for each oxygen position. For this reason, Fig. 3 shows comparison between experimental 17 O CS tensor parameters and "averaged" GIPAW DFT results (averaged over the six crystallographically independent glucose molecules in the asymmetric unit). Because the 17 O chemical shift anisotropies are rather small in glucose, the agreement seen in Fig. 3 is clearly satisfactory. Since the 17 O QC tensor parameters do not show much variation, we will not examine them further, except to note that the GIPAW DFT calculations are consistent with the experimental results. GIPAW DFT computations also yielded further information about the 17 O NMR tensor orientations in the molecular frame. In Fig. 4, we used TensorView 65 to display the ovaloid representation of the 17 O CS and QC tensors for the six oxygen sites in a-D-glucose. Two general types of orientation were found for the 17 O CS tensors, as seen in Fig. 4(a). For O1, O2, and O3, the 29 orientations in the molecular frame, the two seemingly different orientations are essentially the same. This is because the largest QC tensor or EFG tensor component (V zz ) is defned according to its absolute value so that jV zz j $ jV yy j $ jV ## Solid-state 17 O NMR at high magnetic elds One of the major challenges in solid-state 17 O NMR studies of carbohydrate compounds is that all oxygen-containing functional groups are either hydroxyl or ether groups. As a result, they exhibit very similar 17 O NMR parameters. For example, the 17 O isotropic chemical shifts for the six oxygen sites in a-Dglucose, given in Table 1, are within a narrow range of 60 ppm. If multiple oxygen sites are simultaneously 17 O-labeled, it could be very difficult to resolve their 17 O NMR signals because each signal would be signifcantly broadened by the second-order quadrupole interaction. Since the second-order quadrupole interaction is inversely proportional to the applied magnetic feld, it is often advantageous to perform solid-state 17 O NMR experiments at the highest possible magnetic feld. To test the limit of this brute-force approach, we obtained 17 Because the cross-relaxation between CSA and second-order quadrupole interactions becomes more important at high magnetic felds, 17 O QCT spectra will display higher resolution for carbohydrates (with small CSAs) than for proteins (with large CSAs). 67 ## Combination of paramagnetic doping and CPMAS CryoProbe technology Crystalline D-glucose is known for its exceedingly long T 1 ( 1 H) values. It was observed that T 1 ( 17 O) values are also long for Dglucose compounds, hindering rapid repetition of 17 O data acquisition. One common approach that has been widely employed in cross polarization (CP)-based solid-state 13 C NMR studies is to add paramagnetic Cu(II) dopants to shorten T 1 ( 1 H). In this work, we hypothesized that the same paramagnetic doping approach might be useful for 17 further increase sensitivity, we combined the paramagnetic doping with new CPMAS CryoProbe technology. It has recently been shown that a CPMAS CryoProbe provides a 3-4 times higher sensitivity for detecting 13 C and 15 N nuclei compared to a conventional MAS probe. 32 After the submission of this work, we learned that Michaelis and co-workers 73 also obtained some preliminary solid-state 17 O NMR data using the CPMAS Cryo-Probe. For acquiring 17 O MAS spectra for the a-D-glucose compounds, we found that the combination of paramagnetic doping and CPMAS CryoProbe yielded a sensitivity gain by a factor of 6-8. Fig. 7 shows the 2D 17 O 3QMAS spectra obtained for [2- 17 O]-a-D-glucose and [3/5/6-17 O]-a-D-glucose samples doped with Cu-EDTA. This is the frst time that 2D 17 O 3QMAS spectra are reported for carbohydrate compounds. It should be emphasized that the level of 17 O enrichment in the [3/5/6-17 O]-a-D-glucose sample was only about 10%. Thus, the observed sensitivity shown in Fig. 7 is quite remarkable. Interestingly, whereas each of the O3 and O6 signals appears to split into two signals, no signal splitting was observed for the O2 and O5 signals (vide infra). We were able to ft the F2-slice spectra and obtained the following 17 O NMR parameters: O2, MHz, h Q ¼ 0.9; O5, d iso ¼ 56 ppm, C Q ¼ 9.9 MHz, h Q ¼ 1.0. These values are also confrmed by the signal positions in the isotropic dimension of the 17 O 3QMAS spectrum; see ESI. † As expected, the 17 O NMR parameters for O2 and O5 are identical to those extracted from 1D MAS spectra as listed in Table 1. For O3 and O6, in contrast, the unprecedented spectral resolution offered by 2D 17 O 3QMAS spectra revealed fner spectral details. We will further discuss these new details in the next section. Further 17 O and 13 C NMR signal assignments As mentioned earlier, there are six crystallographically independent glucose molecules in the asymmetric unit of D-glucose/ NaCl/H 2 O cocrystal. Thus, in principle, there should be six 17 O NMR signals for each oxygen atom in this compound. However, the six crystallographically independent glucose molecules form three Na + -chelated glucose "dimers" with very similar structures. For this reason, the two different signals observed for each of the O3 and O6 groups in the 2D 17 O 3QMAS spectrum shown in Fig. 7 can be attributed to the two types of a-Dglucose molecules, A and B, within each Na + -chelated glucose "dimer". This also implies that the difference among the three "dimers" cannot be detected with the current spectral resolution. The tentative signal assignments shown in Fig. 7 were based on the GIPAW DFT calculations listed in Table 2. To further confrm this hypothesis, we decided to fully assign the solid-state 13 C NMR signals for the same a-D-glucose sample. To this end, we obtained a 2D refocused INADEQUATE NMR spectrum at the 13 C natural-abundance isotope level for the same compound using the CPMAS CryoProbe. As seen from Fig. 8, a similar signal "doubling" was indeed observed for each carbon atom. Fig. 8 also shows the 13 C NMR signal assignment for Molecules A and B, based on GIPAW DFT results for 13 C chemical shifts (provided in the ESI). In fact, in the 1D 13 C CPMAS spectrum shown in Fig. 8, there are also hints that smaller resonance splittings beyond the signal "doubling" are also present for C1, C2A, C3, C4, and C6B. Unfortunately, within the currently achievable spectral resolution, it is not possible to resolve all six 13 C NMR signals for each site. So, for now we focus on the chemical shift differences between Molecules A and B within the glucose "dimer". Clearly, for different carbon sites, the 13 C chemical shift differences between Molecules A and B show different patterns. We will further examine these patterns for all the carbon and oxygen atoms in a-D-glucose. Fig. 9 shows a comparison between experimental and GIPAW DFT results with the PBE method for both 13 C and 17 O chemical shifts; complete GIPAW DFT results from all four methods are provided in the ESI. † The observed general agreement between experiment and computation suggests that the reported signal assignment is quite reasonable. Now we can understand why no "doubling" or "splitting" was observed for the O2 and O5 signals in the 17 O 3QMAS spectra shown in Fig. 7. As seen from Fig. 9, the GIPAW DFT calculations predict that the 17 O chemical shift difference between Molecules A and B is indeed rather small for O2 and O5 (<3 ppm). It is also evident from Fig. 9 that the 17 O chemical shift is a much more sensitive probe than the 13 C chemical shift to any structural variation. In practice, however, the generally lower spectral resolution encountered in 17 O NMR often makes it difficult to fully utilize such sensitivity. On the other hand, it is also not difficult to imagine that, in some cases, the superior sensitivity of 17 O NMR to molecular structure and chemical bonding can produce information that is unobtainable by 13 C NMR. Ideally, one should utilize all available magnetically-active nuclei in a molecular system as a general approach of "NMR crystallography". 74 Now, what are the reasons for the 17 O chemical shift differences between Molecules A and B to show the patterns displayed in Fig. 9? Why do the O2 and O5 atoms between Molecules A and B exhibit very similar 17 O chemical shifts (within 2 ppm), but the O3 and O6 atoms have so different values (by more than 10 ppm)? To link the structural features to these spectral characteristics, we will need to further examine the crystal structure of the D-glucose/NaCl/H 2 O cocrystal. 17 O NMR parameters to hydrogen bonding interactions. Interestingly, the GIPAW DFT calculations showed that the protons attached to O3B and O6B are also signifcantly deshielded by 2-3 ppm, due to the stronger hydrogen bonding, than the corresponding protons attached to O3A and O6A. ## Conclusions We have carried out a comprehensive solid-state 17 O NMR study for a-D-glucose. In this work, a total of six site-specifcally 17 Olabeled a-D-glucose compounds were synthesized. The 17 O CS and QC tensors were determined for each of the six oxygen sites in a-D-glucose from an analysis of solid-state 17 O NMR spectra obtained at multiple magnetic felds. This is the frst case where all oxygen-containing functional groups in a carbohydrate molecule are site-specifcally 17 O-labeled and have their 17 O NMR tensors fully characterized. We found that paramagnetic Cu(II) doping can signifcantly shorten the T 1 ( 17 O) values for solid a-D-glucose samples, making it possible to rapidly collect 17 O NMR data. By combining the paramagnetic doping effect with the new CPMAS CryoProbe technology and apodization weighted sampling at high magnetic felds, we have achieved a signifcant sensitivity boost that allowed us to obtain the frst set of 17 O 3QMAS spectra ever reported for carbohydrate compounds. The unprecedented resolution offered by 2D 17 O 3QMAS spectra permitted the detection of a subtle structural difference for a single hydrogen bond between two types of crystallographically distinct D-glucose molecules. With the aid of GIPAW DFT calculations, all observed 17 O and 13 C NMR signals were assigned to the two groups of crystallographically distinct a-D-glucose molecules. This combined 17 O and 13 C solid-state NMR approach adds a new dimension to the feld of "NMR crystallography". Successful synthesis of site-specifcally 17 O-labeled D-glucose also paves the way for researchers to consider 17 O NMR as a new spectroscopic tool in glucose-related research, which can range from glucose binding proteins to glucose metabolism of live cells. In a broader context, this work demonstrates that continuing advancement of solid-state 17 O NMR spectroscopy has begun to open the door for studying many biological molecules that are usually considered too difficult for 17 O NMR spectroscopy. It is about time to add 17 O to the NMR toolbox for probing organic and biological molecules.

chemsum

{"title": "Solid-state ¹⁷O NMR study of \u03b1-<scp>d</scp>-glucose: exploring new frontiers in isotopic labeling, sensitivity enhancement, and NMR crystallography", "journal": "Royal Society of Chemistry (RSC)"}

tautomerism_as_primary_signaling_mechanism_in_metal_sensing:_the_case_of_amide_group

## Abstract: The concept for sensing systems using the tautomerism as elementary signaling process has been further developed by synthesizing a ligand containing 4-(phenyldiazenyl)naphthalene-1-ol as a tautomeric block and an amide group as metal capturing antenna. Although it has been expected that the intramolecular hydrogen bonding (between the tautomeric hydroxy group and the nitrogen atom from the amide group) could stabilize the pure enol form in some solvents, the keto tautomer is also observed. This is a result from the formation of intramolecular associates in some solvents. Strong bathochromic and hyperchromic effects in the visible spectra accompany the 1:1 formation of complexes with some alkaline earth metal ions. ## Introduction The design of new organic sensing systems is an undividable part of the development of coordination chemistry . Particularly chromophore ligands have been successfully utilized for colorimetric detection of the majority of metal ions as complex . Some of them are used as standard tools in chelatometric titrations . The design of specific ligands for alkali metal determination is still a challenge. In the case of alkaline earth metal ions, the reagents with reasonable selectivity are still not commonly accepted since they compete with transitional metal ions . The discovery of crown ethers and 3D-based ligands unquestionably helped the development of natural ligand-supported metal investigations. The ion recognition is based on the existence of two molecular states (ligand and complex) with different optical properties and a structure that allows fast transfer from the ligand to the complex upon addition of the desired metal ion . The tautomeric proton exchange has the same properties when the equilibrium is switched from one to the other tautomer. The tautomerism can be controlled by metal ion addition, when an ionophore unit Scheme 1: Conceptual idea for tautomeric metal sensing. is implemented in the tautomeric backbone. The conceptual idea to achieve the pure enol tautomer through intramolecular hydrogen bonding with the ionophore is shown in Scheme 1. The complex formation ejects the tautomeric proton and stabilizes the keto tautomer. Several successful tautomeric ligands, based on 4-(phenyldiazenyl)naphthalen-1-ol (1) (2 and 3, Scheme 2) as a tautomeric unit have been developed by us. We found that compounds 2 and 3 exist in the neutral state solely as enol tautomers due to intramolecular hydrogen bonding involving the tautomeric hydroxy group and that the complexation shifts the equilibrium to the K form. Although 3 exhibits a 3D structure and as a result, shows high stability constants upon complexation, the selectivity is rather low, which can be attributed to the crown ether complexation features in general. Developing the system further, leads to modification of the ionophore part by replacing the crown ether with other ionophores, such as done in the case of 4 and 5. The quantum-chemical calculations for 4 and 5 have demonstrated that the stable enol tautomers exist as intramolecular C=O•••HO bonded system, while in the K forms the ionophore part does not participate in hydrogen bonding and can be considered as a basic 2-alkyl substitution . Consequently, the stabilization between the E and K forms is a result of the competition between the strength of the hydrogen bonding in the enol tautomer and the effect of simple alkyl substitution in the keto form skeleton. The calculations also suggest that the efficient switching towards the enol form can be achieved only when R' = NMe 2 (Scheme 2). Theoretical modelling of structures 4 and 5 have also shown that only one of the carbonyl groups from the ionophore unit really participates in the capturing of the metal ion upon complexation. Therefore, the aim of the current article is to estimate theoretically and experimentally, the tautomeric state and complexation abilities of compound 6, where only one carbonyl group in the ionophore part is present (Scheme 2). It is expected that the enol tautomer stabilization would be achieved in the neutral state as a result of the strong intramolecular hydro- gen bonding between the tautomeric OH group and the carbonyl group in the ionophore part. The complex formation, depending on the size and charge of the metal ion, should shift the tautomeric equilibrium towards the keto tautomer and should provide stabilization of the complex. To the best of our knowledge, such a system has not been synthesized and studied up to now. ## Results and Discussion Compound 1 is a well-studied tautomeric structure featuring a moderate energy gap between the enol and the keto tautomeric forms . For this reason, the tautomeric equilibrium can be easily affected by changing the solvent. However, the tautomeric equilibrium has not been switched fully to either of the tautomers in solution. For instance, the experimentally determined ΔG values at room temperature range from 1.42 kcal/mol, which corresponds to around 8% (in cyclohexane) or 10% (in methylcyclohexane/toluene) of the K tautomer , to −0.71 kcal/mol in chloroform , where this tautomer dominates. The ΔG value of 0.33 kcal/mol in acetonitrile, determined experimentally , have been used to validate the level of theory used in the current study. As seen from Table S1 (Supporting Information File 1) the best result has been achieved by using M06-2X/6-31++G** functional and basis set, which predicts the relative energy of the tautomers (ΔE value, defined as E K − E E ) of 0.33 kcal/mol, perfectly matching the experiment. In the case of 6 the calculations yield a ΔE value of 3.14 kcal/mol in acetonitrile, which leads to the expectation that the tautomeric equilibrium should be fully shifted to 6E. The corresponding most stable structure of the enol form is shown in Figure 1, where hydrogen bonding between the tautomeric OH group and the sidearm carbonyl group can be seen. The tautomeric equilibrium in 1 is strongly solvent-dependent as mentioned above and which can also be seen from Figure 2a. For instance, through intermolecular hydrogen bonding with the carbonyl oxygen atom from the tautomeric backbone, chloroform stabilizes the keto tautomer, absorbing at ≈480 nm, while in acetonitrile the enol form is also presented with a maximum at ≈410 nm. A comparison between the absorption spectra of 1 and 6 shows that the tautomeric equilibrium in 6 is also surprisingly solvent dependent. As shown on Figure 2b, the tautomeric equilibrium in 6 is shifted, but not fully, towards the K form in acetonitrile and chloroform and towards the E form in dichloromethane and toluene. In ethanol and dimethylformamide, the maximum of the enol form visually disappears (the new maximum around 550 nm in the spectrum in dimethylformamide belongs to the deprotonated form, see Figure 4 below). A careful study of the spectra shown in Figure 2b leads to the conclusion that in all solvents the absorption maximum of the enol tautomer is in the range of 415-420 nm, while the maximum of the keto form in acetonitrile (455 nm) is substantially blue shifted compared to the other solvents (≈480 nm in ethanol, dimethylformamide and chloroform). Having in mind the theoretical predictions discussed above, the existence of the keto tautomer in solution is surprising. This behavior might mean that either the enol stabilizing intramolecular H-bonding is not strong enough and can be broken by the solvent or there are intermolecular interactions not taken into account by the calculations. The former could be the reason in some of the solvents, which have the capacity to stabilize 6K as proton acceptor (dimethylformamide), proton donor (chloroform) or both (ethanol). The latter could be the reason for the keto tautomer stabilization in acetonitrile. The explanation for the sudden stabilization of 6K was found by X-ray measurements of its crystal, obtained in acetonitrile. The crystal structure of 6, shown in Figure 3, clearly indicates that the K form is stabilized through the formation of linear intermolecular associates. It can be seen that a hydrogen bond is formed between the nitrogen proton of one keto tautomer and the carbonyl group of another neighboring molecule. Probably, the process of associate formation is facilitated by the position of the chromophore part in the isolated K form (Figure 3, left). Obviously, the formation of the seven-membered hydrogenbonding ring in 6E cannot compete with the flexibility of the system in the case of the intermolecular association. Compared to another tautomeric C=O containing ionophore, recently developed , it seems that the existence of a carbonyl group leads in some cases to stabilization of the keto tautomer through formation of associates. This kind of aggregation reflects to the spectrum of 6 in acetonitrile. The formed aggregate (see Figure S1, Supporting Information File 1) has a H-type structure (also called "sandwich" type) with parallel assignment of the monomer molecules, consequently, its absorption maximum should be blue shifted compared to the monomeric species. If we assume that in ethanol or in dimethylformamide only the monomeric keto form is present, the blue shift of the absorption in acetonitrile indicates that the keto form here exists exclusively as H-aggregates as in the crystal structure. The absorption spectra of 6 in acetonitrile upon addition of Mg(ClO 4 ) 2 are shown in Figure 4. A clear isosbestic point can be seen in the area where the enol tautomer does not absorb, indicating that the tautomeric equilibrium is shifted towards the keto tautomer (in form of the complex) as a result of the general equilibrium scheme below: The complexation provides a substantial red shift (from 455 to 513 nm) compared to the neutral ligand with increased intensi- ty of the new maximum at 513 nm. The comparison between the spectra of the complex and the deprotonated ligand, shown in Figure 4, indicates that the complex formation is not related to deprotonation. These results coincide with the results obtained for compound 3 . The complexation abilities of 6 towards some alkaline-earth metal ions were studied and the obtained spectra of the complexes are given in Figure 5. As seen the λ max values of the complex changes with the change of the type of the metal ion. We assume the formation of a 1:1 complex (the Job's plots are shown on Figure S2, Supporting Information File 1) as shown in Figure 1, which leads to a substantial red spectral shift and allows the recognition of each metal ion based on the complex peak position. The estimated stability constants and the absorption maxima of the complexes are summarized in Table 1. The complex formation causes a substantial red shift, which varies with the metal ion. It is worth mentioning that complexation with any alkaline metal was not observed. As seen, 6 shows strong complexation with Ba 2+ , which fits well with the size of the cavity formed between the two carbonyl groups of the keto form of the ligand, while the corresponding stability constants with Ca 2+ and Mg 2+ are very similar. However, as shown in Figure 5 and Table 1, the difference in the spectral maxima of the complexes allows detection of each of the studied cations. ## Conclusion In the current study, we modeled theoretically and experimentally the tautomerism and complexation abilities of a new tautomeric ligand, based on 4-(phenyldiazenyl)naphthalen-1-ol. According to the theoretical calculations the enol form stabilization could be achieved through a strong intramolecular hydrogen bond formed between the tautomeric hydroxy group and the carbonyl group from the tautomeric backbone. However, intermolecular association plays a role in some solvents as shown by the experimental results. The calculations predict that the complexation with alkali earth metal ions could lead to a full shift of the tautomeric equilibrium towards the keto tautomer, which was finally observed in solution. The formed 1:1 complexes showed large bathochromic and hyperchromic shifts in the visible spectra. ## Experimental Organic synthesis The synthetic route to compound 6 is shown in Scheme 3. ## Preparation of compound c The starting intermediate b was prepared according to the described procedure from commercially available ketone a. Preparation of compound 6 ## 2-(1-Hydroxynaphthalen-2-yl)-N,N-dimethylacetamide (c) Preparation of phenyldiazonium salt solution: Aniline (0.90 mL, 10.00 mmol) was dissolved in a mixture of concentrated hydrochloric acid (5 mL) and distilled water (20 mL). A solution of sodium nitrite (0.83 g, 12.00 mmol) in distilled water (5 mL) was prepared in a test tube. The sodium nitrite solution was added dropwise to the acidic solution of the amine over 5 min at 0 °C. The mixture was stirred at 0 °C for 40 min. Compound c (0.59 g, 2.57 mmol) was dissolved in an aqueous solution of NaOH (1.03 g, 25.73 mmol in 10 mL distilled water) and cooled to 0 °C. The above prepared phenyldiazonium salt solution (6.43 mL, 2.57 mmol) was added dropwise to the solution of c at 0 °C. The resultant deep red mixture was stirred for 1 h at 0 °C. The crude product 6 was precipitated by addition of 20% hydrochloric acid, filtered and washed with distilled water. For further purification the crude product was dissolved in 5 mL dichloromethane and purified by column chromatography -75 g silica gel, phase dichloromethane/methyl tert-butyl ether 10:1. After column chromatography, the product was additionally washed with petroleum ether and dried in vacuum to give 0.720 g (84%) of pure *Due to tautomerism, the NMR spectra in most of the solvents (DMSO-d 6 , CDCl 3 , acetone-d 6 , acetonitrile-d 3 etc.) are not informative. In all cases a complicated mixture of tautomers and lack of signals was observed. Therefore, the NMR spectra were recorded in strong basic media, in order to obtain a single tautomeric skeleton. Nevertheless, some signals in the 13 C NMR spectra do still not appear even after 1024 scans. ## Theoretical calculations Quantum-chemical calculations were performed using the Gaussian 09 D.01 program suite . The M06-2X functional was used with the 6-31++G** basis set for the calculations. This fitted hybrid meta-GGA functional with 54% HF exchange was especially developed to describe main-group thermochemistry and noncovalent interactions. It shows very good results in predicting the position of the tautomeric equilibria for compounds with intramolecular hydrogen bonds as well as describing the ground and excited state proton transfer mechanism . The solvent effect was described using the Polarizable Continuum Model (the integral equation formalism variant, IEFPCM, as implemented in Gaussian 09) . All ground state struc-tures were optimized without restrictions, using tight optimization criteria and an ultrafine grid in the computation of twoelectron integrals and their derivatives. The true minima were verified by performing frequency calculations in the corresponding environment. The TD-DFT method , carried out with the same functional and basis set, was used for predicting vertical transitions. ## Spectral measurements The NMR spectra were recorded on a Bruker Avance II+ 600 spectrometer. In case of CDCl 3 tetramethylsilane was used as internal standard. In case of DMSO-d 6 , the spectra were calibrated to the residual solvent peaks (for DMSO-d 6 : δ = 2.50 for 1 H). 13 C NMR spectra were calibrated in all cases to the residual solvent peaks (for CDCl 3 δ = 77.00, for DMSO-d 6 δ = 39.52). The following additional NMR techniques were used for all compounds: DEPT 135, COSY, HSQC and HMBC. Mass spectra (MS) were recorded on a Thermo Scientific High Resolution Magnetic Sector MS DFS spectrometer. UV-vis spectral measurements were performed on a Jasco V-570 UV-vis-NIR spectrophotometer, equipped with a thermostatic cell holder (using Huber MPC-K6 thermostat with 1 °C precision) in spectral grade solvents at 20 °C. The complexation was studied in acetonitrile. AR grade Mg(ClO 4 ) 2 (Fluka), Ca(ClO 4 ) 2 •4H 2 O (Aldrich) and Ba(ClO 4 ) 2 •xH 2 O (Fluka) were vacuum dried at 90 °C for 3 days. Due to the red shift upon complexation, the estimation of the stability constants was performed at the maximum of the complex using the final complex spectrum (Figure 5). Deprotonation was made with trimethylamine (Aldrich). ## X-ray crystallographic measurements Experimental Single crystals of 6 were crystallized from acetonitrile by slow evaporation. A suitable crystal was selected and was mounted on a loop in oil on a Stoe IPDS2T diffractometer. The crystallographic data of the single crystal were collected with Cu Kα 1 radiation (λ = 1.54186 ). The crystal was kept at 250(2) K during data collection by an Oxford Cryosystem open-flow cryostat. Using Olex2 , the structure was solved with the ShelXT structure solution program using Intrinsic Phasing and refined with the ShelXL refinement package using Least Squares minimization. ## Crystal structure determination of 6 Crystal data for

chemsum

{"title": "Tautomerism as primary signaling mechanism in metal sensing: the case of amide group", "journal": "Beilstein"}

assessing_vesicular_monoamine_transport_and_toxicity_using_fluorescent_false_neurotransmitters

## Abstract: Impairments in the vesicular packaging of dopamine result in an accumulation of dopamine in the cytosol. Cytosolic dopamine is vulnerable to two metabolic processes -enzymatic catabolism and enzymatic-or auto-oxidation -that form toxic metabolites and generate reactive oxygen species.Alterations in the expression or activity of the vesicular monoamine transporter 2 (VMAT2), which transports monoamines such as dopamine from the cytosol into the synaptic vesicle, result in dysregulated dopamine packaging. Here, we developed a series of assays using the fluorescent false neurotransmitter 206 (FFN206) to visualize VMAT2-mediated vesicular packaging at baseline and following pharmacological and toxicological manipulations. As proof of principle, we observed a significant reduction in vesicular FFN206 packaging after treatment with the VMAT2 inhibitors reserpine (IC50: 73.09 nM), tetrabenazine (IC50: 30.41 nM), methamphetamine (IC50: 2.399 µM), and methylphenidate (IC50: 94.33 µM). We then applied the assay to investigate the consequences on vesicular packaging by environmental toxicants including the pesticides paraquat, rotenone, and chlorpyrifos, as well as the halogenated compounds unichlor, perfluorooctanesulfonic acid, Paroil, Aroclor 1260, and hexabromocyclododecane. Several of the environmental toxicants showed minor impairment of vesicular FFN206 loading, suggesting that the toxicants are weak VMAT2 inhibitors at the concentrations tested. The assay presented here can be applied to investigate the effect of additional pharmacological compounds and environmental toxicants on vesicular function, which will provide insight into how exposures to such factors are involved in the pathogenesis of monoaminergic diseases such as Parkinson's disease, and the assay can be used to identify pharmacological agents that influence VMAT2 activity. ## Introduction Proper vesicular dopamine storage is essential for the survival of dopaminergic neurons. When dopamine is not properly packaged into the synaptic vesicle, it accumulates in the cytosol where it is susceptible to oxidation and enzymatic deamination, which when occurring in excess result in oxidative stress and subsequent cell death.1-3 Cytosolic monoamines, including dopamine, are packaged into synaptic vesicles by the vesicular monoamine transporter 2 (VMAT2; SLC18A2). Given the toxic properties of cytosolic dopamine, this process is necessary to reduce potential dopaminergic toxicity.2,4-16 In addition to sequestering monoamines, it has been shown that VMAT2 transports the neurotoxicant 1-methyl-4-phenylpyridinium (MPP+), the toxic active metabolite of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), into synaptic vesicles. MPP+ is a potent inhibitor of complex I of the electron transport chain, and exposure to MPTP results in immediate onset of parkinsonism.17 However, VMAT2-mediated sequestration of MPP+ prevents its accumulation in the cytosol, thereby diminishing its neurotoxic effect and suggesting a neuroprotective role for VMAT2.10,18-22 These data suggest a role of VMAT2 in regulating the neurotoxic effects of both endogenous (e.g. dopamine) and exogenous (e.g. MPP+) toxicants. While genetic deletion of VMAT2 is neonatal lethal, mice with 5% functional expression (VMAT2-LO mice), 50% functional expression (VMAT2-HET), or 200% functional expression (VMAT2-HI mice) are viable. 6,9,11,13,20,23,24 Genetic reduction of VMAT2 expression results in progressive nigrostriatal neurodegeneration as well as olfactory deficits, depressive behavior, and altered sleep latency in mice-symptoms that mimic both the motor and non-motor symptoms of Parkinson's disease (PD).9,25-27 In addition, VMAT2-LO and VMAT2-HET mice exhibit increased cell death upon exposure to the dopaminergic neurotoxicants methamphetamine and MPTP.4,8,9,11-4 methamphetamine and MPTP.4-6 Collectively these data suggest that while losing VMAT2 expression or activity may facilitate neuronal toxicity, an increase in expression or activity may allow VMAT2 to confer neuroprotection. For these reasons, it is essential to understand the factors that modulate VMAT2 expression and activity. Analyses from human studies further implicate VMAT2 as a necessary mediator of neuronal health. Decreases both in the amount of, and activity of, VMAT2 have been detected in postmortem tissue from PD patients,28,29 and mutations in VMAT2 have been identified as causative of infantile parkinsonism.30 Recent work has identified low-activity variants in VMAT2 that may increase the risk of PD, and gain of function polymorphisms in SLC18A2 have been associated with decreased risk for PD.31-33 Furthermore, data shows a decrease in VMAT2 mRNA from platelets of PD patients suggesting that a systemic deficiency in VMAT2 may be a pathologic characteristic of the disease.34 Genetic predisposition only accounts for a portion of PD risk, the rest of which is explained by environmental exposures including manganese and the pesticides rotenone and dieldrin. While the mechanisms by which these toxicants contribute to PD pathogenesis remain unknown, it is possible that these toxicants exert their neurotoxicity in part by affecting VMAT2 function. 40,41 Fluorescent false neurotransmitters (FFNs) were designed as specific substrates for VMAT2 as a tool to visualize vesicular uptake. FFN206 is a fluorescent monoamine analog and substrate of VMAT2 that was first reported in 2013. 42 Here, we replicate the findings of the original study that show FFN206 can be used to investigate VMAT2-mediated vesicular uptake, and report new data at high resolution and in near real-time. Furthermore, we optimized a 96-well plate reader-based screening assay to assess VMAT2 function and the dynamics of vesicle loading under: 1) physiological conditions, 2) during treatment with pharmacological inhibitors of VMAT2, and 3) during treatment with select pesticides and halogenated compounds. Our data show significant reductions in vesicle packaging after treatment with the VMAT2 inhibitors tetrabenazine, reserpine, methylphenidate, and methamphetamine, and modest reduction in vesicle packaging after treatment with the toxicological compounds paraquat, rotenone, unichlor, perfluorooctanesulfonic acid, Paroil, Aroclor 1260, and hexabromocyclododecane. The methods of assessing vesicle function developed here can be used to further screen pharmacological and toxicological factors that alter dopaminergic vesicular storage. Live-cell total internal reflection fluorescence (TIRF) microscopy. HEK+VMAT2 cells were seeded at 60,000 cells per well on laminin-coated glass-bottom 8-well chamber dishes (LabTek) and maintained in selection media until they reached 60% confluence. Upon reaching confluence, the selection media was aspirated and replaced with experimental media containing 1 µM FFN206 (Abcam). Cells were incubated with FFN206 for 1 hour at 37°C with 5% CO2 before the FFN206containing media was aspirated and replaced with experimental media. Cells were then imaged at 37°C with 5% CO2 on the GE Delta Vision OMX total internal reflection fluorescence (TIRF) microscope (FFN206 peak excitation = 369 nm; peak emission = 464 nm). ## Cell Real-time uptake with confocal microscopy. HEK+VMAT2 cells or HEK+mCherry-VMAT2 cells were seeded at 100,000 cells per plate in laminin-coated glass-bottom round 35 mm dishes (ThermoFisher) and maintained in selection media until they reached 80% confluency. Upon reaching confluency, selection media was aspirated and replaced with experimental media. Cells were imaged on a Nikon A1R TE2000 confocal microscope at 37°C with 5% CO2. Cells were imaged for a 30 second baseline before the addition of FFN206 to a final concentration of 20 µM. Imaging lasted for a duration of 5 minutes or 1000 seconds (16.67 minutes; FFN206 peak excitation = 369 nm; peak emission = 464 nm; mCherry peak excitation = 587 nm; peak emission = 610 nm). 96-well plate screening assay. We adapted the protocol utilized by Hu and colleagues42 for use in a 96-well plate reader. HEK+VMAT2 cells were seeded at 40,000 cells per well in half volume, black-walled, laminin-coated 96-well plates (Grenier Bio One) and maintained in selection media at 37oC with 5% CO2 until 90-100% confluent (approximately 24 hours). Upon reaching confluency, selection media was aspirated and replaced with 90 µL of either experimental media or experimental media containing the desired pharmacological compound or environmental toxicant. Plates were incubated with the pharmacological compound and/or environmental toxicant for 30 minutes at 37°C with 5% CO2 before FFN206 (Abcam) was diluted in experimental media and added to the appropriate wells to produce a final concentration of 1 µM FFN206 per well. Plates were incubated with FFN206 for 60 minutes at 37°C with 5% CO2. Wells were then washed with sterile phosphate-buffered saline (PBS, Gibco) and imaged in PBS on a BioTek Synergy H1 multi-mode plate reader (FFN206 peak excitation = 369 nm; peak emission = 464 nm). Statistical analysis. The data were analyzed by t-test, ANOVA, and z factor analysis as appropriate using GraphPad Prism software. The z factor is commonly used in the design of protocols for high-throughput screens and incorporates the positive control mean (𝜇 + ), the positive control standard deviation (𝜎 + ), the negative control mean (𝜇 − ), and the negative control standard deviation (𝜎 − ) 42. ## Results FFN206 packaging is dependent on VMAT2 function. We first sought to confirm that FFN206 fluorescence was a reliable representation of VMAT2-mediated vesicular uptake. To this end, HEK cells stably transfected with mCherry-tagged human VMAT2 (HEK+mCherry-VMAT2) were treated with 1 µM FFN206, and FFN206 fluorescence was recorded after one hour of incubation. FFN206 fluorescence was observed to overlap with VMAT2 fluorescence (Figure 1A), thus confirming that FFN206 was loaded into VMAT2-containing cells. Furthermore, analysis of HEK+VMAT2 cells grown in a glass-bottom 8-chamber dish following incubation with 1 µM FFN206 using a GE Delta Vision OMX Blaze TIRF microscope demonstrated the localization of FFN206 fluorescence within small vesicle-like compartments at high resolution (60x magnification) within live cells (Figure 1B). To further confirm that FFN206 was loaded via VMAT2 into vesicular compartments, HEK cells stably transfected with human VMAT2 (HEK+VMAT2) were treated with 10 µM tetrabenazine (TBZ)a pharmacological inhibitor of VMAT2. Compared to HEK cells lacking VMAT2 that show no FFN206 uptake, treatment with tetrabenazine resulted in almost total loss of FFN206 fluorescence in HEK + VMAT2 cells, indicating that VMAT2 function must be maintained in order to observe FFN206 fluorescence (Figure 1C). Quantification of FFN206 fluorescence displayed as percent control demonstrated a significant increase in fluorescence in HEK+VMAT2 cells treated with FFN206 than all other conditions (Figure 1D). As an additional method of confirmation, cells were treated with bafilomycin, which inhibits the vesicular ATPase that maintains the proton gradient present across the vesicular membrane that VMAT2 depends on for sequestering dopamine. As expected, bafilomycin also depleted FFN206 fluorescence (data not shown). Real-time VMAT2-mediated uptake of FFN206. After examining FFN206 packaging in highresolution in a live cell, we sought to observe the dynamic packaging of FFN206 in living cells in real-time. To that end, HEK+mCherry-VMAT2 cells were grown in glass-bottom round dishes and recorded a baseline of mCherry and background fluorescence for 30 seconds. Cells were then treated with 20 µM FFN206 and fluorescence was recorded for 5 minutes. Image stills from 0.93s, 3.18s, 3.94s, and 5.94s demonstrate that over time, FFN206 fluorescence and its co-localization with the VMAT2-containing vesicles is observed (Figure 2A). We then sought to examine how perturbations in VMAT2 function and vesicle function affected uptake and retention of FFN206. Uptake was recorded under four conditions: HEK cells without VMAT2 (HEK), HEK cells with human VMAT2 (HEK+VMAT2), HEK+VMAT2 cells with VMAT2 inhibitor tetrabenazine (HEK+VMAT2+TBZ), and HEK+VMAT2 cells with the proton gradient dissipater bafilomycin (HEK+VMAT2+BAF). FFN206 fluorescence was recorded for 1000 seconds in each condition The Hu (2013) protocol utilized tetrabenazine as a negative control; thus, we performed a dose response of tetrabenazine-suppressed FFN206 fluorescence to determine the appropriate tetrabenazine dose to utilize in our cell line. HEK+VMAT2 cells were treated with tetrabenazine at 0, 0.0001, 0.001, 0.01, 0.1, 1, and 10 µM concentrations (Figure 3B, D). The dose of tetrabenazine was selected in order to achieve a good dynamic range with a high degree of FFN206 fluorescence suppression with the selected tetrabenazine concentration. With increasing tetrabenazine concentration, FFN206 fluorescence decreased with the greatest suppression of FFN206 fluorescence at 10 µM tetrabenazine; thus, this dose was chosen for the assay. High-throughput assays require a high degree of accuracy and sensitivity, and therefore demand a wide dynamic range and minimal variability within the datasets. The z factor is commonly used in the design of protocols for high-throughput screens and incorporates the positive control mean (𝜇 + ), the positive control standard deviation (𝜎 + ), the negative control mean (𝜇 − ), and the negative control standard deviation (𝜎 − ).43 Z factor calculation ensures that assays with favorable z values (as close to 1 as possible) will have a large band of separation between the distributions of the data for the positive and negative control. A z factor above 0.5 represents a suitable assay. For the FFN206 assay, the positive control was represented by HEK+VMAT2 cells treated with 1 µM FFN206 and 0 µM tetrabenazine, while the negative control was represented by HEK+VMAT2 cells treated with 1 µM FFN206 and 10 µM tetrabenazine. After performing iterative experiments to optimize cell density, incubation time, and reaction volume, a z factor of 0.76 was consistently achieved, indicating that the protocol was suitable for high-throughput screening (Figure 3E). Pharmacological inhibitors of VMAT2. After optimizing the 96-well plate protocol, the assay was used to test a variety of pharmacological VMAT2 inhibitors to demonstrate the utility and accuracy of the assay. Dose-dependent VMAT2 inhibition was observed in HEK+VMAT2 cells treated with reserpine, tetrabenazine, methamphetamine, and methylphenidate from a range of 0.0001 to 10 µM to determine the concentration at which VMAT2 was completely inhibited (Figure 4). HEK+VMAT2 cells treated tetrabenazine yielded an IC50 of 73.09 nM and showed essentially complete inhibition at 1 µM. HEK+VMAT2 cells treated with reserpine yielded an IC50 of 30.41 nM and showed essentially complete inhibition at 0.1 µM. HEK+VMAT2 cells treated with methamphetamine yielded an IC50 of 2.399 µM and showed total inhibition by 100 µM. HEK+VMAT2 cells treated with methylphenidate yielded an IC50 of 94.33 µM. Environmental toxicants and VMAT2 function. The optimized assay was then used to test a variety of pesticides (Figure 5A) and halogenated environmental toxicants (Figure 5B) of interest. All compounds were tested at concentrations of 0.01, 0.1, 1, 10 and 100 µM. The pesticides rotenone and chlorpyrifos caused a minor decrease in FFN206 uptake at the highest concentration, however, paraquat demonstrated a more significant effect on FFN206 fluorescence yielding an IC50 of 12.41 µM (Figure 5A). The halogenated compounds unichlor, PFOS, and hexabromocyclododecane did not show impairment of vesicular loading at the concentrations tested (Figure 5B). However, Paroil and Arochlor 1260 showed modest effects on FFN206 fluorescence with IC50s of 57.03 µM and 95.07 µM respectively (Figure 5B). ## Discussion Here, we demonstrate the utility of using false fluorescent neurotransmitters to investigate the effect of pharmacological and environmental compounds on vesicular uptake in an in vitro application. Importantly, we were able to reproduce key findings from Hu et al. 2013 demonstrating that FFN206 can be used to examine VMAT2 function and the dynamics of vesicle packaging in HEK293 cells. We further went on to optimize a fluorescent 96-well plate assay with a dynamic range that allows for detection of altered VMAT2-mediated vesicular uptake and is amenable to high-throughput screening. We previously developed an assay (Bernstein et al. 2012) to spatially resolve VMAT2-mediated packaging of dopamine utilizing high-content imaging with a fluorescent dye and mCherry-tagged VMAT2, but this assay required time-intensive image analysis to obtain suitable results. 44 The advent of FFN206 allowed us to adapt our assay to a fluorescent plate reader format and to visualize monoamine transport with an ease and in a real-time manner that was previously inaccessible. The methods presented here can be used in experimental applications to understand how pharmacological and environmental manipulation affects vesicle function and thus may contribute to monoaminergic neuron vulnerability. ## Real-time visualization of FFN206 vesicular uptake We next sought to determine whether FFN206 could be used to measure real-time VMAT2mediated monoamine uptake. FFN206 uptake was recorded under four conditions: HEK cells not expressing VMAT2, HEK cells expressing human VMAT2, HEK+VMAT2 cells incubated with tetrabenazine, and HEK+VMAT2 cells incubated with bafilomycin. Quantification of FFN206 fluorescence demonstrated significant differences in HEK cells lacking VMAT2 and HEK cells treated with tetrabenazine and bafilomycin compared to HEK cells expressing human VMAT2. These results were as expected as tetrabenazine inhibits VMAT2 activity, thus preventing the sequestration of FFN206. While cells treated with bafilomycin also exhibit decreased FFN206 fluorescence, this is caused by a different mechanism. Bafilomycin inhibits the vacuolar H+ ATPase present on vesicles, leading to the dissipation of the vesicular proton gradient present across the vesicular membrane. As the proton gradient dissipates, the physiological state of the vesicle is disrupted, and vesicular homeostasis cannot be maintained despite the continued normal function of VMAT2. Thus, the decrease in FFN206 fluorescence of HEK+VMAT2 cells treated with bafilomycin represent the necessity of proton motive force to sequester FFN206 within vesicles. ## Pharmacological and toxicological applications of high-throughput FFN206 assay We adapted the protocol utilized by Hu et al. for use in our 96-well plate assay for highthroughput screening applications.42 We chose concentrations of 1 µM FFN206 and 10 µM tetrabenazine to ensure the assay had the dynamic range necessary to detect both decreases and increases in VMAT2 function. These concentrations were determined by our own dose-response experiments and reproduce the optimal concentrations determined by Hu et al. 2013. Iterative experiments were conducted to refine the protocol until a z score above 0.70 was consistently achieved. We further verified the reliability of the assay by screening pharmacological compounds known to inhibit VMAT2: tetrabenazine, methamphetamine, methylphenidate, and reserpine, and observed dose-dependent VMAT2 inhibition that aligned with previously published results.42 After confirming the reliability of the FFN206 assay, we treated HEK+VMAT2 cells with select pesticides and halogenated environmental toxicants. There is an association between exposure to environmental toxicants and PD and it has long been established that exposure to heavy metals and pesticides contribute to PD risk.36-39,50-54 The extent to which altered VMAT2 function mediates the toxicity of exposure to these environmental toxicants has not been extensively studied. Thus, we tested representative pesticides and halogenated environmental toxicants including rotenone, paraquat, chlorpyrifos, unichlor, PFOS, Paroil, HBCD, and Aroclor 1260 for their effect on VMAT2 function. Of these compounds, exposures to rotenone and paraquat have been most extensively associated with PD in both humans and animal models.52,53,55 Rotenone is a known inhibitor of mitochondrial complex I and exerts its toxicity by oxidizing mitochondrial proteins and causing oxidative stress that leads to cell death.56,57 Similarly, paraquat exerts its toxicity predominantly through oxidative modification of cytosolic proteins, which causes oxidative stress and leads to cell death.55,57,58 Here, we detected a reduction in VMAT2 function with administration of 100 µM of rotenone and with 10 µM and 100 µM of paraquat, though this reduced VMAT2 activity occurred at such high doses as to be physiologically irrelevant. Previous studies examining rotenone and paraquat toxicity have observed 70-80% cell death resulting from mitochondrial complex I inhibition with doses of 100 nM rotenone and 200 µM paraquat.52,57,58 The present study differs from these studies in several respects, the first of which being the length of exposure. We exposed cells to each toxicant for a total of 90 minutes (30-minute incubation with the toxicant followed by a 60-minute incubation with FFN206), while previous studies examined toxicity after 48 hours. We also screened for VMAT2 function and not cell death. As the amount of cell death imposed by a 90-minute length of exposure is minimal, we did not assess cell death in our assay. Finally, given the high IC50 value we determined for paraquat and rotenone exposure, the mild inhibition of VMAT2 caused by exposures to these toxicants is not likely to be mediated by mitochondrial complex I inhibition. For the halogenated compounds, we observed a mild reduction in VMAT2 function at high concentrations. While previous literature indicates exposure to PFOS59, Aroclor 12607, and hexabromocyclododecane60 exerts toxicity on the dopamine system, our results indicate that unichlor, PFOS, Paroil, Aroclor 1260, and hexabromocyclododecane are weak inhibitors of VMAT2 over the course of a short-term exposure, and therefore it is unlikely that these compounds affect VMAT2 function at environmentally and physiologically relevant concentrations. ## Conclusions In conclusion, we have used FFN206 to investigate VMAT2-mediated vesicle uptake at high resolution and in near real-time. Importantly, we were able to reproduce in HEK cells expressing human VMAT2 the findings reported by Hu et al. 2013 performed in HEK cells expressing rat VMAT2. We further optimized a 96-well plate assay that has a dynamic range and is amenable to a high-throughput format, and we used this assay to assess VMAT2 function and the dynamics of

chemsum

{"title": "Assessing vesicular monoamine transport and toxicity using fluorescent false neurotransmitters", "journal": "ChemRxiv"}

ultrathin_nickel_hydroxide_and_oxide_nanosheets:_synthesis,_characterizations_and_excellent_supercap

## Abstract: High-quality ultrathin two-dimensional nanosheets of a-Ni(OH) 2 are synthesized at large scale via microwave-assisted liquid-phase growth under low-temperature atmospheric conditions. After heat treatment, non-layered NiO nanosheets are obtained while maintaining their original frame structure. The well-defined and freestanding nanosheets exhibit a micron-sized planar area and ultrathin thickness (,2 nm), suggesting an ultrahigh surface atom ratio with unique surface and electronic structure. The ultrathin 2D nanostructure can make most atoms exposed outside with high activity thus facilitate the surface-dependent electrochemical reaction processes. The ultrathin a-Ni(OH) 2 and NiO nanosheets exhibit enhanced supercapacitor performances. Particularly, the a-Ni(OH) 2 nanosheets exhibit a maximum specific capacitance of 4172.5 F g 21 at a current density of 1 A g 21 . Even at higher rate of 16 A g 21 , the specific capacitance is still maintained at 2680 F g 21 with 98.5% retention after 2000 cycles. Even more important, we develop a facile and scalable method to produce high-quality ultrathin transition metal hydroxide and oxide nanosheets and make a possibility in commercial applications. Because of high power density, fast charging time, and long lifespan, supercapacitors have recently received considerable attention for the increasing demand in advanced energy storage devices, especially with the emergence of electric vehicles. Pseudocapacitors, mainly employing the fast reversible multi-electron surface Faradaic redox reactions from electrode materials such as Ni(OH) 2 1,2 , MnO 2 3 , NiO 4-6 , and Co 3 O 4 7 , often exhibit very high specific capacitance. However, pseudocapacitive charge storage sites are only limited on surface/ near-surface, the bulk of materials cannot be used, causing very low active material utilization with non-competitive energy density to batteries. They also suffer from limited electrochemical stability under higher rates 2 . They cannot keep good rate capability and high reversibility, simultaneously, for their high dependence on Faradic redox reactions and in general unfavorable reaction kinetics under high rates. In this regard, the option is to explore novel pseudocapacitive materials, where high capacitance is still being retained at fast charge/ discharge rates without any obvious capacity decay or leaving little portion of electrochemical active materials inactive under prolonged cycling, i.e., favorable reaction kinetics and high active material utilization are guaranteed.Recently, the ultrathin two-dimensional (2D) nanomaterials, such as graphene and inorganic nanosheets, have attracted tremendous attention due to their fascinating physical and chemical properties for great potential applications in field-effect transistors, energy storage and conversion, topological insulators, and so on [8][9][10][11] . Generally, the ultrathin 2D nanomaterials are nearly made up of surfaces with molecular-scale thickness, causing a high percentage of surface atoms and high efficient active sites on the exposed surface. Consequently, the theoretical lithium storage capacity of graphene is improved to 744 mA h g 21 , two times as that of graphite, when both sides occupied by Li 1 to form LiC 3 structures 12 . The unique structure and surface properties of the ultrathin 2D nanomaterials make them more attractive in catalysis and energy devices [13][14][15] . More importantly, the ultrathin 2D nanostructures can provide short ion and electron diffusion path distances, large electrochemical active sites and electrode-electrolyte interface, high electronic conductivity, and improved structural stability 16 . So it can be concluded that the ultrathin 2D nanostructures are the most ideal morphological foundation for the surface- dependent electrochemical reactions 17 , representing a very interesting target with great promise for application in next generation batteries and supercapacitors 9,10,18 . However, the exfoliation of layered compounds and high-temperature epitaxial growth are the dominant ways to produce ultrathin 2D nanomaterials. The former is usually uncontrollable in the uniformity of size and shape, thickness, and lateral dimension; while the latter is highly dependent on growth substrates and/or noble metal catalysts accompanied with a low yield and high cost. It has been reported that solution-phase methods, such as solvothermal or colloidal growth reactions, can offer a facile production of 2D materials at gram scale quantities 19,20 . The recent successes in solution synthesis of PbS and CeO 2 ultrathin nanosheets show good promises in producing large quantities of 2D nanostructures 21,22 , but their synthetic processes remain sophisticated and time-consuming. The facile and cost-efficient synthesis of high-quality ultrathin 2D nanomaterials, especially large-area non-layered nanosheets, at large scale has met with limited success and is still extremely urgent. Still, there exists the possibility to create single-atom or few-atom thick 2D layers from any material 23 . With respect to the exploration of various potential 2D nanomaterials, it is a great challenge to develop a universal strategy for fabricating those ultrathin 2D nanostructures. Herein, we report a facile and scalable synthesis of high-quality ultrathin 2D nanosheets of lamellar hydroxides and non-layered metal oxides through microwave-assisted method. Based on the observed formation mechanism, we have successfully synthesized ultrathin Ni-based hydroxide and oxide nanosheets under mild experimental conditions. We also discover that the large-area ultrathin 2D nanostructures can significantly improve the surfacedependent electrochemical processes. In case of our ultrathin a-Ni(OH) 2 and NiO nanosheets, superior electrochemical activity for supercapacitors is achieved. ## Results The preparation process can be summarized in Figure 1a. The fundamental process is dependent on liquid-phase growth of ultrathin lamellar nickel hydroxide precursor under microwave irradiation. This stage undergoes a homogeneous alkalinization of nickel(II) nitrate solutions by urea hydrolysis through inductive effect of microwave irradiation under low-temperature atmospheric conditions. This process is conducted in a 1000 mL three-necked flask in microwave reactor (Figure S1a). The comfortable thermodynamic and kinetic factors are favorable for growth of ultrathin intermediate 24,25 . The formation of ultrathin nanosheets is dominated by a selfassembly and oriented attachment mechanism. The rapid microwave heating can facilitate the super saturation of reactant species, leading to the instantaneous formation of ultrafine nanocrystals and then spontaneous self-assembling or oriented attachment by intrinsic driving force of lamellar nickel hydroxide for 2D anisotropic growth. After heat treatment at 300uC (Figure S2a), a-Ni(OH) 2 nanosheets are completely decomposed into NiO. Figure S2b shows that single crystalline phase of a-Ni(OH) 2 with a hexagonal layered structure (JCPDS 22-0444) was obtained. All the diffraction peaks of NiO are consistent with a face-centered cubic phase (JCPDS 04-0835). The as-synthesized a-Ni(OH) 2 precursor exhibits a geometrically sheetlike 2D structure (Figure 1b). They are uniform and freestanding with a micron-sized planar area. The nanosheets are comprised of ultrafine nanocrystals arranging in planar direction (Figure S3). The sheet-like 2D morphology can be perfectly retained after heat treatment (Figure 1c), and there are no apparent broken or collapsed structures in the final sample, suggesting a good structural stability of this 2D structure. Figure 1d further reveals a highly flexible and gauze-like morphology of non-layered NiO. Figure 2a further shows the freestanding and large-area sheet-like morphology of a-Ni(OH) 2 , none of them assembles into a 3D hierarchical architecture. Figure 2b-c reveal their clear and well-defined outline with highly flexible and transparent features, suggesting fundamental characteristics in common of this kind of ultrathin sheetlike nanostructure. HRTEM investigation in the edge areas of ultrathin nanosheets is a common and direct method to determine the layer thickness microscopically 9 . Figure S4a collected from the folded edge or protuberant ridge of a-Ni(OH) 2 nanosheets demonstrates an average thickness of ,1.52 nm, suggesting that the nanosheets is comprised of 2-3 layers of octahedral Ni(OH) 6 arranging in hexagonal symmetry on planar direction. The XPS analysis is further carried out to determine the composition and the surface electronic state of the as-synthesized a-Ni(OH) 2 nanosheets (Figure S5). Only oxygen and nickel species are detected. SAED pattern (inset in Figure 2c) indicates a polycrystalline nature of the overall NiO nanosheets. The three marked diffraction rings correspond to the {111}, {200}, and {220} planes, respectively. Figure 2d shows that the NiO nanosheet in the selected region shares the same lattice fringe and crystallographic orientation, indicating the same behavior in short range during the crystal transformation. Some observed small pores are irregularly distributed on surface causing a high roughness, which could be created by the evacuation of gaseous contents. FFT pattern (Figure 2e) collected from the area of regular lattice fringes reveals single-crystalline feature. Figure 2f shows some visible lattice fringes with an equal interplanar distance of 2.4 A ˚, corresponding to (111) planes of cubic NiO. Figure 2g demonstrates a ,1.16 nm layer thickness of NiO. As expected, the specific surface area of a-Ni(OH) 2 and NiO nanosheets is as high as 190.15 and 196.01 m 2 g 21 (Figure S6). Such high values are associated with the unique structure of ultrathin nanosheets with extended and rough surfaces. More interestingly, it is found that the NiO nanosheets can form a stable and uniform dispersion in ethanol for weeks (Figure S1d). Compared with traditional wet-chemical syntheses, the microwave-assisted liquid-phase growth can decrease reaction time to less than 20 minutes 26 . The formation of precursor nanosheets is finished in a very short time. After 3 minutes under microwave irradiation with subsequent heat treatment, the perfect non-layered NiO nanosheets can be obtained (Figure S7a). Obviously different from the previously reported wet-chemical syntheses and the analogous microwave-activated procedure 22,26 , the current reaction systems can avoid the morphology transformation to 3D hierarchical structures with prolonged reaction time. Figure S7a-d show that the nonlayered NiO nanosheets synthesized with different microwave irradiating time (3, 5, 15, and 30 minutes) nearly exhibit the same morphology. It could be attributed to the fast local heating from microwave activation and the good structural stability of those nanosheets. This is the first report about a synthesis of freestanding low-dimensional nanostructures, especially ultrathin 2D nanosheets, that independent of reaction time. In the experiments, it is found that the crucial reaction parameter is water. The formation of nanosheets strongly depends on the effects of the water molecules. Under the optimal amounts of water, the NiO samples exhibit a freestanding and large-area sheet-like morphology. Whereas with the contained water decreasing, the final products become folded and assemble into a flower-like quasi-spherical 3D hierarchical architecture (Figure S8a), although retaining their sheetlike building blocks (Figure S8b). In the absence of water, the final morphology further evolves into spherical aggregates (Figure S8c), where nanosheets completely disappearing. Instead, ultrafine nanocrystals tend to spontaneously together forming 3D spherical structures (Figure S8d) rather than arranging in 2D planar direction. So the current experiments demonstrate that directional hydrophobic attraction plays a crucial role in determining morphologies of final products. The formation of ultrathin nanosheets in our microwaveassisted liquid-phase growth procedure is attributed to two factors: layered-structural nature and hydrophobicity. It is well known that the 2D anisotropic growth of nanomaterials needs larger driving force. In case of the layered crystals, they have the tendency to growth into layers. So the intrinsic driving force of lamellar nickel hydroxide is adequate for their 2D anisotropic growth under microwave activation. The layered-structural nature is believed to be a prerequisite for the formation of 2D network in the present facile method. The required hydrophobicity can bring about directional hydrophobic attraction between nanocrystals and water molecules, forming twophase interfaces where the excessive surface energy can be accommodated. A balance of anisotropic hydrophobic attraction and electrostatic interaction can be realized, which is favorable for the spontaneous organization of nanocrystals into nanosheets 27,28 . The resulted interaction could allow for the epitaxial orientation of ultrafine nanocrystals and hinder their potential of shrinking and aggregating. Furthermore, the presence of the hydrophobicity could also terminate their stacking and packing, leading to ultrathin 2D structure rather than 3D graphite-like layered framework. To investigate local atomic arrangements and electronic structures of ultrathin a-Ni(OH) 2 nanosheets, X-ray absorption fine structure spectroscopy (XAFS) measurements at Ni K-edge are conducted. Figure 4a shows the typical layered structure model. Figure 3b shows that the observed spectral peaks for a-Ni(OH) 2 nanosheets slightly shift to the higher energy direction. Particularly, the most remarkable difference appears in the main peak locating at around 8351 eV in terms of position and intensity. This result evidently manifests the affection of interlayer scattering toward the observed spectrum. The ultrathin thickness (1.52 nm) means the few layers (2-3) of the a-Ni(OH) 2 nanosheets, which deliver a little interlayer scattering. The removing of interlayer scattering causes the shift of the three major peaks to the higher energy side 29 . Ni K-EXAFS k 2 x(k) oscillation curve for nanosheets (Figure 3c) exhibits a small reduction in amplitude and a little difference in spectral shape compared with that of bulk counterpart, implying the different local atomic arrangement. The nanosheets exhibit three obvious differences in Fourier transform curve (Figure 3d). First, the Ni-O2/O3 peak intensity of the nanosheets decreases significantly to noise level, which can be attributed to the missing contribution from the O3 atoms. This result once again demonstrates the few layer structure of the ultrathin nanosheets. Secondly, the nanosheets show a shift toward a shorter distance and an intensity decrease in the Ni-Ni3 peak. The shift can be attributed to the presence of surface structure distortion on nanosheets. While the particularly strong Ni-Ni3 peak for the bulk mainly caused by the focusing effect from the Ni1 atom situated halfway on the Ni-Ni3 path 29 . Thirdly, the Ni-Ni1 peak for the nanosheets slightly shifts toward a shorter distance. But the Ni-O1 peak shows no shift with intensity similar to the corresponding bulk peak. These facts suggest that the structural contraction occurs parallel to the Ni layer (i.e. along the ab plane) in nanosheets, which also means a noticeable distortion on surface of the nanosheets that in turn helps to balance their excessive surface energy and then allow them with excellent stability 11 . Ultrathin 2D nanomaterials represent a great promising application in next generation batteries and supercapacitors. With respect to the ultrathin a-Ni(OH) 2 and NiO nanosheets, which are almost completely made up by surfaces with the most active material exposed outside for the highly surface-dependent Faradaic redox reactions, their potential in supercapacitor are systematically investigated. Figure 4 shows typical Cyclic voltammogram (CV) curves of the ultrathin a-Ni(OH) 2 nanosheets. A pair of current peaks can be clearly observed during the cathodic and anodic sweeps, which correspond to the reversible conversion between Ni(II) and Ni(III) . The reaction involves the reversible process of insertion and extraction of OH 21 ions. This result reveals that the charge storage mechanism of the a-Ni(OH) 2 nanosheet electrodes is mainly ascribed to the pseudocapacitance from the Faradaic processes. Furthermore, the redox peaks show a symmetric characteristic, suggesting a high reversibility of a-Ni(OH) 2 nanosheets. Apparently, the current density increases with the increasing scan rate, and all CV curve maintain a similar shape, indicating that the ultrathin a-Ni(OH) 2 nanosheets are beneficial to fast redox reactions. The CV curves show more prominent symmetry at higher scan rates indicating better high-rate response of the a-Ni(OH) 2 nanosheets. In addition, as the scan rate increased, the potential of the anodic and cathodic peaks shifted in more positive and negative directions, respectively, most likely attributed to a high electron hopping resistance or the limitation of the ion diffusion rate to satisfy electronic neutralization during the redox reaction. As shown in Figure 5a, all the nonlinear discharge curves confirm the pseudocapacitive characteristic. Encouragingly, the ultrathin a-Ni(OH) 2 nanosheets deliver a ultrahigh specific capacitance of 4172.5 F g 21 at 1 A g 21 (Figure 5b). This specific capacitance is the highest value reported up to now. The ever reported high specific capacitance is 3500 F g 21 for Co(OH) 2 on ultra-stable Y zeolite 33 , but still lower than our current result. As the discharge current density increased, the ultrahigh specific capacitance is still maintained. The specific capacitances are 3650, 3270, 2820, and 2680 F g 21 at 2, 4, 8, and 16 A g 21 , respectively. The specific capacitance gradually decreases at higher current density due to the incremental (iR) voltage drop. It is found that all the obtained specific capacitances are higher than the corresponding results in previous reports respectively checked at the same current density. It is well known that fabricating direct nanostructured electrodes is the most attractive strategy, such as highly ordered nanostructured array electrode, to achieve ultrahigh electrochemical performances. For example, Yang reported electrodeposited Ni(OH) 2 on nickel foam as direct nanostructured electrodes and obtained a high specific capacitance of 3152 F g21 at 4 A g 21 current density 34 , yet slightly lower than our 3270 F g 21 ; Yuan fabricated the Ni foam supported ultrathin mesoporous NiCo 2 O 4 nanosheets electrodes achieving excellent pseudocapacitance of 1694 F g 21 at 8 A g 21 current density 35 ; Shang reported the coaxial Ni x Co 2x (OH) 6x /TiN nanotube arrays as supercapacitor electrodes, exhibiting a high specific capacitance of 2543 F g 21 calculated from the CV curves (5 mV s 21 ) 36 . In practice, using of direct nanostructured electrodes is favorable for proof-concept studies or for microdevices, but unlikely for commercial applications, especially in electric vehicles or power grid storage. Remarkably, the specific capacitances of our a-Ni(OH) 2 nanosheets are distinctly superior to that of the conceptual designing direct nanostructured electrodes. when current density is increased to 16 A g 21 , the specific capacitance is still 2680 F g 21 , and retain 64.2% of its initial value with a current density increase of 16 times, indicating excellent rate capability due to short diffusion path distances. The specific capacitances are larger than the theoretical value of 2602 F g 21 (within 0.4 V). Besides pseudocapacitance derived from the Faradaic redox reactions, contributions from double-layer capacitance are obvious. Considering the high specific surface area from a-Ni(OH) 2 nanosheets, contribution from double-layer capacitance certainly supply an additional boost to the observed value. As shown in Figure 5c, the a-Ni(OH) 2 nanosheets exhibit excellent cycling stability. The average specific capacitance at 4 A g 21 increases gradually up to 3320 F g 21 in the course of 2000 cycles with a high capacitance retention of 101.5%, due to full activation of electrode material. Even at high rates, 8 and 16 A g 21 , the capacitance retention after 2000 cycles is still maintained at 100% and 98.5%, respectively, indicating stable cycling performance at each current density. The insert in Figure 5c further indicates that the fast charge-discharge process of the electrode is highly reversible. The good electrochemical reversibility and high rate capability can be further explained by EIS measurements (Figure 5d). The Nyquist spectrum can be well represented by an equivalent circuit as the insert shown. The first intersecting point with the real axis in the high-frequency region represents the internal resistance (Rs) including ionic resistance of electrolyte, inherent resistance of active material, and contact resistance at the interface of active material-current collector 37,38 . Throughout experiments, the current collector and the a-Ni(OH) 2 nanosheets are well attached together, except a slightly increased Rs. The Rs are evaluated to be approximately 0.37 and 0.49 V, respectively, before and after 1000 cycles. The slightly increased Rs is mainly attributed to the contact resistance due to the fact that the repetitive insertion and extraction of OH 21 ions lead to active material partially disconnecting from current collector. The high frequency semicircle is attributed to the charge-transfer process of Faradic reactions occurring at the electrode-electrolyte interface. The semicircle diameter reflects charge-transfer resistance (Rf) 39 . After 1000 cycles, the Rf is not significantly altered, indicative of the best electrical conductivity and activity. The a-Ni(OH) 2 nanosheets show nearly vertical Warburg slope in both Nyquist spectrum of the initial one and the 1000th cycle, no significant variation is observed. This result suggests that the Warburg resistance has no determinable affect for the electrode to store charges more efficiently. ## Discussion Based on the liquid-phase growth of lamellar hydroxide precursors, it could summarize an intermediate strategy assisted by microwave irradiation for large-scale synthesis of ultrathin 2D nanosheets of non-layered transition metal oxides for the first time. The combination with microwave heating makes it time-saving. The present procedure requires only simple reagents and equipment, carried out under moderate conditions. In the initial growth stage, the main reaction is conducted in an open system rather than in any highpressure vessels, without seeding protocols or ultrafine control over the temperature and pressure. The weight of the final a-Ni(OH) 2 and NiO nanosheets are 1.48 and 1.14 g at one time synthesis (Figure S1b-c). The process is easily scaled up and the approach can lead to ultrathin 2D nanosheets in high yield, demonstrating it very costeffective and greatly competitive for industrialized production of high-quality ultrathin 2D nanosheets. The ultrathin NiO nanosheets also delivered a high specific capacitance of 2236 F g 21 at 0.5 A g 21 (Figure S9). Even at higher current densities, the specific capacitance is still maintained at 1392 F g 21 at 12 A g 21 , and 1576 F g 21 at 4 A g 21 with 99.1% retention after 2000 cycles, suggesting good rate capability and excellent cycling stability(analysis in detail see Supporting Information). The excellent properties must be associated with fast and efficient ion pathways derived from of the ultrathin structure with a large surface area. In the desirable architecture, a maximum active material utilization and favorable reaction kinetics are guaranteed. Consequently, surfaceenhanced performances are achieved by combining short diffusion distances, large electrochemical active surface area of those ultrathin nanosheets. In summary, we have developed a facial method for large-scale synthesis of high-quality ultrathin 2D nanosheets of hydroxides and transition metal oxides. The nanosheet formation in liquid-phase is attributed to two factors: layered-structural nature and hydrophobicity. The nanosheets exhibit a large planar area and ultrathin thickness. Their special surface properties can facilitate the surfacedependent electrochemical reaction processes. The ultrathin a-Ni(OH) 2 and NiO nanosheets exhibit high specific capacitance as well as good rate capability and excellent cycling stability. This work provides an innovation in fabricating ultrathin 2D nanomaterials and extension to non-layered compound as well as the potential of advanced electrodes for next-generation electrochemical energy storage devices. ## Methods Materials synthesis and characterization. The chemical reagents were of analytical grade and used as received. The a-Ni(OH) 2 nanosheets were synthesized through a facile and scalable microwave-assisted method. Briefly, certain amounts of nickel (II) nitrate hexahydrate and urea with molar ratio of 154 were dissolved in 30 mL deionized water, and then 210 mL ethylene glycol was added in, forming a homogeneous solution. The resulting solution was transferred into a 1000 mL homemade round-bottomed flask and treated under microwave irradiation in a SINEO MAS-II microwave reactor (Figure S1a) at 700 W for several minutes. Finally, puffy grass-green colloid precipitates were obtained, and then cooled down to room temperature naturally. After retrieved by centrifugation and washed several times with distilled water and absolute ethanol, the resulting green product was dried in vacuum at 80uC for 12 h. The final NiO nanosheets were obtained through heat treatment of the nickel hydroxide precursor at 300uC for 2 h. Characterizations of the samples were carried out by X-ray diffractometry (XRD; Bruker D8, CuKa source), field emission scanning electron microscopy (FESEM, Hitachi S-4800) equipped with energy-dispersive X-ray spectroscopy (EDS, Oxford Instruments, INCA), and transmission electron microscopy (TEM, JSM-2100F, 200 kV), and high-resolution TEM (HRTEM, FEI Tecnai G2 F20, 200 kV).The Brunauer-Emmett-Teller specific surface areas (BET) and porosity of the samples were evaluated on the basis of nitrogen adsorption isotherms using a NOVA4200e nitrogen adsorption apparatus (Quantachrome Instruments, USA). Thermogravimetrics (TGA) analysis was carried out by ZRY-2P thermal analysis equipment in air atmosphere with 10uC min 21 . X-ray photoelectron sperctra (XPS) were recorded on a PHI Quanteral II (Japan) with an Al K 5 280.00 eV excitation source. The binding energies were calibrated by referencing the C1s peak to reduce the sample charge effect. The Ni K-edge XAFS were measured at the 1W1B beamline of the Beijing Synchrotron Radiation Facility (BSRF, China). The storage ring energy was operated at 2.5 GeV with a maximum current 250 mA in the decay mode. A Si (111) double crystal was used as the monochromator. Data were collected in transmission mode with ionization chambers filled with a N 2 /Ar mixture at room temperature. The energy calibration was performed with a Ni foil at 8333 eV. In the energy range selected for the experiments a detuning of 30% between silicon crystals was performed to suppress the high harmonic content. Fine samples were brushed onto Kapton tapes that were stacked together to give approximately one x-rayabsorption length at the Ni K-edge, following the traditional sample prepared method 40,41 . Data processing and analysis were made by standard procedures 42 . Electrochemical measurements. The procedures of making electrodes and measurements could be found elsewhere 43 . In practice, The working electrodes were fabricated by mixing the as-prepared active material, acetylene black, and poly(tetrafluoroethylene) (PTFE) in a mass ratio of 75515510 with ethanol, then sonicated for 30 minutes to form a homogeneous slurry. Then the resulting mixture was coated on the pretreated battery-grade nickel foam (0.2 mm thick). The mass variation was reduced using a constant volume of slurries by microinjector (loading mass 1 6 0.03 mg), and dried at 90uC for 12 h in a vacuum oven. The nickel foam with active materials were finally pressed under 10 MPa for 30 seconds to obtain the working electrode. The exposed geometric area of these electrodes is equal to 1 cm 2 and the other surface areas were coated with epoxy resin adhesive. Electrochemical measurements were carried out using a three-electrode system. The 6 M and 2 M KOH solution were used as electrolyte for a-Ni(OH) 2 and NiO nanosheets, respectively. A platinum foil (1 3 2 cm 2 ), and a saturated calomel electrode (SCE) were used as counter electrode and reference electrode, respectively. The cyclic voltammogram (CV) and electrochemical impedance spectroscopy (EIS) were measured on CHI660C electrochemical workstation (Shanghai Chenhua Co. Ltd., China) and IM6e electrochemical workstation (Zahner, Germany). CV tests of a-Ni(OH) 2 nanosheet electrode were carried out between 20.1 and 0.43 V (vs. SCE) at various scan rates, respectively, and the EIS measurements were collected in the frequency range from 0.01 Hz to 100 kHz at open-circuit potential of ,0.23 V with an alternating current (ac) amplitude of 5 mV. Galvanostatic charge/discharge curves of a-Ni(OH) 2 and NiO nanosheets were measured in the potential range of 20.05 to 0.35 V and 20.05 to 0.45 V at different current densities on a CT2001A LAND Cell test system.

chemsum

{"title": "Ultrathin Nickel Hydroxide and Oxide Nanosheets: Synthesis, Characterizations and Excellent Supercapacitor Performances", "journal": "Scientific Reports - Nature"}

azologization_of_serotonin_5-ht₃_receptor_antagonists

## Abstract: The serotonin 5-hydroxytryptamine 3 receptor (5-HT 3 R) plays a unique role within the seven classes of the serotonin receptor family, as it represents the only ionotropic receptor, while the other six members are G protein-coupled receptors (GPCRs). The 5-HT 3 receptor is related to chemo-/radiotherapy provoked emesis and dysfunction leads to neurodevelopmental disorders and psychopathologies. Since the development of the first serotonin receptor antagonist in the early 1990s, the range of highly selective and potent drugs expanded based on various chemical structures. Nevertheless, on-off-targeting of a pharmacophore's activity with high spatiotemporal resolution as provided by photopharmacology remains an unsolved challenge bearing additionally the opportunity for detailed receptor examination. In the presented work, we summarize the synthesis, photochromic properties and in vitro characterization of azobenzene-based photochromic derivatives of published 5-HT 3 R antagonists. Despite reported proof of principle of direct azologization, only one of the investigated derivatives showed antagonistic activity lacking isomer specificity. Introduction 5-Hydroxytryptamine (5-HT), commonly known as serotonin or enteramine , is a monoamine neurotransmitter and hormone which is produced in the brain and in intestines and regulates a large variety of physiological functions in the mammalian central and peripheral nervous system . In the central nervous system (CNS), it modulates sleep-wake cycles, emesis, appetite, mood, memory, breathing, cognition and numerous other functions . In the gastrointestinal (GI) tract, it causes peristalsis via either smooth muscle contraction or enteric nerve depolarization . It is also found in the platelets, where it is presumably involved in blood coagulation and vasoconstriction. Furthermore, serotonin is one of the first neurotransmitters to appear during development and may have an organizing function in the development of the mammalian CNS being involved in cell division, differentiation, survival, neuronal migration and synaptogenesis . Dysfunction of the 5-HT receptor (5-HTR) signalling during early developmental stages my lead to altered cognitive ability, neurodevelopmental disorders, and increased incidence of psychopathologies such as autism and schizophrenia . Serotonin operates via seven classes of 5-HT receptors of which six are G protein-coupled receptors (GPCRs) and only one, the 5-HT 3 R, is a ligand-gated cation channel . When this receptor was identified and cloned , it became clear that 5-HT 3 takes a unique position as pentameric ligand-gated cation-selective ion channel belonging to the Cys-loop receptor subfamily. In vertebrates, this family also includes nicotinic acetylcholine receptors (nAChRs), γ-aminobutyric acid type A receptors (GABA A Rs), and glycine receptors (GlyRs). To date, five subunits of the 5-HT 3 receptor are identified (5-HT 3 A-5-HT 3 E) . Functional receptors are either constructed as 5-HT 3 A homopentamers or as heteropentamers containing 5-HT 3 A and 5-HT 3 B receptor subunits . 5-HT 3 receptors are highly expressed in the brainstem, especially in areas involved in the vomiting reflex and in the dorsal horn of the spinal cord . These receptors are also expressed presynaptically providing regulation of the neurotransmitters release . Besides targeting of 5-HT 3 Rs for the treatment of psychiatric disorders, they are object to counteract postoperative nausea and chemo-/radiotherapy provoked emesis . In the early 1990s, the first potent and selective 5-HT 3 receptor antagonist ondansetron was initially developed . Since then the development of 5-HT 3 R antagonists progressed. The first-generation antagonists are structurally categorized in three major classes: (I) carbazole derivatives (e.g., ondansetron), (II) indazoles (e.g., granisetron), and (III) indoles (e.g., dolasetron) . Generally, 5-HT 3 R antagonists share a basic amine, a rigid (hetero-)aromatic system and a carbonyl group or isosteric equivalent which is coplanar to the aromatic system. Although the antagonists show a general structural motive, they differ in their binding affinities, dose responses, and side effects . To improve prospective antagonists and obtain a systematic tool for receptor investigation, spatial and temporal restriction of ligand binding and concomitant activity regulation is desirable. Fuelled by light, the growing field of photopharmacology provides a noninvasive method to trigger a drug's pharmacological response on demand . To introduce photoresponsiveness into a biological system, different approaches are feasible, e.g., the use of caged ligands (CL) , photoswitchable tethered ligands (PTLs) , photoswitchable orthogonal remotely tethered ligands (PORTLs) or photochromic ligands (PCLs) . The latter ones represent small molecules, which can either be engineered via extension of the chemical structure of a known pharmacophore towards a photochromic moiety or via replacement of certain parts of the biomolecule to generate a photochromic hybrid biomolecule. In this context, various photochromic scaffolds including dithienylethenes, fulgi(mi)des, and azobenzenes are investigated . The latter ones were already discovered in 1834 by E. Mitscherlich but it took around another 100 years till G. S. Hartley revealed their photo-induced trans-cis isomerization representing the time of birth of the azobenzene photoswitch. Benefiting of their accessible synthesis, large change in polarity and geometry upon switching, excellent photochromic properties and tuneability, azobenzenes are amongst the most widely used photochromic scaffolds [31,42, . Since their first use in a biological environment in the late 1960s for the photoregulation of the enzymatic activity of chymotrypsin , their applications in biology widely expanded towards receptor control and fields as bacterial growth , vision restoration , the respiratory chain and lipids . Owing to the reported serotonin antagonists' chemical structures, the use of azobenzene as photochromic scaffold in the presented work seemed axiomatic. Therefore, the primary design of our photochromic derivatives is based on the direct "azologization" of reported non-photochromic antagonists via replacement of the benzenering connecting amide bond and thioether, respectively, by an azo bridge. ## Design and synthesis of azobenzene-based photochromic modulators The reported scaffolds of 5-HT 3 R antagonists are based on an aromatic system either connected to a purine/pyrimidine moiety via a thioether bridge or a quinoxaline moiety via an amide bond. Referring to this work performed by the groups of DiMauro and Jensen , we envisioned that the replacement of the thioether or amide bond (Scheme 1) by an azo bridge would result in highly active photochromic serotonin 5-HTR antagonists controllable by irradiation with light. Based on the suggested receptor binding mode reported for one potent non-photochromic antagonist (lead structure of 16c) we expected the extended trans-isomer as biologically active configuration whereas its bent cis-isomer should be inactive. ## Synthesis of the quinoxaline-based azobenzenes The synthesis of the unsubstituted quinoxaline-based azobenzene derivatives 5a and 5b is based on a Baeyer -Mills reaction (Scheme 2). Therefore, nitrosoquinoxaline 3 was synthesized in a two-step procedure starting from 2-chloroquinoxaline (1), which was transformed into its oxime 2 using hydroxylammonium chloride . The subsequent oxidation Scheme 1: Approach of the direct azologization of reported serotonin 5-HT 3 R antagonists via replacement of a thioether or amide bond by an azo bridge. Scheme 2: Synthesis of the differently substituted quinoxaline azobenzene derivatives 5a and 5b via Baeyer -Mills reaction . Scheme 3: Synthesis of the methoxy-substituted quinoxaline derivative 12a via diazotization . was performed using periodic acid as oxidant . The subsequent reaction with differently substituted anilines in acetic acid provided both quinoxaline azobenzene derivatives in good yields. The methoxy-substituted quinoxaline azobenzene derivative 12a was synthesized via a different synthetic route depicted in Scheme 3. In a first step, p-toluidine (4a) was diazotized using sodium nitrite and subsequently reacted with the 2-chloroacetyl-Scheme 4: General procedure for the synthesis of purine-and thienopyrimidine-substituted arylazobenzenes and depiction of the corresponding structures . acetone ester derivative 7 providing hydrazine 8 . Upon reaction of the chloro-ester 8 with phenylenediamine (9) in the presence of triethylamine the quinoxaline moiety was formed . Oxidation of the hydrazine derivative 10 using hydrogen peroxide under an oxygen atmosphere afforded the quinoxaline azobenzene derivative 11 . Subsequent methylation using methyl iodide mainly resulted in the formation of the N-methylated non-photochromic product 12b but in low yields also the desired photochromic methoxy-substituted quinoxaline azobenzene derivative 12a. ## Synthesis of the purine and thienopyrimidinebased derivatives Scheme 4 depicts the general procedure applied for the synthesis of differently substituted purine-and thienopyrimidine azobenzene derivatives. Differently substituted non-photochromic antagonists were chosen as lead structures delivering photochromic derivatives with varying electronic and thus photochromic properties. The respective arylamines 13a-c were converted into their corresponding hydrazines 14a-c via diazonium-salt formation using sodium nitrite and subsequent reduction using tin(II) chloride . The following nucleophilic substitution at a chloro-substituted purine (15a,b) or thienopyrimidine (15c), respectively, and subsequent oxidation of the hydrazine moiety afforded the corresponding azobenzene derivatives 16a-d . ## Synthesis of azobenzene-extended thiopurine derivatives To further tune the photochromism and compare the properties of direct azologization to azo-extension, two additional derivatives of the in vitro most promising naphthalene azopurine 16c were synthesized either by keeping the original thioether (Scheme 5) or replacing it by an amide bond (Scheme 6) known as common structural feature of 5-HT 3 R antagonists. Scheme 5 reflects the synthesis of the azo-extended thiomethylpurine 23 starting with the synthesis of hydroxymethylazobenzene 19 in a Baeyer -Mills reaction and subsequent nucleophilic substitution using cyanuric chloride (20) providing chloromethyl azobenzene 21. The introduction of the thiopurine moiety in 23 was accomplished upon reaction of 21 with dihydropurinethione 22 . The amide-linked derivative of thiomethylpurine azobenzene 23 was synthesized via Baeyer -Mills formation of the carboxylated azobenzene 25 starting from aminobenzoic acid 24 and nitrosobenzene (18) . Activation using thionyl chlo-Scheme 5: Synthesis of the thiomethyl-linked purine azobenzene 23 [62,63, . Scheme 6: Synthesis of the amide-linked azobenzene purine 28 [62,63, . ride afforded the acid chloride 26 and allowed amide-bond formation for the generation of 28 (Scheme 6). ## Photochromic properties The investigation of the photochromic properties of the potential 5-HT 3 R antagonists 5a, 5b, 12a, 16a-d, 23, and 28 was performed in DMSO and depending on their solubility in phosphate buffer + 0.1% DMSO (16a-d) by UV-vis absorption spectroscopy. The compounds were dissolved at 50 µM in the respective solvent and irradiated with the indicated wavelengths to generate a substantial amount of their cis-isomer. This process can be followed by a decrease of the trans-absorption maximum at around 350-400 nm and an increase in absorption at around 450-500 nm in the UV-vis spectrum representing the cis-isomer (Figure 1, black arrows). The absorption bands of the trans and cis-isomers of compounds 12a, 16c, and 16d overlap to such an extent, that no new maximum representing the cis-isomer was observed and thus cis-trans isomerization only occurs thermally and is not triggerable by irradiation with visible light. Back-isomerization was triggered by irradiation with visible light (5a, 5b, 16a, 16b, 23, and 28) of the indicated wavelength or by thermal relaxation (5a, 5b, 12a, 16a-d, 23, and 28). The irradiation times were determined by following the UV-vis spectrum upon isomerization until no more changes in absorption were observed and the photostationary state (PSS) was reached. The points of intersec- S1 and Table S2. A comparison of the differently substituted purine azobenzene derivatives revealed the beneficial effect of an o-chloro substitution on the photochromic properties of 16b compared to 16c as the electron density at the nitrogen-rich purine core is reduced. Further reduction of the electron density was achieved by using a thienopyrimidine (16a) instead of a purine core (16b-d). Nevertheless, the photochromic properties of those heterocyclic, especially purinebased azobenzenes, are rather poor. In addition to direct azologization, two azo-extended purine derivatives 23 and 28 were synthesized resulting in excellent photochromic properties. Figure 1 compares exemplarily the UV-vis absorption spectra of the naphthalene-azo-purine 16c (left) and its azo-extended azobenzene thioether purine 23 (right). The determination of the thermal half-lives (THL) of the cis-isomers of compounds 5a, 5b, 12a, 16a-d, 23, and 28 was accomplished by monitoring the increase in absorbance which corresponds to the evolution of the trans-isomer after irradiation and exposure to dark. In contrast to the heterocyclic compounds 5a, 5b, 12a, and 16a-d with a thermal half-life in the seconds to minutes range, the azoextended compounds 23 and 28 showed only slow thermal back-isomerization (day range) at room temperature. Depending on the desired application, both properties may be of benefit. For thermally instable compounds, only one wavelength for switching is required. In case of thermally stable cis-isomers constant irradiation to maintain a substantial amount of the cisisomer can be avoided. ## Patch-clamp studies The synthesized azo antagonist derivatives 5a, 5b, 12a, 16a-d, 23, and 28 were tested for their inhibitory activity using the patch-clamp technique on heterologously expressed ionotropic homopentameric 5-HT 3 A receptors. Only upon addition of 16c the amplitude of the 5-HT 3 A mediated currents was decreased (Figure 2, left). Application of a 50 µM solution of trans-16c in its thermal equilibrium decreased the amplitude of 5-HT induced currents on 54 ± 3% (n = 4). However, irradiation-induced trans-cis isomerization with light of λ = 530 nm and 455 nm, respectively, had no significant effect on the amplitude of 5-HT 3 A-mediated currents (Figure 2, right). ## Conclusion In the presented work, we address the design, synthesis, photochromic characterization and in vitro investigation of in total nine azobenzene-based derivatives of reported 5-HT 3 R antagonists. Initially, seven photoligands (5a, 5b, 12a, and 16a-d) either based on quinoxaline (5a, 5b, and 12a) or purine derivatives (16a-d) with varying electronic and thus photochromic properties were synthesized by direct azologization of the respective leads. Especially the purine-based azobenzenes displayed high solubility in aqueous media. The beneficial effect of substituents reducing the overall electron density of the purine moiety (16a, 16b) resulted in higher photostationary states and better band separation compared to 16c and 16d. Still, only one compound (16c) showed antagonistic activity in patch-clamp studies. This might be explained by the fact that its corresponding non-photochromic lead is the inhibitory most active reported antagonist among the investigated ones. The partial rigidization of the thioether via incorporation of an azo bridge might result in a vast loss of activity. Thereby, azologization of the less potent leads resulted in complete loss of inhibitory activity (5a, 5b, 12a, 16a, 16b, 16d) and only the originally most potent derivative 16c kept recordable antagonistic activity. The missing significant difference in activity upon irradiation-induced trans-cis isomerization of 16c is probably due to its moderate photochromic properties and slow trans-cis isomerization (Figure 1, left). During the patch-clamp analysis, the cells are continuously superfused with external solution resulting in a fast exchange of the surrounding media and co-applied tested compounds. Thus, the cis-PSS of 16c might not be reached by irradiation within the short time of compound application despite continuous irradiation. Therefore, two azobenzene-extended derivatives (23 and 28) with improved photochromic properties were synthesized but lost antagonistic activity probably due to their increased steric demand. In ongoing studies, detailed molecular modelling is used to design potential photochromic antagonists fitting the requirements of the receptor's binding pocket. Regarding the analysis method, compounds will be optimized towards either thermally stable cis-isomers to be tested separately upon prior irradiation or faster switching compounds.

chemsum

{"title": "Azologization of serotonin 5-HT₃ receptor antagonists", "journal": "Beilstein"}

experimental_and_computational_chemical_studies_on_the_corrosion_inhibitive_properties_of_carbonitri

## Abstract: The present work aims to study 6-amino-4-aryl-2-oxo-1-phenyl-1,2-dihydropyridine-3,5-dicarbonitrile derivatives namely: 6-Amino-2-oxo-1,4-diphenyl-1,2-dihydropyridine-3,5-dicarbonitrile (PdC-H), 6-Amino-2-oxo-1-phenyl-4-(p-tolyl)-1,2-dihydropyridine-3,5-dicarbonitrile (PdC-Me) and 6-Amino-4-(4-hydroxyphenyl)-2-oxo-1-phenyl-1,2-dihydropyridine-3,5-dicarbonitrile (PdC-OH) as corrosion inhibitors to provide protection for carbon steel in a molar hydrochloric acid medium. Chemical measurements such as (weight loss) and electrochemical techniques such as (Potentiodynamic polarization, electrochemical impedance spectroscopy, and Electron frequency modulation) were applied to characterize the inhibitory properties of the synthesized derivatives. The adsorption of these derivatives on the carbon steel surface was confirmed by Attenuated Total Refraction Infrared (ATR-IR), Atomic Force Microscope (AFM), and X-ray Photoelectron Spectroscopy (XPS). Our findings revealed that the tested derivatives have corrosion inhibition power, which increased significantly from 75.7 to 91.67% on the addition of KI (PdC-OH:KI = 1:1) to inhibited test solution with PdC-OH derivative at 25 °C. The adsorption process on the metal surface follows the Langmuir adsorption model. XPS analysis showed that the inhibitor layer consists of an iron oxide/hydroxide mixture in which the inhibitor molecules are incorporated. Computational chemical theories such as DFT calculations and Mont Carlo simulation have been performed to correlate the molecular properties of the investigated inhibitors with experimental efficiency. The theoretical speculation by Dmol3 corroborates with the results from the experimental findings.Corrosion is a global industrial concern worldwide for decades due to its deteriorating effect on materials not only metallic materials, but also other materials used in the construction industry. Corrosion causes significant economic loss and jeopardizes human safety and leads to operation shutdown, waste of resources, loss of product, reduced efficiency, increased maintenance needs, and cost over design [1][2][3][4] . Carbon steel is an ubiquitous alloy in many industrial fields such as metal processing equipment, petrochemical production, refining, and chemical processing owing to its low cost, ease of machinery, and unique mechanical properties compared to other metallic alloys 5 . Acid solutions are commonly utilized in chemical treatment processes such as acid cleaning, oil-well acidizing, acid descaling, etc. It is a highly aggressive medium for corrosion of carbon steel, as a result, such treatment with an acid brings about undesirable consequences, especially when used in direct contact on C-steel alloys 6 . One acceptable approach to combat corrosion is to incorporate corrosion inhibitors that reduce corrosion rates to the required level with minimal impact on the environment 7,8 . It is therefore, to date, the addition of corrosion inhibitors remains an essential procedure to mitigate the destructive attack of corrosive acid on the metal surface 9,10 . In a major number of studies, effective inhibitors are those organic compounds with hetero atoms such as Nitrogen, Oxygen, Sulfur, which show promising efficiency in mitigating the aggressive attack of corrosive species on metals 11,12 . These inhibitors act at the interface between the metal and the acidic solution and their interaction with the metal surface through the adsorption process that stops the dissolution of metal surface 13 . Weight loss measurements. The weight-loss method is a high-accuracy method for laboratory corrosion study to determine the corrosion rate ( C.R ) and the inhibition efficiency (η). Pre-treatment and pre-weighed carbon steel specimens were immersed in the aggressive media in the absence and presence of the studied additives with the concentration range of ((2-10) × 10 -5 M). Maximum immersion time was 180 min. After equal time intervals (30 min), specimens were taken from the acidic medium, rinsed with bi-distilled water, air dried, and precisely re-weighed. All measurements were performed in triplicate to ensure the reproducibility of the results. Average weight loss values ( W ) were recorded and the corrosion rate was calculated using equation 24 : Table 1. %IE of some carbonitrile compounds in 1 M HCl and for steel corrosion. Mild steel 1 M HCl 95.75 24 The combined admixture of benzene carbonitrile and 5-bromovanillin (BNV 0.25%) Carbon steel 1 M HCl 97.95 25 The combined admixture of benzene carbonitrile and 5-bromovanillin (BNV 0.25%) Carbon steel 1 M HCl 97.95 25 (i) 6-Amino-2-oxo-1,4-diphenyl-1,2-dihydropyridine-3,5-dicarbonitrile (PdC-H) Carbon steel 1 M HCl 80.5 Our work (ii) 6-Amino-2-oxo-1-phenyl-4-(p-tolyl)-1,2-dihydropyridine-3,5-dicarbonitrile(PdC-Me) 78.5 (iii) 6-Amino-4-(4-hydroxyphenyl)-2-oxo-1-phenyl-1,2-dihydropyridine-3,5-dicarbonitrile (PdC-OH) 77.8 where C.R is the corrosion rate (mg/cm 2 /min), W is the average weight loss (mg) for the carbon steel specimens, A is the surface area (cm 2 ) and T is the immersion duration. From the calculated C.R values, the inhibition efficiency (η) can be calculated by the following equation 25 : where CR 1 and CR 2 are the corrosion rate values in the absence and presence of different concentrations of carbonitrile derivatives, respectively. Electrochemical measurements. Further corrosion tests were investigated electrochemically in a threecompartment glass cell comprising a working electrode (carbon steel electrode) with an uncovered area of 1 cm 2 , reference electrode (saturated calomel electrode), and counter electrode (platinum foil). Electrochemical measurements were performed under static conditions in a naturally aerated solution of 1 M HCl in the absence and presence of varying concentrations of three studied derivatives at 25 ± 1 °C. Before each electrochemical experiment, the system was maintained in an unperturbed state for 1800s to reach a stable value of the open circuit potential (OCP) . The polarization curves started from the cathodic direction to the anodic direction and (1) Table 2. Chemical structures of the tested compounds. www.nature.com/scientificreports/ were carried out by automatically sweeping the electrode potential from − 0.5 to + 0.5 V segment to OCP at a scan rate of 0.5 mV s −1 . Polarization parameters were obtained by extrapolation the anodic and cathodic regions of the Tafel plots. EIS measurements were performed after the attainment of steady-state OCP by analyzing the frequency response of the electrochemical system with a range-extending from 0.01 Hz at low frequency to 100,000 Hz at high frequency and the excitation signal is a 5-mV sine wave. In the EFM technique a single, lowfrequency, and low-distortion, sinusoidal voltage is applied to the corrosion interface (amplitude 10 mV with 2 and 5 Hz sine waves). Gamry PCI4-G750 Potentiostat/Galvanostat/ZRA. Echem Analyst V6.30 Software has been applied for fitting the electrochemical data. Surface analysis by ATR-IR, AFM and XPS. Morphological analyses methods were performed to evaluate the impact of corrosion and other characteristics of films on the surface of carbon steel. The mechanically prepared steel specimens were polished by emery paper in sequence to 2000 grade, degreased with absolute ethyl alcohol and acetone, washed with distilled water, and dried in a vacuum system. The pre-treated carbon steel surface is engrossed in 1 M HCl solution in the absence and presence of the optimal concentration of each studied derivative (10 −4 M) at 25 °C for 24 h. The chemical composition of the investigated compounds can be determined using ATR-IR (Thermo Fisher Scientific, NicoletiS10 model) analysis; Infrared spectra were recorded using ATR (Attenuated Total Reflection) in the wavenumber range 400-4000 cm −1 . The morphological changes of the corroded carbon steel surface before and after treatment with the studied carbonitrile derivatives were assessed using non-contact mode atomic force microscopy (AFM) (Model: Thermo Fisher Nicolet IS10 (Scanning probe microscope)). The elemental composition and chemical states of the inhibitive film formed on the surface of carbon steel were examined by XPS (Model: Thermo Fisher Scientific) via Al Kα X-ray source (− 10 to 1350 eV) spot size 400 micro m at pressure 10-9 bar with full-spectrum pass energy 200 eV and at narrow-spectrum 50 eV (produced in USA K-ALPHA). ## Quantum chemical calculations. To inspect the correlation between the molecular structure and the reactivity of 6-amino-4-aryl-2-oxo-1-phenyl-1,2-dihydropyridine-3,5-dicarbonitrile derivatives, theoretical calculations were performed using the DMol3 module adopted in Materials Studio version 7.0. Within the DMol3 module, a basis set of double number polarization (DNP) plus the exchange-correlation functions of Becke One Parameter (BOP) with generalized gradient approximation (GGA) and solvent effects were treated using COSMO controls 26 . Quantum calculations are also performed using Orca 4.1, and the calculation inputs are set at the DFT level with DEF2-SVP as basis Set and B3LYP as functional. The aim of using two different methods and software is to verify the consistency between the experimental and theoretical results in anticipation of the ranking of inhibition efficiency of studied compounds 6 . Monte Carlo simulations. MC simulations were performed to explore the interaction between the inhibitor molecules and the iron surface. The adsorption simulation of the compound on the carbon steel surface in HCl medium was performed using the module of adsorption locator applied in Materials Studio 2107. Carbon steel is significantly made of iron atoms, so the surface was obtained by the cleaved plane of iron identified as the stable crystal plane for iron. The optimized structures of the inhibitor molecules from the DFT study were used in Monte Carlo Simulations. The simulation proved to be related to the experimental study, in which the adsorption of each inhibitor molecule on Fe (1 1 0) was simulated in the presence of hydrochloric acid solution characterized by H 3 O + and Cl − ions in the abundance of water molecules. The simulation was carried out by Monte Carlo method . ## Results and discussion Weight loss method. Effect of concentrations. To assess the effect of varying concentrations of carbonitrile derivatives ((2-10) × 10 -5 M) on the effectiveness of corrosion inhibition, weight loss measurements were performed in 1 M HCl solution. Figure 2 presents the weight loss-time curve for carbon steel in 1 M HCl in the absence and presence of different concentrations of PdC -OH at 25 °C. The weight loss-time curves for PdC-Me, PdC-H are found in the supplementary material (Fig. S1). It is evident that the weight loss of carbon steel in the inhibitor-containing solutions decreases over time with increasing the concentration of inhibitors 30 . The corrosion rate (C.R) , the degree of the surface coverage (θ) , and the inhibition efficiency (%η) are summarized in Table 3. These results indicate that the addition of the studied derivatives in the acidic medium reduced the corrosion rate, which led to an increase in the degree of surface coverage of the inhibitor on the steel surface. In addition, the inhibition efficiency %η increases as the inhibitor concentration increases indicating the formation of an adsorbed barrier layer upon the steel surface, wherein the inhibitor acts as an adsorbate and the metal surface behaves as an adsorbent. Interestingly, the PdC -OH compound shows higher values of %η than PdC-Me and PdC-H, signifying that the molecular structure of the inhibitors affected the adsorption properties of organic molecules on the steel surface. The strong conjugation between benzene and the pyridine ring and the specific hetero-aromatic ring (-OH) promotes the adsorption of the derivative on the steel surface, thus increasing %η 31 . The effect of KI added to PdC-OH compound. Adding KI can increase the inhibition efficiency as cited 32 . The addition of KI to PdC-OH results in higher inhibition efficiency. As indicated from Table 4, the inhibition efficiency of individual PdC-OH at 10 -4 M is 75.7% (from weight loss results), while the inhibition efficiency of PdC-OH in combination with KI is 91.67%. In the presence of the auxiliary inhibitor and KI, the inhibition efficiencies increase with increasing the concentration, which can be shown from the curves in Fig. 3. The combination of PdC-OH and KI shows a synergistic effect. The synergism parameter S θ is calculated as follows 33 is the surface coverage by the main inhibitor; θ 2 is the surface coverage by the auxiliary inhibitor; θ′ 1+2 is the measured surface coverage by both the main inhibitor and the auxiliary inhibitor (the concentration of the main inhibitor 1 (PdC-OH in this case) and the auxiliary inhibitor 2 (KI in this case) in the mixture should be the same as used responding to separate situations). If S θ approaches 1, it indicates that there are no interactions between the two inhibitors, while if S θ > 1, the synergistic effect exists; in the case of S θ < 1, the antagonistic interaction might prevail. The value of the synergism parameter for PdC-OH and KI at 6 × 10 −5 M studied from weight loss measurement is 1.2 and the value of the synergism parameter for PdC-OH and KI at 8 × 10 -5 is 1.21 both are larger than 1. This synergistic process occurs through the oxidation of Iions in the solution by the dissolved oxygen, resulting in the generation of I 2 . Then I 2 combine with I − to form soluble yellowish 33 which acts as a bridge and connects between the inhibitor and the steel surface. The protection of the steel surface and thereby greater inhibition ability is mainly contributed to the presence of the main inhibitor (PdC-OH) and the existence of the adsorbed iodide ions. Adsorption isotherm. The performance of the inhibitor molecules in the aqueous medium results from their adsorption propensity on the corroded surface of the metal and interfering with the electrochemical reactions over the area covered by the inhibitor molecule. Several isotherms equations such as Frumkin, Langmuir, Temkin, and Freundlich were deduced by fit of the surface coverage ( θ ) from experimental data as a function of concentration ( c ) to determine the nature of interactions between the investigated organic molecules and the corroding metal surface during the corrosion inhibition process. The Langmuir isotherm equation is the best-fit equation to the results where the values of the regression coefficient R 2 approaches from unity. Langmuir Isotherm Plots is provided in Fig. 4. This behavior indicates that a monolayer of the adsorbed inhibitor was formed on the surface of the metal substrate according to the Langmuir equation (4) : As shown in Fig. 4, C/θ is the ordinate (Y-axis), C is the abscissa (X-axis), 1/K ads is the intercept where the equilibrium constant (k ads ) values can be determined from this interception. Furthermore, the standard adsorption Gibbs free energy ( G o ads ), enthalpy ( H o ads ) and entropy ( S o ads ) of adsorption can be assessed using the following Eqs. ( 5), ( 6) and (7), respectively 38-40 : All the above-calculated adsorption parameters are mentioned in Table 4. The negative values of G o ads approves the spontaneous of the adsorption process and strong interactions between the inhibitor molecules and the metal surface. Besides, the high values of k ads revealed an effective adsorption process with high efficiency of the inhibitor molecules 41 . From Table 4, the absolute values of G o ads of the three inhibitors were − 35.7 and (3) Commonly, the values of G o ads are less negative than − 20 kJ mol −1 which belongs to physisorption (electrostatic interaction) between the inhibitor molecules and the metal surface. Conversely, negative values than − 40 kJ mol −1 indicate chemical adsorption, which is owed to adsorption of organic molecules on the metal surface accompanied by charge transfer to form chemical bonds. As above mentioned, the free energy of all studied inhibitors was in the range of − 20 kJ mol −1 to − 40 kJ mol −1 , indicating that the type of adsorption process belongs to the mixed adsorption type (including physisorption and chemisorption) 42,43 . Also as shown in Table 5, the values of G o ads were very close to − 40 kJ mol −1 and increased with temperature from (25 to 40) ºC which was attributed to the adsorption of carbonitrile derivatives on the steel surface dependent mainly on chemical absorption 44 . In the literature, the enthalpy of adsorption ( H 0 ads > 0) and close to (100 kJ mol −1 ) reflects the endothermic behavior of the adsorption. While it is exothermic at ( H 0 ads < 0), the enthalpy less (40 kJ mol −1 ) may involve either physisorption, or chemisorption, or a combination of both 45 . As seen in the current study, the calculated enthalpy H 0 ads values have positive signs and less (40 kJ mol −1 ), indicating that the endothermic nature of the adsorption process and two types of adsorption interactions are present on the steel surface. The values of S o ads are increased in a positive direction which is mainly typical of the endothermic adsorption process that is equivocally related to chemical adsorption. This increase can be attributed to the increased randomness due to the formation of the adsorbed layer of the inhibitors and the desorption of a high number of water molecules at the carbon steel/solution interface 46 . ## Effect of temperature. Temperature is a predominant factor in the corrosion inhibition process, in which corrosion is accelerated at elevating the solution temperature and affects the action of corrosion inhibitors. To investigate the effect of temperature on the dissolution of carbon steel in the presence and absence of different concentrations of carbonitrile derivatives, weight loss measurements were performed at varying temperatures ranged (25-40 °C). The variation of %η is plotted vs. the above-mentioned temperatures for PdC-OH derivative in Fig. 5, similar plots for PdC-Me and PdC-H are found in the supplementary material (Fig. S2). The corrosion rate (C.R) is obviously exaggerated and the inhibition efficiency increases with increasing solution www.nature.com/scientificreports/ temperature 43 . This action indicated that the investigated inhibitor molecules adsorbed on the steel/solution interface 30 . Moreover, the slight increase or constancy in the inhibition efficiency with increasing temperatures is attributed to the chemical adsorption of the inhibitor species only or is owed to a combination of chemical and physical adsorptions 47 . ## Thermodynamic kinetic parameters. To get more insights on the mechanism of the adsorption process of inhibitor molecules on the surface of the metallic material, the value of apparent activation energy E * a is frequently utilized to determine the type of adsorption mechanism. It can be calculated from the plot of the Arrhenius equation shown in Fig. 6, The Arrhenius plots of PdC-Me and PdC-H are found in the supplementary material (Fig. S3). Various parameters of the corrosion process based on Arrhenius equation ( 8) and the transition state theory equation ( 9) presented in the following equations 48 : where A is the pre-exponential factor (Arrhenius constant), H * is the activation enthalpy, S * is the activation entropy, T is the absolute temperature in Kelvin, h is the Plank constant, N is the Avogadro number and R (8 www.nature.com/scientificreports/ is the molar gas constant. As depicted in Table 6, the value of E * a in the blank solution is relatively higher than that in the presence of three inhibitors; this is ascribed to the phenomenon of chemical adsorption, whereas the opposite is true for physical adsorption 49 . The positive values of activation enthalpy ( H * ) indicate that the activation process is an endothermic corrosion process 41 . This phenomenon indicates that high temperature accelerates the corrosion process of carbon steel, and thus is attributed to chemical adsorption. The change in activation entropy S * increases in a positive direction inferring that the activated complex in the transition state is formed by association rather than dissociation process. In other words, more ordering will occur when the reactants convert to the active complex 50 . ## Potentiodynamic polarization technique. Polarization measurements were performed to understand the nature of electrochemical kinetics reactions. Figure 7 shows the polarization behavior of the carbon steel electrode in 1 M HCl in the absence and presence of various concentrations of the PdC-OH compound. The current potential plots of PdC-Me and PdC-H are found in the supplementary material (Fig. S4). Relevant electro- 51 . Equation (10) represents the correlation between inhibition efficiency and corrosion current density is represented as follows: where I o corr and I i corr represent the corrosion current densities of the uninhibited and inhibited solutions, respectively. All these parameters are tabulated in Table 7. As can be seen in Fig. 7, the cathodic and anodic Tafel lines are parallel upon adding these derivatives into acidic solution relative to the blank sample and have no substantial changes with each other. Thus, the adsorbed inhibitor merely hinders the active site of the anodic and cathodic reactions on the metal surface without affecting the actual corrosion mechanism, and only causes inactivation of part of the surface with respect to the corrosive medium 52 . Inspection the data in Table 7, the slopes of the anodic (β a ) and cathodic (β c ) Tafel lines slightly changed upon addition of these derivatives. It could be argued that these organic derivatives have the function of controlling the activation of hydrogen evolution and the anodic dissolution of the metal without any variation in the dissolution technique 53 . Another finding from Table 6 is that the E corr values for the inhibitory systems shifted to positive potential with the change in inhibitor concentrations less than 85 mV. This observation indicated that the studied derivatives behaved as mixed-type inhibitors and affect both the cathodic and anodic polarization curves 54 . As well, the maximum shift in the corrosion potential E corr with respect to E corr (blank) is more than 85 mV (observed for PdC-OH at 8 × 10 -5 M, PdC-Me at 6 × 10 -5 M, and PdC-H at 1 × 10 -4 M). This attributed to that the three derivatives can be classified as cathodic or anodic inhibitors according to the previous literature 55 . The decrease in the corrosion current density (i corr ) with the incremental concentrations of the investigated inhibitors leads to an increase in the inhibition efficiency. This phenomenon pronounced that the investigated inhibitors adsorbed on the active sites and formed a more stable layer on the surface of carbon steel 56 . As summarized in Table 7, the %η order is followed as PdC-OH > PdC-Me > PdC-H. However, the difference in %η between the three compounds was small which can be attributed to the similar structures of the three derivatives. ## Electrochemical impedance spectroscopy (EIS). The EIS or AC impedance technique provides important mechanical and kinetic information for the electrochemical system under study. Impedance measurements were utilized to evaluate the corrosion resistance of the carbon steel electrode in 1 M HCl solution in the absence and presence of different concentrations of the investigated derivatives at 25 °C. The impedance spectra include (Nyquist and Bode) plots for PdC-OH are shown in Fig. 8a,b, Nyquist and Bode plots of PdC-Me, and PdC-H are found in the supplementary material (Figs. S5 and S6). Obviously, In the Nyquist plot, the appearance of an individual capacitive loop is represented as a slightly depressed semi-circle. This capacitive loop indicates a nonideal capacitor performance at the metal/solution boundary phase 57,58 . Besides, the diameters of these capacitive loops increase significantly when the concentration of inhibitors in the test solution is increased without affecting their characteristic features. This is an indication that the adsorption of the studied inhibitors retards the corrosion of carbon steel in 1 M HCl without altering the electrochemistry of the corrosion process 59 . Moreover, for all tested inhibitors, the Bode-Phase plots showed a single peak within the studied frequency range, revealing that the impedance measurements were fitted in a one-time constant equivalent model with CPE . Furthermore, This may be attributed to the charge transfer process that occurs at the metal-electrolyte interface. The increase of the impedance modulus and the gradual increase of the phase angle maxima at the intermediate frequency with increasing concentration of the inhibitor were also shown 60 . This is due to more molecules being adsorbed on the surface of the electrode with increasing concentration and forming a protective layer as a barrier to the dissolution of the metal in acidic solutions. Besides, the phase angle values around 80, this deviation from the ideal corrosive system (phase angle = 90) is ascribed to the surface roughness as a result of both structural and interfacial origin 61 For a more in-depth understanding of the impedance spectra; A simple electrical equivalent circuit in Fig. 9 was employed to fit these experimental spectra 62 . The circuit consists of constant phase element (CPE) , charge transfer resistor (R ct ), and solution resistance (R s ) . Inhibition efficiency (η%) is calculated from R ct values using the following formula Eq. ( 11) 58 : where R 0 ct , R ct is the charge transfer resistance without and with inhibitor, respectively. However, R ct is a measure of electron transfer at electrode/electrolyte interface, which is inversely proportional to the corrosion rate 59 . Since the metal solution interface behaves as a double layer but without ideal capacitive behavior. Thus, the constant phase element (CPE) was used to improve the capacitance of the electrical double layer instead of utilizing the absolute capacitance double layer ( C dl ) . The impedance of the CPE (Z CPE ) can be calculated using Eq. ( 12) 58 : where Y O is the magnitude of the CPE , ̟ signifies the angular frequency (̟ = 2πf) , n is the CPE exponent which depends on the nature of the metallic surface representing the deviation from the perfect capacitive performance that value is between 0 and 1. j = (−1) 1/2 represents an imaginary number. The capacitance double layer ( C dl ) values of Y O and n are calculated as follows Eq. ( 13) 58 : where ̟ is the angular frequency when the imaginary component of the impedance is at its maximum value. The preceding impedance spectroscopy parameters were listed in Table 8, and it is worth mentioning that the values of R ct increased with increasing the concentration of the inhibitor and consequently the inhibition efficiency increased η%. This is an evidence of the effective corrosion protection of steel in the presence of the studied compounds. The dramatic drop in C dl values upon addition of the inhibitor is owed to the replacement of the water molecules by the inhibitor molecules adsorbed at the interface. Hence, the formation of an adherent film on the metal surface leads to the increase in the thickness of the electric double layer and /or the reduction in the local dielectric constant at the metal/solution interface 50 . These outcomes approve that carbonitrile derivatives exhibit good carbon steel inhibitory properties in acidic solutions. Furthermore, the corrosion inhibition efficiencies calculated from electrochemical impedance spectroscopy measurements presented in Table 8 shows an agreement trend with the calculated data from weight loss experiments and potentiodynamic polarization measurements. ## Electrochemical frequency modulation. Electrochemical frequency modulation technique is a power- ful tool for monitoring metal corrosion in aqueous solutions. The foremost advantages of EFM are its rabid and non-destructive properties when applied to the corroded electrode. EFM has a substantial feature since the corrosion current can be determined from small polarization and AC signals without prior knowledge of the Tafel constants 63 . Intermodulation spectra acquired from the EFM technique for evaluating different concentrations of the studied derivative (PdC-OH) versus corrosive media 1 M HCl on the steel electrode are characterized in Fig. 10, EFM spectra for PdC-Me and PdC-H are found in the supplementary material (Figs. S7 and S8). Each depicted spectrum represents two current response peaks appearing at 2 and 5 Hz intermodulation frequencies which were analyzed to obtain the relevant corrosion kinetic parameters 64 . Electrochemical parameters such as corrosion current density (i corr ) , Tafel constants ( β a and β c ), and the causality factors CF-2 and CF-3 were measured and listed in Table 9. It is found that the magnitude of i corr is suppressed by adding the investigated organic compounds in acidic solution and accordingly the inhibition efficiency is increased. This behavior indicates the effectiveness of the tested inhibitors due to the stability of the protective barrier layer on the steel surface 65 . The values of ( β a and β c ) were found to change with increasing concentration, therefore carbonitrile derivatives are (12) Quantum chemical parameters. Quantum chemical calculations were performed to inspect the effect of the structural and the electronic properties of the material on the corrosion inhibition performance. Besides, to gain insights into the donor-acceptor interactions between inhibitor molecules and metal atoms. According to the frontier molecular orbital theory FMO, analysis of the density distributions of HOMO and LUMO can determine the donation-acceptance ability and molecular reactivity of the investigated inhibitors. E HOMO denotes the ability of a molecule to donate electrons, whereas E LUMO represents the ability of a molecule to accept electrons. Figure 11 shows the optimized molecular structures, HOMO and LUMO electronic density distributions, respectively. All the calculated quantum chemical parameters are listed in Table 10 including the energy values for the molecular orbitals (HOMO and LUMO), the energy gap (ΔE), and the dipole moment (μ). Examination of Fig. 11, the electron density of the HOMO orbitals of the carbonitrile derivatives is concentrated over the entire pyridine ring, and the high distribution on the phenolic ring in the PdC-OH compound. The contribution of the whole molecules in electron transfer is attributed to the presence of high electron density throughout the π-electrons of the aromatic and pyridine moieties, which are the sites most susceptible to electrophilic attacks in the molecules. However, it can also be seen that the presence of groups (-NH 2 -C= O and -CN) in the pyri- www.nature.com/scientificreports/ dine ring is a relatively softer part of the molecules, mainly involved in electron transfer. Moreover, the LUMO electron density of the three derivatives is mainly spread over the pyridine ring and the aromatic rings. Thus, the tendency to accept charges accumulated on the metal surface increases to form a feedback bond between the donor iron atoms and the acceptor anti-bonding orbital of the inhibitor 63,67 . Therefore, the determination of the electronic density of the HOMO and LUMO orbitals revealed that the studied inhibitors could adsorb on the steel surface by donating π-electrons from pyridine and aromatic moieties (Nucleophilic attack) to the vacant d-orbital of the metal and another possibility is that these derivatives may adsorb by acquiring electrons from the metal surface (Electrophilic attack) 68 . In Table 9, the calculated values for E HOMO and E LUMO show that PdC-OH has a relatively higher value for E HOMO and a relatively lower value for E LUMO compared to PdC-H and PdC-Me. This finding indicates that PdC-OH has the highest propensity to adsorb on the carbon steel surface. It is generally believed that molecules with low E LUMO values and high E HOMO values tend to present better inhibition efficiency 69 . The energy band gap ΔE (ΔE = E HOMO − E LUMO ) 70 is the reactivity coefficient in theoretical studies. Murulana et al. postulated that the molecules which have a small energy gap value are considered highly reactive molecules and have good corrosion performance on the metal surface 71 . PdC-OH has the smallest value of ΔE as reported in Table 9. The dipole moment μ is a descriptor of the polarity in the covalent bond of molecules 71 . Abdallah et al. 72 alluded that the dipole moment is an indication of the electronic distribution in the molecule. The efficiency of corrosion inhibition increases with increasing value of μ, due to stronger dipole-dipole interactions with the metal surface resulting in strong adsorption and effective corrosion inhibition 73 . The µ values follow the order: PdC-OH > PdC-Me > PdC-H which may be attributed to the largest value of the adsorption preference for the inhibitor molecule on the metal surface. The DFT results by DMol3 in Table 9 showed that the compound with the OH Phenolic group was the most reactive compound of the tested group. This observation is consistent with previous experimental methods. Classification of these inhibitors according to their inhibition efficiency is PdC-OH > PdC-Me > PdC-H. ## Monte Carlo simulation. Monte Carlo simulations are performed to describe how inhibitor molecules behave and interact at a metal interface. This method can effectively help reduce the cost of experiment, and can also help in the experimental design of inhibitor molecules that effectively inhibit metal corrosion 74 . Figure 12 shows a side view and a top view of the optimized equilibrium configurations of the three carbonitrile derivatives adsorbed on a carbon steel substrate. Considering this output in Fig. 12, the geometrically optimized molecules under study being loaded on Fe surfaces indicated that these molecules tended to adsorb almost planar orientation on the surface. This flat orientation provides close contact with the active sites to impede the corrosion reaction 75 . The results in Table 11 give the output of energies computed by MC simulation such as total adsorption, adsorption energy, rigid absorption, and deformation energies. The outlined adsorption energy (E ads ) was calculated mathematically by the summation of the rigid adsorption energy and the deformation energy of the adsorbate molecules 76,77 . In general, E ads is used to express the strength of the adsorption process of inhibitors on the surface of iron. The rigid adsorption energy was described as the energy released when the inhibitor molecules are adsorbed on the metal surface and the deformation energy recognized as the energy released when the adsorbed-adsorbate components undergo relaxation on the surface 65 . As depicted in Table 11 the order of the calculated E ads values is PdC-OH (− 4073.86) > PdC-Me (− 4024.09) > PdC-H (− 4006.85), indicating the inhibition performance follows: PdC-OH > PdC-Me > PdC-H. Negative magnitudes of E ads for all three studied inhibitors denote the spontaneous and strong adsorption process that occurred on the iron substrate. As shown in Table 11, PdC-OH reached a maximum value of dE ads /dNi (− 208.85) in the simulation method indicating its highest contribution to the total adsorption energy 78 . This theoretical study supports previous experiments and proves coincident well with all acquired data. It can be concluded that all the three studied inhibitors can adsorb on the steel surface through the π-charge of pyridine and the moieties of the aromatic ring providing strong bonding to the metal surface. Finally, the use of carbonitrile derivatives as a corrosion inhibitor has a great potential for corrosion prevention, and consequently better inhibition efficiency. www.nature.com/scientificreports/ Atomic force microscopy (AFM) analysis. In the field of corrosion research, AFM analysis has been used to elucidate the effect of inhibitors on corrosion product development or corrosion progression at the metal/solution interface. The resulting topographical images directly reflect the surface of carbon steel at the nanometer scale 79 . AFM characterizes the morphology of the corroded metal in 3D images. Figure 13a shows a micrograph from AFM analysis of a carbon steel surface after immersion in 1 M HCl for 24 h in the absence of an optimal concentration of inhibitors 72 . The surface was severely damaged and corroded by acid attack, which was estimated by the average roughness (Ra) that recorded a height of 272.8 nm. Figure 13b shows the smooth and uniform surface of the free sample with Ra 49.8 nm 72 . However, the average roughness value was reduced to 152.11, 76.06, 64.55 nm after the treatment of the test solution with carbonitrile derivatives PdC-H, PdC-Me, and PdC-OH, respectively as shown in Fig. 13c compared to the absence of these inhibitors. The improvement shown in the surface topography is due to the adsorption of the investigated carbonitrile molecules onto the steel surface and the formation of a protective layer. In view of the above findings, the inhibition action tendency of inhibitors acquired from surface analysis is related to those acquired from experimental results. www.nature.com/scientificreports/ Attenuated total refraction infrared (ATR-IR) analysis. This method concerns identifying the adsorbed functional groups of the organic compounds upon the metal substrate, ATR-IR was performed with a range of 4000 to 400 cm −1 . Figure 14 signifies the ATR-IR spectrum of the PdC-OH compound and the construction of a protective film on the carbon steel surface after soaking for 24 h in 1 M HCl with the optimum concentration 10 -4 M of this compound. ATR-IR spectra for PdC-Me and PdC-H are shown in are found in the supplementary material (Figs. S9 and S10). It can be seen in Fig. 13 that the ATR-IR spectra of the protective film formed on the steel surface showed all the characteristic peaks of the pure inhibitor, indicating that the inhibitor adsorbed on the steel surface. The characteristic peaks of the active function groups of the free organic compound before (pure inhibitors) and the other peaks in the presence of this compound are discussed and summarized in Table 12. From the obvious peaks, function groups such as (Nitrile C≡N, -C = O and -NH-) appear on the carbon steel surface. Besides, there were small changes, and some frequencies of weak function groups were shifted significantly such as the peak of (C-O) Phenolic and stretching (C=C) aromatic as shown in Table 12. The above findings clearly illustrate that the specific hetero-aromatic ring (-OH) and the π electrons of the aromatic rings were involved in the adsorption process of the PdC-OH compound. X-ray photoelectron spectroscopy analysis. X-ray photoelectron spectroscopy (XPS) was performed to confirm the adsorption of the studied organic compounds on the carbon steel surface and to determine the chemical nature of the inhibitors/carbon steel interface. Figure 15 shows the high-resolution XPS spectrum survey obtained for the surface of corroded carbon steel in 1 M HCl solution in the presence of the PdC-OH derivative. The XPS spectrum shows complex forms, which were assigned to the corresponding species through a deconvolution fitting procedure. High-resolution XPS spectra obtained for carbon steel surface corroded in 1 M HCl composed of (Fe 2p, O 1s, Cl 2p, C 1s) are illustrated in Fig. 16. While in the presence of the studied carbonitrile compounds, the XPS spectra consisted of the same elements (Fe 2p, O 1s, Cl 2p, C 1s) in addition to N 1s core level as shown in Fig. 17. The XPS spectrum of Fe 2p shows six peaks, the higher peak at low binding energy (711.2 eV) corresponding to metallic iron 80 . The peak at 714.6 eV is attributed to Fe 2p3/2, and the small peak at 719.40 eV is ascribed to the Fe 3+ satellite 81 . In addition, the peaks at 724.3 eV ,and 727.9 eV can be attrib- . The second has a binding energy of 531.8 eV, related to hydroxide bonds chemisorbed on the surface 87,88 . The third peak, at 532.4 eV, can be assigned to the oxygen of the adsorbed water and OH − in FeOOH 89 . In the case of the inhibited samples, the oxygen spectrum shows three peaks at binding energies of 530.8 assigned to iron oxide, the peak at 532.8 for OH − in FeOOH, and the last peak at 534.2 corresponding to C-OH and surface adsorbed-water molecules. Moreover, in the presence of the investigated inhibitors, the O1s core level signal decreases significantly which is consistent with the adsorption of the inhibitors on the steel surface. Also, the XPS spectrum for Cl 2p shows the best fit in two components locate at around 198.9 eV for Cl 2p3/2 and 200.6 eV for Cl 2p1/2 as reported by Gu et al. 90 .The same peaks are obtained in the absence and presence of the studied inhibitors due to the arrival of some chloride ions to the surface and are responsible for the corrosion of the alloy. Finally, the XPS spectrum of N 1s appears with a single peak at 399.9 eV, and this peak can be attributed to the neutral imine (-N=) and amine (-N-H) nitrogen atoms as previously reported 91 . The appearance of N peak in the spectra of the protected sample surface confirms the adsorption of the studied inhibitors on the sample surface. According to the XPS results, we can conclude that the composite film formed on the surface contains iron oxide/hydroxide and carbonitrile compounds. These components provide a protective film that can effectively isolate the corrosion medium and reduce the corrosion of carbon steel. Possible corrosion inhibition mechanism. The adsorption process of organic inhibitor molecules depends on many physical and chemical properties such as; Electron density, chemical structure, metal nature, charges at the metal/solution interface, and type of aggressive medium (pH and/or electrode potential) 92,93 . www.nature.com/scientificreports/ These properties affect the mode the molecules interact on the metal surface. Adsorption of organic molecules on solid surfaces cannot be considered purely physical or chemical; A combination of both processes can occur in adsorption 83,84 . Physical interactions are accepted as the first step for adsorption of molecules on the metal surface and then chemical adsorption may occur via different charge sharing processes 94 . In a solution of hydrochloric acid, the inhibitor molecules are adsorbed on the metal surface through the following interactions: (i) the electrostatic interaction between the positively protonated inhibitor and the chloride ions adsorbed on the metal surface (physisorption process); (ii) the chemical interaction between the lone-pair electrons on the heteroatoms (N, S, O, P) and the unoccupied d-orbital on the metal surface (chemical adsorption process); (iii) the donoracceptor interaction between π-electrons of the aromatic ring and the vacant d-orbitals on the metal surface (chemical adsorption process); and (iv) Retro-donation interaction between the excess negatively charged metal surface and the π*anti-bonding of the inhibitor molecule. To elucidate the corrosion inhibition mechanism of the investigated carbonitrile derivatives, the inhibition action is due to adsorption of these compounds at the metal/solution interface. Adsorption may occur through a donor-acceptor interaction between conjugated π-charge of (two aromatic and pyridine rings moieties) and an unoccupied d-orbital of iron atoms to form coordination bonds (chemical adsorption process) 95 . The electron density of the donor atom in the carbonitrile functional group depends on the substituents present in these compounds. The direction of the inhibition potentials of three derivatives is determined by the value of Hammett sigma constant (σ) for the substituent groups (OH, CH 3 and H). This is because; in this type of derivatives the adsorption center is conjugated with the ring. The presence of electron donation (OH, CH 3 ) (σ = − 0.17 for p-CH 3 and σ = − 0.37 for p-OH) increases the electron density of the neighboring aromatic ring and makes the π-electrons more available for interaction with the C-steel surface thus strengthens the adsorption of PdC-OH and PdC-Me on the steel surface. PdC-OH has the highest inhibition efficiency, which is due to the presence of the OH group which added an extra adsorption center to the molecule compared to the CH 3 group in PdC-Me compound. PdC-H ranked below the two inhibitors in the inhibition order due to the presence of a hydrogen atom (H-atom with σ = 0.0) which is considered an electron-withdrawing atom 87 . Meanwhile, the carbonitrile derivatives can accept electrons from the d-orbital of iron atoms through their π* anti-bonding orbital to form a feedback bond (retro-donation process), thus promoting the adsorption of the inhibitor molecules on the steel surface. In addition, the carbonitrile molecules may be adsorbed on the metal surface through the van der Waals force by interacting neutral inhibitor molecules with iron ions to form [Fe-PdC] complexes (physical adsorption process). It can be concluded that the good inhibition efficiency of the studied derivatives is due to the presence of two aromatic and pyridine rings, the polar functional groups such as (-CN, -NH 2 ) that act as adsorption centers, and the specific hetero-aromatic ring (-OH) that present in PdC-OH compound. Hence, the type of adsorption of carbonitrile derivatives on the C-steel surface is more than just physical adsorption but not purely chemical . ## Conclusion According to the present study, the synthesized carbonitrile compounds can be used as effective corrosion inhibitors for C-steel in 1 M HCl. The corrosion inhibition efficiency increases with increasing concentrations of the studied inhibitors, and ranked as PdC-OH > PdC-Me > PdC-H. The adsorption of the inhibitors on the surface of C-steel in 1 M HCl follows the Langmuir isotherm and the calculated thermodynamic parameters propose that the adsorption is predominantly chemisorption. Addition of KI to the PdC-OH shows a synergistic effect that significantly improves its inhibition efficiency. The magnitudes of the synergism parameter (Sө) showed that the corrosion inhibition produced by PdC-OH and iodide mixture is synergistic in nature. Tafel polarization data showed that the corrosion current density decreases and the corrosion potential changes slightly with the addition of carbonitrile compounds, therefore these compounds are mixed type of inhibitors. The surface examined by ATR-IR, AFM, and XPS showed the formation of adsorbed film on C-steel surface. There is a good correlation between theoretical and experimental data.

chemsum

{"title": "Experimental and computational chemical studies on the corrosion inhibitive properties of carbonitrile compounds for carbon steel in aqueous solutions", "journal": "Scientific Reports - Nature"}

ambient_microdroplet_annealing_of_nanoparticles

## Abstract: Conversion of polydisperse nanoparticles to their monodisperse analogues and formation of organized superstructures using them involve post synthetic modifications, and the process is generally slow. We show that ambient electrospray of preformed polydisperse nanoparticles makes them monodisperse and the product nanoparticles self-assemble spontaneously to form organized films, all within seconds. This phenomenon has been demonstrated with thiol-protected polydisperse silver nanoparticles of 15 AE 10 nm diameter. Uniform silver nanoparticles of 4.0 AE 0.5 nm diameter were formed after microdroplet spray, and this occurred without added chemicals, templates, and temperature, and within the time needed for electrospray, which was of the order of seconds. Well organized nanoparticle assemblies were obtained from such uniform particles. A home-made and simple nanoelectrospray set-up produced charged microdroplets for the generation of such nanostructures, forming cm 2 areas of uniform nanoparticles. A free-standing thin film of monodisperse silver nanoparticles was also made on a liquid surface by controlling the electrospray conditions. This unique method may be extended for the creation of advanced materials of many kinds. ## Introduction Chemical processes in microdroplets are a rapidly evolving subject. Examples include synthesis of molecules 1,2 and pharmaceutical products 3 as well as conformational changes in proteins. 4,5 Such synthesis can also produce nanoparticles (NPs) without reducing agents starting from metal ions in solutions. This synthetic method can be tuned with the application of an electrical potential and can lead to assemblies of nanomaterials, and a viable method for forming 1D structures with potential applications was demonstrated. 7 Materials science with charged microdroplets can produce nanoholes on 2D materials 9 and metallic thin flms on liquid surfaces. 10 The science of nanomaterials, especially noble metal nanomaterials, has expanded into almost every area of materials science. The properties of these materials are heavily dependent on their size, shape, and distribution. These aspects are especially important for their electronic structure and consequently applications involving chemistry, physics, and biology. 14 As a result, several methods have been developed to control the size dispersity of such materials. Digestive ripening is one of the most commonly used methods in this regard for noble metal NPs. The method typically involves high-temperature annealing involving refluxing of NP suspensions for extended periods to achieve monodispersity. 15,18,19 However, extremely precise conditions for a long time, of the order of days, are required to obtain uniform particles. Therefore, it is important to develop a facile and fast method for making monodisperse nanostructures. Ultrafast acceleration of chemical reactions and synthesis of NPs starting from metal precursors in droplets suggest that new chemical bonds of diverse kinds can be formed and broken under such synthetic conditions. 1,2, This prompted us to explore the possibility of spontaneous dissociation and reassembly of preformed NPs in microdroplets. To our surprise, such dropletinduced dissociation and reassembly of polydisperse silver NPs protected with thiols resulted in highly monodisperse NPs in the microsecond time scale. The deposition of such monodisperse particles produced a flm of uniform NPs, and this process is millions of times faster than digestive ripening. Exploring the science through a series of control experiments showed that metal thiolates are transient precursor species formed in this process. We have also optimized conditions under which such a process is feasible to make cm 2 area flms of uniform particles. As this method is similar to high-temperature annealing, leading to monodispersity, we term it microdroplet annealing. ## Observation of microdroplet annealing In the present experiment, we have utilized a home-built nanoelectrospray source to deliver charged microdroplets containing polydisperse 2-phenylethanethiol (PET)-protected silver NPs (Ag@PET NPs) in dichloromethane (DCM) onto a transmission electron microscopy (TEM) grid placed on an indium tin oxide (ITO)-coated collector plate. The collector was grounded through a picoammeter to monitor the deposition current, and a potential in the range of 4.5-5.0 kV was applied to the solution held within a glass spray tip through a platinum (Pt) wire electrode. The spray plume was ejected from the nanoelectrospray tip, which can be visualized with a laser torch. Further details are available in the experimental section and ESI. † Various other substrates could be used in place of the TEM grid (see below). In the present work, an organized assembly of uniform NPs was formed on the TEM grid, starting from the corresponding polydisperse NPs. The as-synthesized Ag@PET NPs have a broad size distribution, as shown schematically in Fig. 1a. The particle size distribution covers a broad range, from 2-25 nm, as determined by TEM (Fig. 1b(i)). This distribution is typical of such NPs. Further details of the characterization of the starting material are presented in the ESI (Fig. S1 †). A d-spacing of 0.23 nm is due to Ag(111) and it suggested the growth of pure metallic NPs. 20 The characteristic Fourier-transform infrared spectroscopy (FTIR) features of the as-synthesized Ag@PET NPs confrm the attachment of PET on the AgNPs (Fig. S1c, ESI †). The as-synthesized Ag@PET NPs exhibit prominent surface plasmon resonance at 451 nm (Fig. S1d, ESI †). Details of synthesis and characterization are also presented in the experimental section and ESI. † Ag@PET NPs in DCM were transferred to a home-built borosilicate spray capillary of 50 mm inner diameter (ID) for microdroplet-induced reaction. A schematic representation of the home-built electrospray set-up is shown in Fig. 1. In the course of electrospray deposition of Ag@PET NPs, a dark circular spot of 0.5 cm diameter due to the impinging plume appeared on the substrate kept at a distance d from the spray tip, shown in the bottom panel of Fig. 1. The drastic change observed in the course of the reaction is shown in Fig. 1b. Positively charged microdroplets convert polydisperse Ag@PET NPs to an ordered 2D superlattice structure of uniform NPs. A superlattice in the current context refers to a localized periodic assembly of uniform NPs which possess inherent periodicity of the element. Such structures are possible only with uniform particles, and the Ag@PET NPs formed have a particle size distribution of 4.00 AE 0.50 nm, which is evident from the expanded portion of the TEM image (inset of Fig. 1b(ii)). From separate experiments it is known that droplet formation and deposition occurs in hundreds of microseconds to milliseconds. 3, Surprisingly, microdroplet chemistry via ambient electrospray converts irregular particles into a superstructure of uniform particles within a fraction of a second. Optimizing microdroplet annealing In the process of optimization, a series of trials were made to achieve ordered NP assemblies. Finally, ordered structures of Ag@PET NPs were obtained at an applied voltage of 5 kV, d of 1.5 cm and flow rate of 1.0 mL h 1 , and the data are shown in Fig. 2. A large-area view of monodisperse Ag@PET NPs is shown in Fig. 2a, which form a nearly continuous uniform 2D structure. The corresponding magnifed images are shown in Fig. 2b-f, which portray the orientation of the superlattice domains. The inter-particle gap in the superlattice structure was $1.2 nm (Fig. 2e), slightly less than the length of two PET ligands (the distance between S and Ph-C4 in the PET ligand is $0.7 nm (ref. 32)), suggesting ligand interactions driving the assembly. Ligands usually form interdigitated structures in these types of assemblies. 33 The Ag(111) lattice spacing of 0.23 nm (Fig. 2f) and the TEM-energy dispersive spectroscopy (TEM-EDS) data (Fig. S2 †) confrm the formation of crystalline AgNPs. The FTIR spectrum (Fig. S2b †) shows that PET protection remains intact after such ambient solution-state conversion. The reduced width of the optical absorption spectrum confrms the narrowness of particle sizes after electrospray (Fig. S2a †). This abrupt change in particle distribution postelectrospray suggests a digestive ripening-type of phenomenon through microdroplet chemistry. Inside the microdroplets, polydisperse Ag@PET NPs of size between 2 and 25 nm get reorganized and become uniform particles of 4.00 AE 0.50 nm size (inset of Fig. 2a). Images at different magnifcations (Fig. 2b-d) suggest that there is scope for the formation of a 3D superlattice assembly due to overlayer deposition. Negatively charged microdroplets did not produce such superlattice assemblies. Events within microdroplets can be modifed by altering d, under a particular electric feld. Monodispersity was achieved at d ¼ 1-2.5 cm at a spray voltage of 5 kV (Fig. S3b-d, ESI †). At shorter and longer d, polydispersity was seen, as shown in Fig. S3a and e-h, ESI †. This observation suggests that at d < 1 cm there was not enough time for the reorganization of the NPs in the microdroplets. At d > 2.5 cm, droplets were not producing NP assembly owing to the destabilization of microdroplets, due to evaporation of solvents. It could be demonstrated that when d was 1.5 cm, Ag@PET NPs achieved the perfect condition for self-assembly leading to superlattices. An optimum structure was obtained at 1.5 cm, and hence this d was maintained for other experiments. After fxing d, the electrospray voltage was optimized between 0.5 and 7.0 kV. In this process, the electric feld and deposition current of microdroplets were adjusted accordingly. The monodisperse assembly of Ag@PET NPs was obtained at 5.0 kV (Fig. S4e, ESI †). The rate of flow of Ag@PET NPs from the spray emitter was also an essential parameter. The size of microdroplets depended on the flow rate. For the given condition, microdroplets produced ordered structures at a flow rate of 1.0 mL h 1 , as shown in Fig. S5b, ESI. † Solvent plays a vital role in the generation of a suitable environment within the microdroplets. The generation of the spray-plume and materials formed in microdroplets are highly dependent on the polarity of the solvent. The dielectric constant (3), pH, and pressure of the droplet environment are important electrospray parameters. 31,34 In this regard, solvents with different 3 were studied under the above-optimized conditions to check the effect of solvents on the formation of such structures within microdroplets. Monodispersity was achieved for a few solvents having a particular range of 3 under optimized conditions in the case of PET-protected silver NPs, as shown in Fig. S6c-g. † Organized assembly was observed when such a process was continued with DCM having an 3 of 8.93. Other solutions, with solvents such as carbon tetrachloride, 3 ¼ 2.24, diethyl ether, 3 ¼ 4.33, chloroform, 3 ¼ 4.81, tetrahydrofuran, 3 ¼ 7.82, pyridine, 3 ¼ 12.40, acetone, 3 ¼ 20.70, dimethylformamide, 3 ¼ 37.50, and acetonitrile, 3 ¼ 38.25 were unable to produce such monodisperse assembly. This study was continued with four different capillaries with IDs in the range 30-60 mm under the optimized conditions. An ID of 50 mm was most appropriate under these conditions. ## Extension to other ligands It is clear that an unusual phenomenon in microdroplets was observed with PET protected AgNPs under a particular condition. In order to check the possibility of such transformations with other protecting ligands, trials were made with different ligands, especially thiols with different denticities and chain lengths under the optimized conditions achieved for Ag@PET NPs. Different polydisperse AgNPs were synthesized with 2,4dimethylbenzenethiol (DMBT) and water-soluble sodium citrate (Cit), and their TEM images are shown in Fig. S7a and c, ESI. † Previously optimized conditions did not produce monodispersity in these cases (Fig. S7b and d, ESI †). Polydisperse dithiol-protected AgNPs were synthesized with 1,4-benzenedithiol (BDT) and 1,6-hexanedithiol (HDT), and spray experiments were performed subsequently. In the case of Ag@BDT NPs, monodispersity without such assembly was seen (Fig. S8a and b, ESI †). However, spray of Ag@HDT NPs was inefficient even to produce a signifcant change in the particle size, as presented in Fig. S8c and d. † These experiments were extended to comparatively long-chain thiols like octadecanethiol (ODT)-and dodecanethiol (DDT)-protected AgNPs. A broad size distribution was observed after the spray of ODT and DDT-protected AgNPs (Ag@ODT NPs & Ag@DDT NPs), which may be due to the difficulty in reorganization of long carbon chain thiols within microdroplets (Fig. S9, ESI †). Optimization of spray conditions is necessary to get monodisperse assemblies for these NP systems. However, ambient microdroplet annealing produced monodisperse assembly of Ag@DMBT NPs at 8.5 kV and d ¼ 1.5 cm, as shown in Fig. S10. † The method was also successful in creating uniform NPs of ethanethiol (ET)protected Ag NPs from polydisperse NPs at an applied voltage of 4.0 kV and d ¼ 1.0 cm (Fig. S11 †). Therefore, it is clear that this process is general and is applicable for different NP systems under specifc electrospray conditions. Optimization of spray conditions is essential for a particular NP system. The concentration of polydisperse Ag@PET NPs taken for the spray experiments is also important. Fig. S12 of the ESI † shows the TEM images of Ag@PET NPs after spraying at different concentrations. These experiments demonstrate that uniformity is dependent on the starting concentration of the polydisperse Ag NPs. In summary, monodispersity needs optimization of a range of parameters. ## Nebulization spray The aerosol spray can also be produced by nebulization. 31,34 This process does not require an applied electric potential. Ag@PET NPs (Fig. 3a) were sprayed using dry N 2 gas with four different pressures, using a microcapillary of 50 mm diameter. A schematic of this spray set-up is presented in Fig. 3b. Microdroplets produced in the spray can create uniform particles with an average particle size of 4 nm, as shown in Fig. 3c. Thus, monodispersity can be achieved by voltage-free spray as well. However, the spray does not produce uniform ordered superstructures, which may be due to the high-pressure of N 2 gas from the emitter. This may require additional optimization of the pressure needed. Fig. S13 † shows the TEM images of Ag@PET NPs after spray at different N 2 nebulization pressures. For all these cases, a nearly similar particle size distribution was observed. Fig. 3 Generation of monodisperse Ag@PET NPs from polydisperse Ag@PET NPs using microdroplets. (a) TEM of the as-synthesized Ag@PET NPs, (b) schematic representation of the spray of polydisperse Ag@PET NPs using dry N 2 gas without an electric field; the ID of the silica microcapillary was 50 mm, the same as the spray tip used in the electrospray. (c) TEM image of monodisperse Ag@PET NPs after spray (size distribution is in the inset), (d) schematic of the formation of soluble Ag-PET thiolates and their electrospray at three different d values, (e-g) TEM images of Ag@PET NPs obtained at a d of 0.50 cm, 0.75 cm, and 1.00 cm, respectively. Nearly ordered assemblies are formed from thiolates at d ¼ 0.75. An expanded view of the same image is shown in the inset of f, showing the Ag(111) lattice. ## Thiolate intermediates A probable mechanism for the formation of such organized structures was hypothesized with inputs from control experiments. We know that Ag + and thiols in solution form thiolates during the formation of NPs 35,36 and clusters. 37,38 Silver thiolates are formed when clusters decompose. 39,40 Therefore, we propose that silver thiolates are formed in droplets as intermediates during the formation of uniform NPs. To prove this, we conducted a spray experiment by taking Ag-PET thiolates as precursors (see the experimental section for their synthesis), instead of polydisperse NPs. Thiolates were sprayed at three different d values (Fig. 3e-g). All the experiments produced silver NPs, but the organized structures with silver lattice planes were seen at d ¼ 0.75 cm (Fig. 3f). This suggests that microdroplet spray of solvated polydisperse particles resulted in highly ordered assemblies, most likely through a transient thiolate species, as presented in Fig. S14. † ## Large-area lms This ambient ion-based been introduced for the creation of free-standing metallic monodisperse nanoparticles-nanosheets (NPs-NSs). In an earlier study, we electrodeposited metal salts on a liquid surface and got selforganized flms of the particles under the influence of electrohydrodynamic flow. 10 A similar approach was used here. The set-up, schematically represented in Fig. 4a is the same as that shown in Fig. 1, except that multiple nozzles and a liquid substrate were used. The multi-nozzle electrospray enlarges the magnitude of the NPs-NSs on the liquid substrate. In this objective, a series of trials were performed using different solvents. It was observed that among the solvents studied, the ordered NP structure was formed on the water surface. These uniform NPs-NSs, composed of superlattice structures of silver NPs, can be used for diverse applications. The large-area TEM image in Fig. 4b confrms the compactness of the assembly. The particle size distribution was in the range of 4.5 AE 0.5 nm. The formation of a brown colored thin flm of NPs-NSs was observed and is shown in the inset of Fig. 4b. Fig. 4c shows the corresponding magnifed image. The lattice spacing of these NPs-NSs matches with the (111) plane of AgNPs, proving the metallic nature of AgNPs. Similarly, polydisperse NPs were deposited on an ITO surface under optimized spray conditions and were redispersed in DCM. A uniform flm of monodisperse NPs was observed using a TEM, as shown in Fig. S16 (ESI †). Polydisperse silver NPs can also form 3D assemblies on a metal substrate as shown in Fig. 5a. If we look closely at the structures, a layer-upon-layer assembly of nanoparticles is observed, which is indicated by arrows in Fig. 5b. During spray for an extended period, the incoming microdroplets with the transient metal-thiolates are continuously reduced to NPs and get deposited on preexisting nanoparticle layers. 7 The 3D crystal structures formed this way could generate exceptional properties. 41,42 Fig. S17 in the ESI † shows the growth of a multilayer assembly with respect to time. The developed superlattice structures can be used as an ink for electrospray-based printing of a range of crystalline materials. This ambient approach can also print bulk composites and porous architectures. The simplicity of Fig. 4 (a) Schematic illustration of the electrospray deposition of Ag@PET NPs on a water surface. The formation of a thin film of Ag@PET NPs on a liquid surface is also shown, (b) a large-area TEM image of monolayer Ag@PET NPs from the deposited film (inset: photograph of the free standing film deposited on water after spray and its particle size distribution), (c) HRTEM image of NPs which confirms the ordered structure and formation of metallic Ag particles. The well-defined Ag(111) lattice is shown in the inset of c. this ambient process is expected to create interest for the development of such 3D structures directly from polydisperse NPs. A series of control experiments were performed to optimize the ordered structure by varying different parameters, especially, spray distance (d), voltage and flow rate. The experiment was repeated under the optimized conditions. It was confrmed that an electric feld and a deposition current of $35 nA were also vital for the creation of such nanostructures. This methodology may provide an opportunity to produce large scale monodisperse silver NP assemblies via such ambient microdroplet annealing. Moreover, this ambient technique could certainly be used as a facile synthetic approach for the development of unique nanomaterials of other materials with emerging properties for a broad range of applications. The possibility of assembly over large areas along with uniform synthesis offers new possibilities. ## Conclusions In summary, we developed a fast method of making monodisperse silver NPs by electrospraying highly polydisperse NPs synthesized in solution, under ambient conditions. Microdroplets from the spray formed a uniform assembly of nanostructures composed of ordered AgNPs upon deposition on a substrate. Control experiments proved that the electric feld, the tip-to-collector distance, and the flow rate are the key factors for the oriented growth of such superlattices. This process does not require any sophisticated instrumentation, and it transforms polydisperse NPs to superlattices composed of monodisperse NPs within seconds. This process may be considered green as no solvents or additional processing is involved, unlike solution state post-processing methods. Microdroplet spray on a water surface makes a thin flm of monodisperse metallic sheets of uniform NPs assisted by electrohydrodynamic flow. The method described in this work may be utilized for the development of multimetallic superlattice structures or high entropy alloys by efficient control of the composition of metals, leading to new properties. ## Materials and chemicals used Silver nitrate (AgNO 3 ) was purchased from RANKEM. Sodium borohydride (NaBH 4 ), 2-phenylethanethiol (PET), 1,4-benzenedithiol (BDT), 2,4-dimethylbenzenethiol (DMBT), sodium citrate (Cit), ethanethiol (ET), 1-octadecanethiol (ODT), and dodecanethiol (DDT) were purchased from Sigma Aldrich Chemicals. Pure ethanol, methanol, dichloromethane, pyridine, carbon tetrachloride, acetone, diethyl ether, chloroform, acetonitrile, and N,N-dimethylformamide were sourced from Merck India and used as solvents for the electrospray deposition experiments. ## Synthesis of PET protected silver NPs (Ag@PET NPs) Polydisperse Ag@PET NPs were synthesized by our previously described method with slight modifcations. 33 Briefly, PET (0.58 mL) was added to pure methanol (30 mL) at 25 C. Subsequently, AgNO 3 (50 mg) in Millipore water (0.5 mL) was added to the above solution. The mixture was gently stirred for 15 min to form silver thiolates. After that, NaBH 4 (25 mg) was dissolved in 8 mL ice-cold water and was subsequently added slowly into the flask and stirred for another 12 h to allow the complete growth of Ag@PET NPs. The as-synthesized Ag@PET NPs were washed with methanol and extracted in different solvents for further experiments. ## Synthesis of other ligand protected silver NPs Different ligand protected AgNPs were synthesized for comparative experiments. BDT, DMBT, ET, HDT, Cit, ODT, and DDT were used separately during the synthesis of AgNPs in individual trials. For every case, the experimental procedure was similar, except for the addition of a particular thiol in methanol at the beginning of the synthesis. After synthesis, the products were washed with methanol and extracted in DCM for electrospray experiments. ## Electrospray deposition experiments A home-made electrospray set-up was used for the generation of microdroplets (Fig. 1a). 7 A micropipette puller (P-97, Sutter Instruments, U.S.A.) was used for pulling a borosilicate glass capillary of 1.5 mm outer diameter (OD) and 0.86 mm ID. It was cut into two pieces having tips of 50 mm ID and 150 mm OD. Each tip of the capillary was checked using an optical microscope to ensure the size and quality of the cut. Polydisperse Ag@PET NPs (100 mg mL 1 ) in DCM were flled in the nanoelectrospray tips using a microinjector pipette tip, and Pt wire (0.5 mm diameter) was inserted into the solution, making an electrode for high voltage connection. A multi-nozzle spray setup was used for scaling up the method. A syringe pump controlled the flow rate of the microdroplets to achieve the best structures. The syringe needle was connected to a high voltage power source through a copper clip. The experiments were carried out in different solvents and solvent mixtures having different 3 and at varying potentials. Among the solvents used, spray with DCM showed the best result. More control experiments were carried out with silver thiolates in DCM. ITO-coated glass slides were procured from Aldrich. Millipore water was used in all the experiments. ## Microdroplet generation through nebulizer gas For nebulization spray, a home-built set-up was used as presented in the schematic (Fig. 3b). A Hamilton syringe was connected to a silica capillary (50 mm ID) through a union connector. Microdroplets were generated by spraying solvents through a fused silica capillary of 50 mm ID with 10, 20, 30, and 40 pounds per square inch (psi) N 2 gas for nebulization. ## Characterization of the nanostructures TEM and HRTEM were performed at an accelerating voltage of 200 kV on a JEOL 3010, 300 kV instrument equipped with a UHR polepiece. A Gatan 794 multiscan CCD camera was used for image acquisition. Energy-dispersive X-ray spectroscopy (EDS) spectra were collected on an Oxford Semistem system housed on the TEM. The formation of NPs during microdroplet deposition was examined directly using 300-mesh carbon-coated copper grids (spi Supplies, 3530C-MB) under different experimental conditions. Silver NPs were deposited directly on the TEM grids under different experimental conditions. Particle size distributions were obtained from TEM images using ImageJ.

chemsum

{"title": "Ambient microdroplet annealing of nanoparticles", "journal": "Royal Society of Chemistry (RSC)"}

electrochemical_hydroxylation_of_c3–c12_n-alkanes_by_recombinant_alkane_hydroxylase_(alkb)_and_rubre

## Abstract: An unprecedented method for the efficient conversion of C 3 -C 12 linear alkanes to their corresponding primary alcohols mediated by the membrane-bound alkane hydroxylase (AlkB) from Pseudomonas putida GPo1 is demonstrated. The X-ray absorption spectroscopy (XAS) studies support that electrons can be transferred from the reduced AlkG (rubredoxin-2, the redox partner of AlkB) to AlkB in a two-phase manner. Based on this observation, an approach for the electrocatalytic conversion from alkanes to alcohols mediated by AlkB using an AlkG immobilized screen-printed carbon electrode (SPCE) is developed. The framework distortion of AlkB-AlkG adduct on SPCE surface might create promiscuity toward gaseous substrates. Hence, small alkanes including propane and n-butane can be accommodated in the hydrophobic pocket of AlkB for C-H bond activation. The proof of concept herein advances the development of artificial C-H bond activation catalysts. Artificial direct oxidation on terminal C-H bond of linear alkanes leading to primary alcohols in a highly selective manner is acknowledged to be challenging . On the contrary, alkane monooxygenase (AlkB), an integral membrane-bound diiron ω-hydroxylase obtained from the Gram-negative bacterium Pseudomonas putida GPo1 (formerly known as Pseudomonas oleovorans) found in nature, can regio-selectively introduce molecular oxygen onto the unreactive terminal methyl group of C 5 −C 12 linear alkanes to yield primary alcohols and water in the presence of reducing equivalents (eq. 1). Taking advantage of this protein, P. putida GPo1 is able to utilize linear medium-chain length alkanes (C 5 −C 12 ) as the sole source of carbon and energy . n 2n 2 2 n 2n 1 2 AlkB belongs to a family of integral membrane non-heme diiron proteins including integral membrane fatty acid desaturase (for instance, integral membrane stearoyl-acyl carrier protein ∆ 9 -desaturase, SCD1) and xylene monooxygenase (XylM) 10 . Nine conserved histidine residues are found as their featured motif. This class of non-heme diiron proteins is distinct from soluble non-heme diiron proteins and their diiron active sites are ligated by multiple histidine residues. A comparison between soluble and membrane bound non-heme diiron proteins as well as the sequence alignment data among AlkB, SCD1 and XylM are illustrated in Supplementary Information, Tables S1 and S2 to reveal their functional features as well as structural and biochemical characteristics. Besides AlkB, a soluble rubredoxin-2 (AlkG) and a soluble NADH-dependent rubredoxin reductase (AlkT) are also required to mediate the transfer of reducing equivalents for the activation of linear alkanes . However, AlkB, just like many other membrane proteins, is difficult to be isolated and purified into homogeneity, and its activity is hard to be maintained outside of the bilayer membrane . Hence, there is a lack of detailed structural information. To date, the metal active-site, the environment around the active-site and the framework of this protein are still elusive 22 . Based on studies using diagnostic probes, it is known that a long-lived radical intermediate is generated during the catalysis mediated by AlkB . The specific activity for the direct conversion from n-octane to 1-octanol is determined to be 26.6 min −1 14 ; whereas the NADH consumption rate is within the range of 2.0-5.2 U/mg AlkB (one unit of activity is defined as the oxidation of 1 μmol of NADH per minute; i.e. around 90-230 min −1 ) . To explore the structural properties and the electron transfer pathway of AlkB, X-ray absorption spectroscopy (XAS) is employed to reveal the environment close to the active site of AlkB and AlkG, respectively. Based on the information obtained, a novel electrochemical hydroxylation of C 3 -C 12 n-alkanes mediated by membrane-bound AlkB through an AlkG immobilized screen-printed carbon electrode (SPCE) is developed (Fig. 1). So far, SPCE has been extensively used for numerous purposes including analysis in evironmental assessment, food processing and pharmaceutical industry. One particular application is to serve as biosensors for the detection of biomolecues via direct electron transfer (DET) . For instance, the detection of glucose via DET using a glucose oxidase immobilized SPCE was reported 30 . Recently developed SPCE with better operating performance significantly reduces the intrinsic barrier of DET from the electrode surface towards the redox center of proteins, that is usually hampered by the relatively thick molecular layer with poor conductivity as well as low electron transfer efficiency 31 . It is expected that DET can be conducted on a SPCE even with highly insulated membrane-bound AlkB. This study demonstrates the electrochemically controlled redox-driven functionalization of chemically inert alkanes by AlkB on a SPCE as well as the possibility in the future chemical transformation through similar systems. ## Construction and purification of AlkB and AlkG proteins. Constructed vectors, pET21alkBStrep and pACYCDuetalkG, were generated from the insertion of alkB and alkG genes into pET-21a(+) (Novagen) and pACYCDuet-1 (Novagen), respectively. To follow the preparation of Roujeinikova, A. et al. 21 , a Strep II tag (Trp-Ser-His-Pro-Gln-Phe-Glu-Lys) and a linker (Ser-Ala) were added to the C-terminus of AlkB within the constructed vector pET21alkbstrep. The shuttle vectors were then transformed into E. coli BL21 (DE3) and expressed. The integrity of expressed AlkB and AlkG was demonstrated by peptide mass fingerprinting analysis of trypsin-digested protein fragments of respective proteins that appeared at 45 and 23 kDa on the SDS-PAGE. The Strep II-tagged AlkB was purified into homogeneity via a Strep-Tactin column (GE Healthcare) in the presence of the detergent, n-dodecyl-N,N-dimethylamine-N-oxide (LDAO). The purified AlkG was obtained via a HiLoad Q Sepharose Fast Flow column (GE Healthcare) followed by a Phenyl Sepharose HP column (GE Healthcare). The UV-vis spectrum for the purified oxidized AlkG exhibits characteristic peaks at 378 and 495 nm, respectively (Figure S1(a) in Supplementary Information). The protein content of AlkB was quantified using Enhanced Chemiluminescence (ECL) of Western blotting against Strep-tag; whereas that of AlkG was determined by UV-vis absorption at 378 and 495 nm. Also, iron quantification conducted by inductively coupled plasma optical emission spectroscopy (ICP-OES) demonstrates that the iron content of purified AlkB and AlkG are 1.5 ± 0.4 and 1.6 ± 0.4 per molecule, respectively. The data is consistent with previous findings that AlkB is a diiron monooxygenase 18 with an anti-ferromagnetically coupled high-spin Fe III − Fe III evidenced by Mössbauer studies, whereas AlkG contains two [Fe − (CysS) 4 ] cores characterized by spectroscopic methods 32,33 . X-ray absorption spectroscopic study of AlkB and AlkG proteins. Fe K-edge X-ray absorption near edge structures (XANES) of AlkB and AlKG are displayed in Fig. 2(a), providing the information of iron oxidation states and the geometry of iron atom within the proteins. Enlarged pre-edge region (inset of Fig. 2(a)) illustrates the difference in XANES among the purified AlkB (alternatively termed AlkB oxidized from this point on in the text), the fully oxidized AlkG (AlkG oxidized ) and the fully reduced AlkG (AlkG reduced ). The mid-point energies of rising edge, defined as K-edge energies in the following text (confer Table 1), for AlkB oxidized and AlkG oxidized absorption appear to be 7125.7 eV and 7120.9 eV, respectively (Fig. 2(a) and Table 1). The K-edge energy of FeCl 3 located at 7126.2 eV suggests that the purified AlkB (AlkB oxidized ) is in the ferric state. In addition, the K-edge energies obtained from several iron-sulfur cluster containing proteins in the oxidized form, such as succicinate:caldariellaquinone oxidoreductase (SdhC) 34 , phthalate dioxygenase (PDO) 35 and superoxide responsive protein (SoxR) 36 , are 7119.3 eV (the energy calibration is set with Fe foil at 7111.3 eV), 7120.6 eV (the energy calibration is set with Fe foil at 7111.2 eV) and 7120.2 eV, respectively. Moreover, both K-edge energies for the oxidized form of non-heme diiron hydroxylase in soluble methane monooxygenase (sMMO), MMOH ox , and its complex with another component in the system, MMOH ox + MMOD, are 7124.7 eV (the energy calibration is set with Fe foil at 7111.1 eV) 37 . These data imply that the purified AlkB is indeed in ferric state. It has been known that AlkG can be reduced either by NADH in the presence of rubredoxin reductase, or sodium dithionite 17 . In this study, it is found that at least 20 eq. of dithionite is required to completely abolish the UV-vis feature of AlkG oxidized . Normalized XANES profile of AlkG reduced gives the K-edge energy of 7119.6 eV (Fig. 2(a) and Table 1). This result is comparable with the K-edge energies of several iron-sulfur cluster containing proteins in the reduced form, such as PDO 35 and SoxR 36 , which are 7118.7 eV (the energy calibration is set with Fe foil at 7111.2 eV) and 7119.3 eV, respectively. Although it is shown that AlkG oxidized can be fully reduced using a high dose of dithionite, AlkB oxidized cannot be entirely reduced by dithionite (Figure S2). The treatment of AlkB oxidized with 10 eq. of dithionite only slightly leads to the shift in the K-edge energy of −0.4 eV. Instead, AlkB oxidized can be reduced in the presence of AlkG reduced . The titration of dithionite into AlkB oxidized -AlkG oxidized mixture in 1:1 ratio was performed and the shift to lower energy in XANES profile supports that AlkB oxidized can be reduced by AlkG reduced (Fig. 2(b)). The XANES spectrum of AlkB in the reduced form (AlkB reduced ) can be obtained after the subtraction of XANES spectrum of AlkG reduced from that of the AlkB reduced -AlkG reduced adduct (Fig. 2(a), red line). The obtained K-edge energy of AlkB reduced is 7122.3 eV, which is comparable to that of FeSO 4 (7122.2 eV) (Fig. 2(a) and Table 1). The K-edge energies for MMOH red and the hydroxylase of toluene monooxygenase, TMOH red , are 7121.5 eV and 7121.8 eV (the energy calibration is set with Fe foil at 7111.1 eV for both proteins), respectively 38 , which are quite close to the obtained K-edge energy from AlkB reduced . To decipher the redox chemistry occurred in between AlkB-AlkG pair, the linear combination of experimental XAS data obtained from the individual XANES profile was conducted (Figure S3 in Supplementary Information). The spectral change deduced from the linear combination suggests that two phases are involved in the electron transfer mediated by AlkB-AlkG pair. During the first phase observed through the linear combination of XANES spectra among AlkG reduced , AlkG oxidized and AlkB oxidized , all AlkG oxidized are fully reduced to AlkG reduced prior to the reduction of AlkB oxidized (Figure S3(d)). During the second phase observed through the linear combination of XANES spectra among AlkG reduced , AlkB oxidized and AlkB reduced , the transition from AlkB oxidized to AlkB reduced in the presence of AlkG reduced is noted (Figure S3(e)). The merged XANES profile from two phases (Figure S3(f)) agrees with the experimentally given XANES profile obtained from AlkB-AlkG (1:1) mixture (Fig. 2(b)). The results herein imply that AlkG serves as a redox partner to transfer electrons towards AlkB through the transition in the order of AlkB oxidized AlkG oxidized → AlkB oxidized AlkG reduced → AlkB reduced AlkG reduced (Fig. 1). The strong pre-edge absorption at around 7113 eV for the iron-sulfur complex in both AlkG oxidized and AlkG reduced is assigned as the 1 s→3d transition, reflecting the extent of d-p mixing that satisfies the selection rule The integrated area is reported in unit, which is the real value multiply by 100 . b ND: Not detected due to the poor resolution of the spectrum after linear combination. c K-edge energy 41 (eV) is defined herein as the energy at half its maximum normalization intensity for the rising edge of normalized XANES spectra. (Fig. 2(a), inset) 34 . Pre-edge absorption peaks obtained from AlkG oxidized and AlkG reduced were normalized and their integrated areas were calculated following an established procedure, resulting in the integrated areas for AlkG oxdized and AlkG reduced of 25.0 units and 17.5 units, respectively (one unit is 10 −2 eV; Fig. 3 and Table 1) 41 . The integrated area for pre-edge absorption peaks is acknowledged to be correlated with the coordination number of iron center. Model compound studies show that, for high spin Fe III model complexes, the range for the normalized integrated areas are 6-9 units, 12-19 units and 20-25 units for 6-, 5-and 4-coordinated complexes, respectively 41 . For high spin Fe II model complexes, the range for the normalized integrated areas are 4-6 units, 8-13 units and 16-21 units for 6-, 5-and 4-coordinated complexes, respectively 39 . The obtained integrated areas for AlkG . The decreased integrated area of AlkG reduced , for ca. 70% of AlkG oxidized , is attributed to the higher occupancy of 3d orbitals in Fe II relative to that of Fe III . Moreover, the difference in the K-edge energy between the oxidized state and the reduced state of AlkG is 1.3 eV (Table 1), which is smaller than the data obtained from Rieske-type clusters (ca. 2 eV) 35 but are closer to the one electron reduction of [2Fe-2S] cluster in SoxR (0.9 eV) 36 . This result suggests that the reduction of AlkG from Fe III to Fe II may not dramatically affect the charge of iron centre for its coordination with cysteinyl sulfur atoms. In addition, the Fe extended X-ray absorption fine structure (EXAFS) data of AlkG merely show a minute difference between the Fe-S bond length (or Fe-S backscattering in distance) of AlkG oxidized (2.25 ) and AlkG reduced (2.28 ) (Figure S4 and Table S3 in Supplementary Information). This is consistent with the earlier EXAFS study indicating that the reduction of rubredoxin yields only a slight increase of average Fe-S bond length by ca. 0.05 42 . The lack of Fe-Fe backscattering suggests that the complex possesses a tetrahedron-like Fe-4S core. Similar to common iron-sulfur proteins that participate in electron transfer, e.g. SoxR 36 , the reorganization energy for the redox switch from AlkG oxidized to AlkG reduced is minute, which is facile for electron transfer from AlkG to AlkB. Furthermore, the pre-edge energies (integrated intensities) for SdhC oxidized and PDO oxidized are 7113.0 eV (the energy calibration is set with Fe foil at 7111.3 eV; 24.8 units) and 7113.1 eV (the energy calibration is set with Fe foil at 7111.2 eV; 26.0 units), respectively; whereas that of PDO reduced (Fe II -Fe III ) is 7112.5 eV (18 units) 34,35 . The obtained pre-edge energies from AlkG oxdized and AlkG reduced are 7113.7 eV and 7112.8 eV, respectively (Table 1), which are fairly consistent with the reported values obtained from aforementioned iron-sulfur proteins in their respective oxidation states. The integrated pre-edge area of AlkB oxidized is 11.2 units, which suggests that iron geometry in AlkB oxidized is either in a 5-coordination or a distorted 6-coordination environment. The 1s→3d transition can be further resolved into two peaks at 7114.1 eV and 7115.9 eV with the integrated intensity of 6.7 and 4.5 units, respectively (Fig. 3(b)). The pre-edge profile of AlkB oxidized is reminiscent of distinct pre-edge signals for diiron model complexes such as ([FeOH(H 2 O)Chel] 2 (H 2 O) 4 and [FeOH(H 2 O)Dipic] 2 with an intense feature at the low-energy but a shoulder at the high-energy 40 . The pre-edge profiles of MMOH ox and the complex, MMOH ox + MMOD, also give two resolved peaks with featured intensities of 7.0 (1.2) units and 7.5 (0.3) units at 7113.1 eV as well as 1.1 (0.1) units and 2.1 (0.7) units at 7114.6 eV, respectively 37 . The slight inconsistency both in the position and intensity of pre-edge peaks is due to the fact that, in the study of MMOH ox and MMOH ox + MMOD, the energy calibration is set with Fe foil at 7111.1 eV, which is 0.9 eV shorter. In addition, the normalization of XANES spectra for MMOH ox and the corresponding model studies were selected at 7130 eV with an edge jump of 1. Therefore, the pre-edge intensity of AlkB oxidized at around 7114.5 eV has to be lowered down by 1.1 times according to the present employed method 41 . The pre-edge intensities or areas observed for other non-heme diiron proteins such as metHrN 3 (~10.4 units; Hr: hemerythrin), soluble stearoyl-acyl carrier protein Δ 9 -desaturase (10.8 units or 11.5 units) and RNR R2 met (~10.1 units; RNR R2: ribonucleotide reductase subunit R2) suggest that AlkB oxidized exhibits the features of non-heme diiron proteins with the iron active sites in diferric state 43,44 . Similar to mammalian integral membrane stearoyl-CoA desaturase (SCD1), AlkB contains four histidine-containing motifs (two HXXHH motifs, one HXXXH motif and one NXXH motif) that presumably consist of the ligands to coordinate the bimetallic centre (see sequence alignment in Supplementary Information). Recently revealed SCD crystal structure indicates that the distance between those two metal ions (Zn) is 6.4 and each metal ion is associated with four to five histidine residues including the TM4 motif (N 261 XXXH 265 ) 45 . To obtain the local structural information for the metal active site of AlkB oxidized , EXAFS data analysis was conducted and the resulting k 3 -weighted EXAFS raw data (k 3 χ, i.e. the radial distribution function), are displayed in Fig. 4. To employ the mammalian SCD crystal structure as a model system, seven possible fitting results are listed in Table 2. The first-shell data fitting (2.0 < R < 3.0 ) to the Fourier transforms of k 3 χ mainly arises from the backscattering from 3-5 ligated N/O atoms with the average distance of 2.02-2.03 . In Fit 5, a Fe-O/N of 2.40 with the goodness of fit, R fit , of 1.6% is considered acceptable but poorly fitted. Previous data of RNR R2 met show that, except for the short Fe-O' (bridged-oxo), the average length of Fe-O/N bonds for the first coordination shell is 0.07 shorter than that of metHrN 3 (2.13 ) 46 . Since Fe-O bonds are much shorter than Fe-N bonds in high-spin ferric complexes, this result suggests that at least one histidine on the iron in metHrN 3 is replaced by an oxyanion ligand in RNR R2 met . The shorter Fe-O/N bonds of AlkB than that of histidine ligated metHrN 3 (~0.10 shorter) is also observed for the inner shell coordination and that may be attributed to the participation of an oxyanion ligand in the coordination. Besides, previous studies demonstrate that the average Fe-O/N bond length of MMOH ox or MMOH ox + MMOD is within the range of 1.99-2.01 37,47 . The second-shell data fitting to the Fourier transforms of k 3 χ (3.0 < R < 4.0 ) yields 1 × Fe-Fe of 3.08-3.11 (Fits 1, 2, 5 and 7) and 3 × Fe-C α from histidine with the average distance of 2.95-3.17 (Fits 3, 4, 6 and 7), along with all the fitting results considered acceptable (R fit < 2.0%). The third-shell data fitting can be extended to give 1 × Fe-C α of 4.10 (Fit 2) or 1 × Fe-Fe of 4.07 (Fit 4). Since Fe III -Fe III distance for most non-heme diiron enzymes is within the range of 3.0-3.4 (Table S1 in Supplementary Information), it is surmised that the active site of AlkB contains two irons with Fe III -Fe III distance of 3.08-3.09 and each iron coordinates with 3-4 histidines (Fits 1, 2 and 7 in Fig. 4 and Table 2). Among Fits 1, 2 and 7, the obtained Fit 7 with the additional second-and third-shells giving 3 × Fe-C α of 2.95 and 3 × Fe-C/N of 4.08 could be more promising in revealing the geometry and the structure for the iron active site of AlkB. Previous multi-shell restricted fits for non-heme diferric iron proteins, such as metHrN 3 and RNR R2 met , yield 1 × Fe-Fe of 3.19 and 3.22 , respectively, which is ca. 0.10 longer than that value obtained from AlkB 46 . In addition, previous fits for metHrN 3 and RNR R2 met result in the second-shell backscatters of 3.05 (4.2 × Fe-C α ) and 3.03 (3.1 × Fe-C α ) as well as third-shell backscatters of 4.33 (5.1 × Fe-C/N) and 4.30 (3.8 × Fe-C/N), respectively, which are 0.09 (in the case of metHrN 3 ) and 0.24 (in the case of RNR R2 met ) longer than the value obtained from Fit 7 of AlkB 46 . Based on the results that Fe-O/N bond length obtained from the first-shell fitting of AlkB is 0.03-0.10 shorter than that of metHrN 3 and RNR R2 met (vide supra), the active site of AlkB seems to be with a more congested coordination environment, presumably, similar to the active site of MMOH ox that with a short first-shell Fe-O/N bond (ca. 2.0 ) and short Fe-Fe distance (ca. 3.0 ) 37,47 . However, the possibility that the average distance of Fe-Fe backscatter in the purified AlkB may not be resolved by EXAFS study cannot be excluded. The third-shell data fitting in Fit 6 reveals 3 × Fe-C/N bonds with an average distance of 4.17 that might be resulted from the imidazole rings. Plus, in Fit 6, there is no Fe-Fe backscattering observed and this model is more consistent with the result observed from the three-dimensional structure of SCD1 with a dimetallic center of metal-metal distance >6 . In summary, the obtained seven Fits of EXAFS analysis (Fig. 4 and Table 2) may directly or indirectly contribute to the understanding of the iron active site in AlkB. Further studies, including single crystal X-ray crystallography, Mössbauer and/or Resonance Raman Spectroscopy, are definitely essential to unravel the three-dimensional structure and the non-heme diiron active site features of membrane-bound AlkB protein in order to comprehend how AlkB executes medium-chain length n-alkane oxidation. ## Electrochemical properties of AlkB and AlkG proteins. XANES studies show that fully reduced AlkG can efficiently reduce AlkB towards a higher reduced state (Fig. 2(b), vide supra). Based on the redox sequence concluded from XANES studies, it is expected that the direct supply of electrons to AlkG oxidized using an electrochemical approach followed by the shuttling of electrons to AlkB oxidized can further catalytically oxidize linear alkanes to primary alcohols. Fully reduced AlkG can be prepared electrochemically through a direct electron transfer (DET) process on a disposable SPCE (with the surface roughness (Ra) of 2.02 nm) immobilized with AlkG. It is noted that oxidized AlkG immobilized on a SPCE shows the UV-vis absorption features at 410, 495 and 595 nm (Figure S1(b) in Supplementary Information). As shown in Fig. 4(a), low charge transfer resistance (R ct = 1.81 kΩ) of SPCE enables the DET from the immobilized AlkG to AlkB. The reduction and oxidation potential of AlkG are determined to be −515 and −266 mV, respectively, in a quasi-reversible manner (Fig. 4(b)). The surface concentration of electroactive AlkG on a SPCE is estimated to be 0.057 × 10 −9 mol/cm 2 based on the observed cyclic voltammogram (CV) 48 . CV of recombinant AlkB-enriched membrane obtained from E. coli without further purification gives redox peaks at −452 mV and −285 mV after deoxygenation (Fig. 4(c)). In the presence of oxygen, obvious electrocatalytic reduction is observed, indicating that AlkG and AlKB-enriched membrane immobilized on a SPCE are electrocatalytic active. The surface coverage for AlkB-enriched membrane on a SPCE is estimated to be 1.06 × 10 ─9 mol/cm 2 , which is higher than theoretical monolayer coverage (i.e. 1.89 × 10 −11 mol/cm 2 ) 49 . The ΔE p,1/2 for the redox peak of AlkB-enriched membrane is calculated as 36.9 mV, which is close to an ideal Nernstian adsorbate layer with the transfer of two electrons under Langmuir isotherm conditions (i.e. ΔE p,1/2 = 45.3 mV) 50 . Presumably, this result is attributed to the transition of two-electron transfer from Fe III Fe III → Fe II Fe II in the absence of substrates. Results show that AlkG on the surface of electrodes can be maintained at the reduced state when the operation potential is kept at −0.6 V (vs. Ag/AgCl) for the catalytic turnover mediated by AlkB-enriched membrane. It is expected that the AlkG immobilized electrode can interact with suspended AlkB-enriched membrane micelles to form an AlkB-AlkG complex and then efficiently convert C 5 -C 12 linear alkanes to the corresponding primary alcohols. To carry out the electrochemical conversion, first, the surface of SPCE is coated with a layer of Nafion ® . Then, an aliquot of AlkG solution (0.13 nmol/cm 2 ) is immobilized on the top of electrode via physical adsorption. To optimize the conversion from n-octane to 1-octanol, the electrochemical reaction is suspended by varying the concentration of AlkB-enriched membrane at 43-690 μM in 50 mM Tris buffer solution, pH 7.5. No activity is observed from AlkB-enriched membrane using a SPCE without AlkG (Figure S6 in Supplementary Information). It is found that the highest specific activity can be achieved if operated within 30-min duration. The obtained specific activity in turnover frequency (TOF) per AlkB-AlkG complex (min −1 •protein −1 ) is related to the actual concentration of AlkB bound with the immobilized AlkG on the SPCE. A one-site binding model can fairly describe the change in the specific activity with the concentration of AlkB bound to AlkG, yielding the coefficient of determination (R 2 ) of 0.984, the maximum activity of 1,100 min −1 •protein −1 as well as K d of 268 μM (data not shown). However, if a two-site binding model is employed (Fig. 5), the change in the specific activity with the concentration of AlkB bound to AlkG of 322 and 776 min −1 •protein −1 can be obtained. Both dissociation constants, K d , from AlkB-enriched membrane are determined to be 268 μM. Alkane hydroxylation via electrocatalysis. The effective surface coverage ratio of AlkG for DET from a SPCE to AlkG is within 0.5-2% 48 . Data obtained from catalytic conversion mediated by SPCE electrodes show that the apparent TOFs (based on the amount of AlkG placed on top of SPCE instead of electroactive AlkG, i.e. the effective surface coverage ratio of AlkG) of C 5 -C 12 range within 250-1000 min −1 per unit AlkG, where the concentration ratio of AlkB to AlkG is about 2 (Fig. 6; confer Figures S6 and S7 in Supplementary Information for GC chromatograms). The specific activities (in TOF) for the conversion of C 6 , C 7 and C 9 n-alkanes to primary alcohols are in the range of 500-1,000 min −1 (Fig. 7). The TOFs for the conversion of C 5 , C 8 and C 10 −C 12 n-alkanes to corresponding primary alcohols are within 250-400 min −1 . Herein, unprecedented results for the conversion of gaseous propane (C 3 ) and n-butane (C 4 ) to 1-propanol and 1-butanol, respectively, are noted; TOFs for the gas-to-liquid (GTL) conversion are 16 ± 2 and 106 ± 6 min −1 , respectively. ## Discussion The analysis of XANES spectra obtained from AlkG oxidized , AlkG reduced , AlkB oxidized and AlkB-AlkG pair yields the plausible electron transfer pathway employed by AlkB through the transition of AlkB oxidized AlkG oxidized → AlkB oxidized AlkG reduced → AlkB reduced AlkG reduced , further facilitating the hydroxylation of n-alkanes (Fig. 1). This revelation allows the development of a system for artificial n-alkane hydroxylation using AlkG immobilized SPCE, in which AlkT is replaced by the electrode. Therefore, the reaction can be controlled via the supply of electrons; both AlkT and the costly reducing equivalent are no longer required. The kinetics data suggest that one AlkG with two Fe-4 S electron transfer centres can bind with two membrane-bound AlkB to form an AlkG:2AlkB adduct, facilitating an efficient conversion of n-octane to 1-octanol. With a high concentration of AlkB-enriched membrane binding to AlkG, conformational changes or molecular assembly (presumably, dimerization), might not greatly vary the dissociation constant of AlkB-AlkG adduct. However, the catalytic efficiency driven by AlkB-AlkG adduct with excess of AlkB significantly improves, i.e. more than two-fold. This result agrees with the observation that the ratio for the expressed AlkB to AlkG in P. putida GPo1 and E. coli under the regulation of alk operon are within the range of 1:0.34 and 1:0.08-0.10, respectively 16 . The alkane hydroxylation activity given by P. putida GPo1 is usually higher than that obtained from the alkBGT expression in E. coli. It is noted that differences in the specific activity among all chain length linear alkanes tested are varied; there is no particular relationship between the specific activity and the chain length of alkanes. This phenomenon may be attributed to that the recombinant AlkB is heterogeneously expressed on the membrane of E. coli and may cause the promiscuity in the selection of substrates 16 . The adsorption of the protein complex on the SPCE with high surface area and its intrinsic properties may also alter the shape or the morphology of the hydrophobic pocket in AlkB so that the relationship between the reaction activity and the chain length is difference from that observed from the wild type host cells. It has been shown that P. putida GPo1 can metabolize propane and n-butane 51 . In addition, several mutations implemented in recombinant AlkB in vivo demonstrate the feasibility in the conversion of n-butane to 1-butanol 52 . In this study, the electrocatalytic conversion of gaseous propane and n-butane to 1-propanol and 1-butanol, respectively, is shown. To date, there is no direct chemical method reported for the efficient oxidation of C 3 −C 4 n-alkanes to the corresponding primary alcohols. The catalytic process presented is environmentally benign and, in principle, can combine with the solar energy/photovoltaic cell for the sustainable conversion of either natural gas or LPG to the corresponding primary alcohols. Presumably, other small oxygenated products, such as propylene oxide or butene oxide, that are classified as fundamentally basic chemicals for the future supply in chemical industry can be generated through a similar approach 53,54 . In this study, spectroscopic and electrochemical evidence is presented to show that an iron-sulfur protein, AlkG, can efficiently transfer electrons towards the non-heme iron monooxygenase, AlkB, for the subsequent conversion of medium-chain length n-alkanes to primary alcohols. Immobilized AlkG on SPCE can interact with the AlkB-enriched membrane to form a complex for efficient conversion of C 5 −C 12 n-alkanes to primary alcohols with the specific activity in TOF of 250-1000 min −1 . Structural distortion of protein complex upon the adsorption on the surface of electrode might cause the promiscuity towards substrates, leading to the interaction with propane and n-butane. The proof of concept in this study provides the future feasibility in the production of green alcohols that can potentially serve as feedstock for chemical industry. ## Methods Materials. All chemicals were purchased from Sigma-Aldrich and were used as received unless otherwise specified. The gas cylinders of propane and n-butane were obtained from Huei Chyi Gas Co., Ltd. The vectors pET-21a(+) and pACYCDuet-1 were purchased from Novagen. Escherichia coli strain DH5α used as a carrier for the plasmid constructs and strain BL21 used for expression of plasmid constructs were purchased from Bioman Scientific Co. Ltd. (Taiwan). Primers were synthesized by Tri-I Biotech Inc. (Taiwan). UV-visible spectra were recorded on a HP 8453 diode array spectrometer and the path length for the UV-vis cuvette is 1 cm. All electrochemical experiments were carried out using a CH Instrument electrochemical workstation (CHI 611 C). The three-electrode system contains a SPCE (5 mm in diameter and 0.196 cm 2 in area; Zensor R&D, Taiwan), an Ag/ AgCl reference electrode and a platinum wire counter electrode. The working electrode was anodized prior to use by applying a potential at 2.0 V versus Ag/AgCl. The iron quantification was conducted by Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES 720S) (Agilent Technology). In the series of C 5 −C 12 alkane/ alcohol analysis, gas chromatography (GC, Agilent HP6890 plus) was equipped with a flame ionization detector (FID) and compounds administered were separated by a HP-5 capillary column (60 m × 0.25 mm × 0.25 μm film thickness). The analysis conditions were presented as the following unless stated otherwise: carrier gas, nitrogen, was at 1.0 mL/min; for the activity assay using octane as the substrate, the oven temperature was set isothermally at 120 °C for 10 min and then raised to 150 °C with the rate of 5 °C/min and hold for another 4 min; the injection port was at 300 °C in splitless mode. In the series of C 3 and C 4 alkane/alcohol analysis, gas chromatography (GC-2010 Plus, Shimadzu) was equipped with a barrier discharge ionization detector (BID) and compounds administered were separated by a DB-624 capillary column (30 m × 0.53 mm × 3 μm film thickness). The analysis conditions were presented as the following unless stated otherwise: carrier gas, helium, was at 1.0 mL/min; the oven temperature was set isothermally at 30 °C for 10 min and then raised to 200 °C with the rate of 10 °C/min and hold for another 5 min; the injection port was at 280 °C in split mode. ## Construction of AlkB and AlkG protein. The genomic DNA of P. putida GPo1 (ATCC 29347) was extracted using Gene-Spin Genomic DNA kit (Protech). The PCR amplification of the construct alkB was achieved by using Deep VentR DNA polymerase (New England Biolabs) and the following four primers: 5′-AAT TCA TGC TTG AGA AAC ACA GA-3′; 5′-CAT GCT TGA GAA ACA CAG AGT TCT-3′; 5′-TCG AGA TAC GAT GCT ACC GCA-3′ and 5′-GAT ACG ATG CTA CCG CAG-3′. The cohesive-end alkB gene with EcoR I and Xho I site was generated, cloned into the corresponding restriction sites of pET-21a( + ) and then sequenced. The Strep II tag (WSHPQFEK) and a linker (SA) were added to the C-terminus by a two-step PCR using constructed plasmid pET21alkB as a template with first pair of primers: 5′-GGA ATT CCA TAT GCT TGA GAA ACA CAG AGT TCT GG-3′ and 5′-TCA TTT TTC GAA CTG CGG GTG GCT CCA AGC GCT CGA TGC TAC CGC AGA GGT ACT-3′. The resulting product was used as template for the second pair of primers: 5′-CCG CTC GAG TCA TTT TTC GAA CTG CGG GT-3′ and 5′-GGA ATT CCA TAT GCT TGA GAA ACA CAG AGT TCT GG-3′. After the digestion with restriction enzymes Nde I and Xho I, the end product was cloned into the corresponding sites of pET-21a(+) to produce the gene construct of pET21alkBStrep and then sequenced to confirm its integrity 55 . The rubredoxin-2 gene, alkG, was amplified from the genomic DNA using the following four primers: 5′-CAT GGC TAG CTA TAA ATG CCC G-3′; 5′-GCT AGC TAT AAA TGC CCG GA-3′; 5′-AAT TCT CAC TTT TCC TCG TAG AGC-3′ and 5′-CTC ACT TTT CCT CGT AGA GCA C-3′. The cohesive-end alkG gene with Nco I and EcoR I sites was generated, cloned into the corresponding restriction sites of pACYCDuet-1 and the pACYC-DuetalkG construct was then sequenced. The plasmid of pET21alkBStrep or pACYCDuetalkG was transformed into E. coli BL21 (DE3). The E. coli transformant of pET21alkBStrep was grown in Luria-Bertani (LB) medium containing 100 mg/L of ampicillin at 37 °C in a 10 L fermentor. When the turbidity of cell culture monitored by OD 600 reached 0.7, 60 mg/L of FeSO 4 was supplemented and 0.4 mM of isopropyl ß-D-thiogalactopyranoside (IPTG) was added to induce the overexpression of AlkB-Strep. The temperature was then lower down to 20 °C and cell cultures were incubated for another 22 hr. The E. coli transformant of pACYCDuetalkG was grown in LB medium containing 34 mg/L of chloramphenicol at 37 °C in a 10 L fermentor. When OD 600 reached 0.7, 60 mg/L of FeSO 4 was supplemented and 0.4 mM of IPTG was added to induce the overexpression of AlkG. The temperature was then lower down to 20 °C and cell cultures were incubated for another 22 hr. ## Purification of AlkB protein. All procedures for the isolation of AlkB-enriched membrane and the purification of AlkB were performed at 4 °C. Cells were resuspended in 50 mM Tris-HCl containing 10 μg/mL of DNaseI, pH 7.5, and disrupted in a pre-chilled French pressure cell (SLM Aminco) via three passages at 20,000 lb/ in ref. 2. The lysate was subjected to centrifugation at 12,000 rpm for 20 min to remove cell debris. Subsequently, the supernatant was subjected to ultracentrifugation at 36,000 rpm for 1 hr and the AlkB-enriched membrane portion was collected after the ultracentrifugation. To purify Strep-tagged AlkB, the collected membrane portion together with the pellet portion was solubilized in 50 mM Tris-HCl, 300 mM NaCl, 1 mM FeSO 4 , 0.23% N,N-dimethyldodecylamine N-oxide (LDAO, Sigma), pH 7.5, homogenized and then stirred in a beaker for 1 hr. Followed by the solubilization, the membrane extract was diluted with 50 mM Tris-HCl, 300 mM NaCl, 1 mM FeSO 4 , pH 7.5, to reduce the concentration of LDAO to 0.07%. The membrane extract was subjected to centrifugation at 12,000 rpm for 20 min and the supernatant was loaded onto a Strep-Tactin column (GE Healthcare) running by gravity flow. The unbound was washed with five bed volumes of 50 mM Tris-HCl, 300 mM NaCl, 0.07% LDAO, pH 7.5, and proteins were eluted with two bed volumes of the same buffer containing 0.5 mM D-desthiobiotin (Figure S8 in Supplemental Information). The AlkB-containing fractions were pooled and concentrated using a concentrator (Vivaspin 20, MW cutoff 30 kDa, GE Healthcare). Protein concentrations were determined through Bio-Rad DC protein assay (Bio-Rad) using bovine serum albumin as the standard. The purified AlkB can be stored at 4 °C for a day and the activity diminishes afterward. Purification of AlkG protein. All procedures were performed at 4 °C. Cells were resuspended in 50 mM Tris-HCl containing 1 mM of phenylmethylsulfonyl fluoride (PMSF) and 10 μg/mL of DNaseI, pH 7.5, and then disrupted in a pre-chilled French pressure cell (SLM Aminco) via three passages at 20,000 lb/in2. The lysate was subjected to ultracentrifugation at 36,000 rpm for 1 hr. Ammonium sulfate was slowly added to the supernatant until the final concentration reaches 20% and then centrifuged at 6,000 rpm for 15 min to remove the contaminating proteins in the precipitate. The clear reddish solution was concentrated and dialyzed against 1 L of 50 mM Tris-HCl, pH 7.5, for three times to remove the excess salt. The resulting solution was loaded onto a HiLoad Q Sepharose Fast Flow column (GE Healthcare) equilibrated with 50 mM Tris-HCl, pH 7.5. An ascending linear gradient of sodium chloride (0-1 M) was applied and AlkG-containing fractions were pooled. Ammonium sulfate was added into the pooled solution until the final concentration reaches 35% and then centrifuged at 6,000 rpm for 15 min to remove the contaminating proteins in the precipitate. The clear reddish solution part was loaded onto a Phenyl Sepharose HP column (GE Healthcare) equilibrated with 50 mM Tris-HCl, 35% ammonium sulfate, pH 7.5. A descending linear gradient of ammonium sulfate (35-0%) was applied and eluted fractions containing active AlkG were pooled. The purified AlkG was dialyzed overnight against 50 mM Tris-HCl, pH 7.5 and then concentrated. The concentration of AlkG was determined using the characteristic absorbance at 378 and 495 nm with the extinction coefficient of 12,400 and 10,600 M −1 cm −1 , respectively. ## Western blot analysis. Detection of Strep-tagged AlkB was achieved by Western blot analysis using Strep-Tactin HRP conjugate against the Strep tag. Proteins from membrane extracts were separated by SDS-PAGE and then transferred onto a hydrophobic PVDF membrane (0.45 μm Hybond-P, GE Healthcare). Strep-tagged AlkB was detected using Strep-Tactin HRP conjugate (IBA GmbH), which recognizes the C-terminal Strep tag attached on AlkB, following the manufacturer's instruction. The resulting bands were visualized with the luminescence elicited from ECLTM Western blotting reagent (Bionovas Inc.) and analysed using a CCD camera system (UVP Inc.). The quantitation of the recombinant AlkB was assessed using the calibration curve established by Strep-tagged marker proteins (Bionovas Inc.) 56,57 . ## Determination of the iron content in AlkG and Strep-tagged AlkB by ICP-OES. Aliquots (0.2 mL) of protein samples were dissolved in 4.8 mL of saturated 65% HNO 3 solution (suprapure grade, Merck). The samples were digested in a MARS5 microwave digestion system (CEM Inc.). The temperature was stepped up incrementally from room temperature to 180 °C in 15 min, and maintained at 180 °C for another 15 min. The process of nitrate digestion was then terminated as the temperature was gradually lowered down to room temperature. The digested samples were then diluted with 20 mL doubly distilled water (Millipore) prior to the analysis. The iron concentration of samples was determined by interpolating a linear plot of a series standard solutions of Fe(NO 3 ) 2 in 0.10 N HNO 3 . A solution of 0.10N HNO 3 in distilled water was used as the iron-free control. The digested samples were measured by ICP-OES 720S (Agilent Technology). Fabrication of purified AlkG modified electrode. The purified recombinant AlkG was immobilized on the surface of a bare SPCE. The SPCE with a working area of 0.196 cm 2 and a conductive track radius of 2.5 mm was purchased from Zensor R&D (Taichung, Taiwan). The measured average resistance is 85.64 ± 2.10 Ω/cm. The redox potential of AlkB-enriched membrane and purified AlkB solution were determined. The AlkB-enriched membrane is easier to be employed and exhibits more efficient capability in the electron transfer to oxidize a series of C 5 −C 12 n-alkanes. The SPCE was pre-treated by gently washed with deionized water and then air-dried prior to the modification by protein. Purified AlkG modified electrode was prepared by drop-coated 10 μL of 4.436 mM purified AlkG protein on a bare SPCE and air-dried for 3 hr. Then, 10 μL of 0.5% Nafion ® /MeOH solution was coated onto as-prepared AlkG protein modified electrode and the modified electrode was subsequently stored at 4 °C for 1 hr. Electrochemical hydroxylation of n-alkanes, determination of apparent turnover number (TON) and turnover frequency (TOF). Electrochemical measurements were performed with a CHI 611C electrochemical workstation in a three-electrode cell assembly. A Nafion ® /AlkG modified SPCE working electrode, an Ag/AgCl, 3 M KCl reference electrode and a platinum auxiliary electrode were used to complete the cell setup. The Nafion ® /AlkG modified SPCE was employed in the electrochemical hydroxylation of C 3 -C 12 n-alkanes (C 5 −C 12 : 0.4-1.1 mM; C 3 and C 4 : 1.0 mL condensed liquid by dry ice) under a potential of −0.6 V versus Ag/AgCl in the presence of 427 μM AlkB-enriched membrane in the electrochemical cell. In the series of C 5 −C 12 alkane/alcohol analysis, after the conversion, the reaction mixtures were transferred to a 1.5-mL Eppendorf tube and extracted with 0.005% p-xylene in 100 μL methylene solution, in which p-xylene was served as an internal standard. Aliquots of 0.050, 0.10, 0.21, 0.41 and 0.82 mg of primary C 5 −C 12 alcohols were dissolved, respectively, in 100 μL of p-xylene containing methylene chloride. The corresponding GC intensities of primary alcohols were normalized by the intensity derived from the internal standard, p-xylene. In the series of C 3 and C 4 alkane/alcohol analysis, methyl t-butyl ether (MTBE) was employed as an internal standard with the addition of 1 μL MTBE into 1 mL aqueous solution. The calibration curve was established by adding aliquots of 0.15, 0.29, 0.44, 0.59, 1.76, 2.93 and 5.85 mg of primary C 3 −C 4 alcohols into ddH 2 O (1 mL) containing 0.1% MTBE. The amount of product (primary alcohols) formation is determined through gas chromatography using the established calibration curve. The apparent turnover number (TON) is calculated by dividing the mole of product formation by the mole of AlkG, as the limiting factor is AlkG. The apparent turnover frequency (TOF) is obtained by dividing TON by the duration of electroanalysis, i.e. 30 min. X-Ray absorption spectroscopy. X-ray absorption data were collected at the National Synchrotron Radiation Research Center in Hsinchu, Taiwan (Beamline Wigger 17C1) using a Si(111) double crystal monochromator in the region of the Fe K-edge (the energy calibration is set with Fe foil at 7112 eV). Protein samples were loaded onto either a sample holder (1.4 cm × 1.4 cm × 0.2 cm) covered with sheets of kapton or sealed in a polyethylene bag. During measurements, the samples were kept at 277 K (cooling by cold air device). The data collection was carried out in combination with the sleeping mode (i.e., without shining x-ray on the sample) for 4-18 s after every 6-24 s of X-ray irradiation. Samples tested were sufficiently thin to allow the total penetration of incident X-ray beam. Fluorescence data were collected on an argon-filled ionization chamber. Under these conditions, the edge jump could be regarded as a measure for the content of the corresponding absorbing element in the sample. The background subtraction and normalization of XAS data were implemented in the program ATHENA. The pre-edge peak areas or intensities were calculated followed the method established by Roe, A. L. et al. 41 , and the peak was isolated from the normalized XAS spectra after the subtraction of a sigmoidal function from a nonlinear curve fit of the raw data using the program OriginPro 8.6.0 (OriginLab Co.). After background subtraction, the area of pre-edge peak was obtained from the peak integration at 7108-7118 eV (AlkG oxidized and AlkG reduced ) or 7110-7120 eV (AlkB oxidized ). The pre-edge peak of AlkB oxidized at ca. 7114.5 eV was further deconvoluted into two Gaussian functions 36 using "Multiple Peak Fit" and "Simulate Curve" of OriginPro 8.6.0. It is noteworthy that many previous characterization studies using XANES in iron-sulfur and non-heme iron proteins including SdhC, PDO, MMOH, TMOH do not usually provide the detailed data to specify the mid-point K-edge and pre-edge energy. To extract the corresponding information from the original XAS spectra in the reported literatures 34,35,37,38 , the converted image files from the pdf documents were digitalized and reproduced by a function of "Digitize Image" in OriginPro 8.6.0. The obtained raw data were further processed and normalized to obtain the K-energy at the midpoint of absorption edge as well as the pre-edge peaks were identified by the same program. For all samples presented in this study, no photoreduction or photodamage was observed from the comparison between the first and the last spectra collected from a given sample. The fitting of experimental extended X-ray absorption fine structure (EXAFS) data was done by a nonlinear least-square fitting algorithm implemented by the IFEFFIT program . Data fitting quality was evaluated with the goodness-of-fit factor defined as: where χ = k 3 χ and n is the number of evaluations of f i , with χ χ =   f i i i data model (and hence R fit ) minimized in the nonlinear least-square fitting algorithm.

chemsum

{"title": "Electrochemical Hydroxylation of C3\u2013C12 n-Alkanes by Recombinant Alkane Hydroxylase (AlkB) and Rubredoxin-2 (AlkG) from Pseudomonas putida GPo1", "journal": "Scientific Reports - Nature"}

relation_between_molecular_electronic_structure_and_nuclear_spin-induced_circular_dichroism

## Abstract: The recently theoretically described nuclear spin-induced circular dichroism (NSCD) is a promising method for the optical detection of nuclear magnetization. NSCD involves both optical excitations of the molecule and hyperfine interactions and, thus, it offers a means to realize a spectroscopy with spatially localized, high-resolution information. To survey the factors relating the molecular and electronic structure to the NSCD signal, we theoretically investigate NSCD of twenty structures of the four most common nucleic acid bases (adenine, guanine, thymine, cytosine). The NSCD signal correlates with the spatial distribution of the excited states and couplings between them, reflecting changes in molecular structure and conformation. This constitutes a marked difference to the nuclear magnetic resonance (NMR) chemical shift, which only reflects the local molecular structure in the ground electronic state. The calculated NSCD spectra are rationalized by means of changes in the electronic density and by a sum-over-states approach, which allows to identify the contributions of the individual excited states. Two separate contributions to NSCD are identified and their physical origins and relative magnitudes are discussed. The results underline NSCD spectroscopy as a plausible tool with a power for the identification of not only different molecules, but their specific structures as well.Recent years have seen both experimental progress 1-5 in, as well as many theoretical proposals 6-16 of, spectroscopies based on nuclear magneto-optic phenomena (termed nuclear magneto-optic spectroscopy, NMOS). NMOS methods, which manifest themselves through modulation of the polarization state of light passing through a nuclear spin-polarized sample, can be thought of as analogues of the classical magneto-optic phenomena 17 , formally obtained by exchanging the externally applied magnetic field by the field originating from a macroscopic magnetization of an ensemble of nuclear spins. In this way, Faraday rotation 18 becomes nuclear spin-induced optical rotation (NSOR) 1 , the classical Cotton-Mouton effect 19,20 becomes nuclear spin-induced Cotton-Mouton effect (NSCM) 6,9,13,14 , or nuclear quadrupole-induced Cotton-Mouton effect (NQCM) 12,15 , and the classical magnetic circular dichroism (MCD) [21][22][23] gives rise to the nuclear spin-induced circular dichroism (NSCD) 16 .The NMOS effects offer possibilities to develop new optical spectroscopy approaches with a high, nucleus-specific and spatial resolution, due to the fact that hyperfine interactions localized at magnetic nuclei are involved, in analogy to nuclear magnetic resonance (NMR) spectroscopy 24 . In addition, owing to the optical detection, NMOS has the potential to offer increased sensitivity and, in extensions to imaging, better resolution as compared to the relatively insensitive NMR method. Moreover, the NMO effects are based on observables differing from those of NMR. Hence, NMOS can also provide new insight into the physical properties of molecules and materials.The first observed NMO phenomenon, NSOR, i.e., the rotation of linearly polarized light by an ensemble of spin-polarized nuclei, was described in 2006 by Savukov, Lee and Romalis 1 . NSOR is caused by the nuclear spin-induced antisymmetric polarizability 6,25,26 of the molecules and is expected to feature an enhanced signal at light wavelengths close to optical transitions. Furthermore, the excited state in question may be localized to a chromophore containing the probed nucleus, thereby realizing a high spatial resolution 7 . NSOR was also predicted 7 to feature an optical chemical shift, expressing different rotation between different molecules, as well as resolving specifically polarized inequivalent nuclear sites within one molecule. The differences in rotation between molecules have already been experimentally verified 5 , while high-resolution nuclear site specificity still awaits experimental confirmation. While NSOR is the only experimentally observed NMOS effect to date, its discovery has inspired a number of theoretical works suggesting existence of further effects 6,9, , among them the NSCD, the focus of this paper. NSCD manifests itself as the difference in the absorption of left-and right-circularly polarized light or, equivalently, as ellipticity induced into the incident, linearly polarized light, in samples with macroscopic nuclear spin polarization in the direction parallel with the light beam. It should be stressed that the NSCD effect is not a perturbation of natural or magnetic circular dichroism, but has a different physical origin. NSCD arises from the real part of the nuclear spin-inflicted antisymmetric polarizability, which gains non-vanishing values only at frequencies in the vicinity of optical transitions 16 . This is in contrast to NSOR, which arises from the corresponding imaginary part, and which can be non-zero at any finite frequency. It may be speculated that NSCD will be the method of choice over NSOR in spectroscopic investigations of molecules by the analogy of NSCD with natural and magnetic circular dichroism, which are much more popular methods than the corresponding birefringent effects of optical rotation and Faraday rotation. Both the magnitude of the NSCD effect, its likely information-richer nature and the ability to single out a particular molecular species in the mixture -as it is, unlike NSOR, intimately dependent also on the specific excited state in question -underlie this anticipation. Theoretical studies of NSCD were so far mainly concerned with small model organic molecules 16 or, more recently, with fullerenes 27 . While it has been shown in the case of fullerenes that NSCD is capable of distinguishing different atom types within a molecule, a detailed study of the underlying effects giving rise to and influencing the shape of NSCD spectra has not yet been performed. This paper thus dives deeper into the effects underlying NSCD spectra in order to further the understanding of this phenomenon. We show that NSCD is an effect with pronounced sensitivity to the electronic structure and that even changing the structure to a different conformer may significantly influence the spectra. This sensitivity can be linked to the changes in the electron density upon excitation. Conversely, even molecules with rather different structures can give rise to similar spectra if the spatial distribution of their excited state densities is comparable. By using a sum-over-states approach we also show that the NSCD signal can be decomposed into two contributions, which we call B d and B a terms based on their band-shapes resembling those characteristic for dispersion and absorption, respectively. We describe how the spectra exhibit the local nature of NSCD via the mutual orientation of the transition moments involving the orbital hyperfine operator ĥPSO and electric dipole operator µ ˆ. This situation is easily observable in the case of B d -term dominated spectra. ## Results The studied structures of all four nucleic acid bases are presented in Fig. 1. They are numbered as in the original study they were taken from 28 and are grouped according to their structural similarity. In particular, the most similar structures, consisting of conformers corresponding to different orientations of hydroxyl or deprotonated amino groups, will be discussed together as subgroups. We will focus on cytosine and thymine in the main text and draw conclusions from there. Additional results for adenine and guanine supporting our conclusions are described in Supplementary Information (Supplementary Figs S1 and S2). Cytosine. We begin with the discussion of the NSCD spectra of the pair cyt2/cyt3 shown in Fig. 2 (top) together with the difference densities corresponding to the changes of the electron cloud upon excitation from the electronic ground state to each excited state in question. As can be seen, the cyt2/cyt3 conformers have remarkably similar NSCD signal as well as very similar difference densities, as could be expected since the hydroxyl group does not participate extensively in the excitations. The spectra are simple and all atoms give rise to a similar couplet feature with the two lobes at the energies of the first two excitations. The signals corresponding to the C 1 , C 2 and C 4 nuclei have the positive sign on the low energy side of the couplet, whereas the opposite situation pertains for the C 3 nucleus. The reason for this behaviour lies in the different orientation of the 〈 | | 〉 ĥ 2 1 PSO transition matrix elements for the atom C 3 , as compared to others, as will be discussed later. In contrast to cyt2/cyt3, the pair cyt4/cyt5 (Fig. 2, bottom) show marked differerences in their NSCD spectra. In the case of cyt4, the C 1 , C 2 and C 3 nuclei give rise to spectra superficially resembling those of cyt2 and cyt3. However, as can be seen by comparing the corresponding excitation energies, the strong higher-energy component of the signal comes from the third excited state rather than the second one, as was the case for cyt2/cyt3. The C 4 nucleus also behaves very differently, having mostly positive sign and lacking the couplet structure. The spectra of the cyt5 structure also significantly differ from those of cyt2 and cyt3 as well as from that of the structurally similar cyt4. The C 1 and C 4 nuclei do not reveal any couplet structure and in the case of C 2 and C 3 the sign pattern of the couplet is reversed as compared to cyt4. The comparison of the difference densities of cyt4 and cyt5 shows similarity between the first two excited states, but in the third excitation the deprotonated amino group significantly influences the distribution of the density. Moreover, no clear correspondence can be found between the difference densities of cyt4/cyt5 and those of cyt2/cyt3 pair. These observations suggest a relation between the shape of the electron density and the NSCD spectra. Thymine. For thymine we can identify three groups of chemically similar structures thy4/thy5/thy7/thy8, thy3/thy9 and thy6/thy12 (Fig. 1). The first group features four different orientation combinations of the two hydroxyl groups (Fig. 3). At first glance it can be seen that, while structurally very similar, there are two pairs of molecules giving rise to two distinct sets of spectra. The first pair consists of thy4 and thy5 and the other pair is thy7 and thy8. The NSCD spectra of thy4 and thy5 (Fig. 3, top) are dominated by a couplet feature with the two lobes positioned at the first and second excited state. The carbon nuclei C 1 , C 2 , C 4 and C 5 produce a positive lobe on the lower energy side, while C 3 corresponds to a negative one. Note that the intensity of the NSCD signal of the C 5 nucleus is several times smaller than for the other nuclei, probably because of its location outside the aromatic chromophore system. On the other hand, the thy7/thy8 pair features an order-of-magnitude weaker spectra, which are, furthermore, mostly composed of isolated bands due to the wider energy separation of the excited states, as compared to the couplet-dominated case of thy4/thy5. Even though the excited-state energies are quite different from those of the thy4/thy5 pair, the overall sign pattern of the spectra follows a similar trend as in the first pair, with C 3 switching sign with respect to all the other nuclei, for the first two excited states. The only exception here is the unexpected behaviour of the C 1 nucleus of thy8. The difference densities for the structures within each of the pairs thy4/thy5 and thy7/thy8 are similar up to the third excited state. In fact, the densities are rather similar among all of the four structures, but their energy order is different. For example, the change in density corresponding to the first excited state in thy4 is very similar to that for the first excited state in thy5, but resembles those of the second excited state for thy7 and thy8. Interestingly, a striking similarity can be observed when comparing the spectra of the thy4/thy5 pair to those of cyt2/cyt3. The overall shape, position and even the intensity of the NSCD signal for carbon nuclei C 1 -C 4 are almost identical. This appears to be somewhat more than just a fortuitous occurrence, since a comparison of the corresponding difference densities reveals a large similarity among these four structures as well. While the distributions are slightly different near the nitrogen and oxygen nuclei, their overall shapes are very similar for the first two states, which correspond to the strong couplet feature. This suggests that even structurally different molecules can give rise to similar NSCD as long as their electronically excited states are quite similar. The second group of thymine structures (thy3/thy9) is shown in Fig. 4 (top). It is apparent that the spectra do not share many common features. The excited state energies are very different and no nucleus shows even a similar spectral pattern. The overall dissimilarity of the spectra is also reflected in the rather different difference densities in the pair. The third group (Fig. 4, bottom) consists of thy6 and thy12. Similarly to the previous pair, the excitation energies and the NSCD signals of these two structures are very different and no strong systematicity among the two sets can be found. It should also be noted that the difference densities of these two structures are quite different, further suggesting a rather prominent role of the nature of the excited states in NSCD, and only a minor influence of the rather similar molecular structure. Finally, it should be also noted that no strong overall similarity can be observed between the NSCD signals of molecules from different groups of thymine structures, correlating with their dissimilar difference densities. 1 H NSCD of nucleobases. The discussion so far concerned only the NSCD of 13 C nuclei. This was done in order to demonstrate the NSCD for atoms that do not change their bonding situation between different structures -all carbon atoms are still connected to the same chemical neighbourhood in all structures of a particular base. There is, however, another set of nuclei suitable for experiments, protons. The analysis in the 1 H case is complicated by the fact that they migrate to various bonding sites in the different tautomers. Therefore, it is not possible to construct a one-to-one comparison of the different structures similar to that of 13 C above, because of the fact that certain kinds of hydrogen nuclei do not exist in other structures (e.g., hydroxyl hydrogen absence in cyt4 Analysis of NSCD. To rationalize the NSCD signals of different nuclei, we performed an analysis in terms of the individual excited states. As the complex polarization propagator (CPP, see the Methods section) calculations used to produce the spectra of the previous figures do not allow to simply distinguish such individual contributions, we used the sum-over-states (SOS) approach, instead. The NSCD signal is computed through quadratic response functions and can be expressed, for nucleus K, as 16 where ω, μ 0 , c 0 , N A and 〈 I K 〉 refer to the angular frequency, permittivity of vacuum, speed of the light in vacuum, Avogadro constant and nuclear spin polarization, respectively. For a particular transition from the ground electronic state |0〉 to the excited state |m〉 , NSCD is proportional to the residues of the quadratic response function 29 ∑ where |k〉 is a member of a set of further excited states with the corresponding energies of the states E 0 , E m and E k , is the angular frequency corresponding to the excitation to |m〉 , ω b is the frequency of the external perturbing field, µ ˆ and ĥPSO are the electric dipole and orbital magnetic hyperfine operators, respectively, and α, β, γ denote Cartesian coordinates. The two contributions in eq. ( 2) are qualitatively different. In the first one, which we call the B d -term (as it appears as a dispersion bandshape, see below), both electric dipole transition moments are between the ground state and an excited state, while the matrix element of ĥPSO is between two excited states. In addition, the denominator of the B d -term contains the difference of the energies of the two excited states. The second term in eq. ( 2), which we denote as the B a -term (since its bandshape resembles an absorption band), involves the ground state in one matrix element of the electric dipole moment, as well as in that of the ĥPSO operator, whereas the transition moment between the two excited states |m〉 and |k〉 involves now the operator µ ˆ. In addition, the denominator is the energy difference between the ground and the excited states in this case. The energy denominators play a significant role in the overall NSCD intensity and the relative significance of the B d and B a contributions. The denominator in B a -type terms corresponds to excitation energies, hence it can never be smaller than the relative energy of the first excited state. In addition, the B a -type terms can be expected to contribute on account of the denominators progressively less to the NSCD spectrum as excited state energies rise. On the other hand, the energy difference of the two excited states in the B d -type terms can in principle be very small, allowing a very large contribution from this type of terms. For this reason, it may be reasonably expected that the B d -terms are dominant in cases of excited states packed closely together. Another interesting feature is related to the way the B d and B a terms of individual transitions from the ground state contribute to the overall spectrum in eq. ( 1). Due to alternating Levi-Civita tensor, the imaginary ĥPSO operator, and the energy difference in the denominator, the B d contributions to the NSCD signal of a pair of excited states |k〉 and |m〉 are identical in magnitude but of opposite signs when the states are interchanged. Where the electric dipole operators appear "symmetrically" in the first term of eq. ( 2), the transition moments of the ĥPSO operator are between two excited states, and will change sign upon exchanging the two states. The same will occur with the energy denominator. Because of this and the properties of Levi-Civita tensor, the B d terms of a pair of excited states, close to one another in energy, contribute to the spectrum (eq. ( 1)) with a couplet-like feature, with the two lobes of the NSCD signal at the energies of the states |k〉 and |m〉 . No such symmetry exists for the B a terms, so no general conclusions on the spectral shape can be easily drawn for them. Note, however, that both described terms are arising from the same perturbation of the electronic states by the local magnetic field and are physically analogous to the B-term in MCD. There is no A term-like contribution in our systems as the molecules do not possess the symmetry required for presence of degenerate states. In order to gain further qualitative insight into the NSCD signal, we calculated NSCD spectra using the SOS expansion of eqs ( 1) and ( 2), with {|k〉 } running over the first 10 excited states. To compare the results of the SOS procedure with those of CPP, we have numerically fitted Lorentzian lineshapes centered at each excitation energy to the CPP spectral points. The intensities obtained correlate well with the SOS results (see Supplementary Figs S9 and S10). The correlation is particularly strong when an intense B d term is present due to closely spaced electronic excited states. On the other hand, weak spectra composed of mainly B a terms are described less accurately by SOS with only 10 excited states included in the summation. In the case of strong B d -term dominated spectra with a large magnitude of the NSCD signal, we can discuss in more detail the information content of the NSCD spectrum for such systems. As a test case we have chosen cyt2, as it provides a simple spectrum dominated by just one couplet, for all the 13 C nuclei. For this case we can neglect the B a -term contribution altogether and approximate the B d terms by including just the two interacting states -the first and the second one in eq. ( 2): Due to the symmetry discussed above, only eq. ( 3) will be discussed in the following as eq. ( 4) yields exactly the same numerical results, just with the opposite sign. The properties of the Levi-Civita symbol allow eq. ( 3) to be rewritten as showing that the NSCD signal contains information about the magnitude of the matrix element 〈 | | 〉 ĥ 2 1 PSO of the orbital hyperfine operator between the two excited states, as well as its direction with respect to the cross-product of electric dipole transition moments between the ground state and the excited states |1〉 and |2〉 . Eq. ( 5) offers a deeper insight into the appearance of the NSCD spectra of the individual nuclei for cyt2. Since the cross-product µ µ × ˆ0 2 1 0 is always nucleus-independent, the only parameters modulating the appearance of the NSCD signal for different nuclei are the direction and magnitude of 〈 | | 〉 ĥ 2 1 ## PSO . Judging by the features of the spectra of cyt2, it can be deduced that the projection of ĥ 2 1 onto the product µ 02 × µ 10 for nucleus C 4 is anti-parallel to the projection of the other three carbon nuclei. Figure 5 shows the calculated matrix elements of ĥPSO for the different carbon nuclei of cyt2, revealing that this is indeed the case. Hence, in the B d -term dominated spectra, NSCD can provide a direct measure of the component of the transition moment of the orbital hyperfine operator ĥPSO perpendicular to the plane defined by the two electric dipole transition moments involved. Some interesting cases are particularly well-suited for analysis of NSCD in terms of the transition moments (matrix elements) of ĥPSO discussed above. For example, the excitation spectra of porphyrins in the Soret or Q-band region are often described in terms of two in-plane polarized orthogonal transitions . For such system, NSCD would directly probe the component of matrix element of ĥPSO between these states perpendicular to the porphyrin ring. ## Discussion In this paper we have performed a computational study, using a recent implementation of the complex polarization propagator-quadratic response theory method, of the theoretically predicted, but not yet experimentally observed nuclear spin-induced circular dichroism effect. Via NSCD, a sensitive optical spectroscopy with atom-specific resolution might be realized, through the modulation of the dynamic dipolar polarizability of the molecule by magnetic hyperfine interaction localized at the nuclei. We have demonstrated on a set of 20 structures of nucleic acid bases that different molecular structures of related systems provide distinct NSCD features. In contrast to NMR spectroscopy, which provides nucleus-specific resolution with the spectral parameters determined by the electronic ground state of the systems, the shape of the NSCD spectrum is heavily correlated to the differences in the electron density between the ground and excited states. Molecules that undergo similar changes in density upon optical excitation give rise to qualitatively similar NSCD signals, even when their molecular structure is rather different, such as in the case of certain structures of thymine and cytosine. Consequently, there is no simple relation between the local chemical neighbourhood of the nucleus and its spectral response, in contrast to NMR. We have also performed a sum-over-states perturbational analysis of NSCD arising from the lowest excited states, in order to gain insight into the physical origins of the effect. It turns out that the NSCD signal can be understood as being composed from two qualitatively different contributions, which we name the B d -and B a -terms. The B d -term gives rise to couplet-like features in the spectra, which are particularly distinctive in molecules involving a pair of narrowly spaced excited states. The B d -term is related to the transition moment of the magnetic hyperfine operator between the two excited states, and it provides a measure of the component of this moment perpendicular to the plane defined by cross product between the two electric dipole transition moment vectors between the ground state and the two excited states. Moreover, for systems dominated by the B d -term, the sum-over-states procedure (with a limited range of states) provides a very good agreement with the rigorous response theory-based calculations. For the time being, an analogous physical interpretation relating the B a -term to a small number of readily interpreted contributions, remains elusive to us. Due to its sensitivity to the structure of the excited states, NSCD shows potential for the identification of molecules based on their excitation properties. Moreover, direct insight into the relation between spatially extended electric dipole transition moments and the transition moment of the localized orbital magnetic hyperfine operator can be obtained in systems with narrowly spaced excited states, when NSCD is dominated by the B d -terms. Finally, we would like to briefly comment on the experimental feasibility of the NSCD measurement. Our results predict the magnitude of the NSCD effect to be on the order of tens to hundreds of microradians for fully polarized nuclei. Therefore, the NSCD effect is actually orders of magnitude stronger than the already experimentally measured NSOR. It is reasonable to expect it to be also measurable using a similar instrumentation. For example, the inherent shortness of a single measurement limited by nuclear magnetization relaxation time can be compensated for by repeating the data acquisition or by using spin-locking techniques. This approach is clearly viable as evidenced in previous experiments with NSOR. Additionally, NSCD is modulated at the Larmor frequency of the nuclear precession, while other potentially interfering effects like natural circular dichroism or classical magneto-optic effects are static in nature and can thus be easily separated and filtered out. We believe that a continued development of this branch of nuclear magneto-optic spectroscopy may provide novel opportunities for the investigation of molecular structure and properties. ## Methods A development version of the DALTON program package 33,34 and the Turbomole 6.5 program 35 have been used in the calculations. All molecular structures of the four bases, optimized at the B3LYP/def2-TZVPPD level, were taken from the study of Ovchinnikov and Sundholm 28 . The excitation and NSCD spectra in the present work were calculated in DALTON employing the BHandHLYP functional 36,37 and a custom-made basis set derived from the def2-SVP 38 , by adding two diffuse Gaussian primitives to the angular momentum values corresponding to the occupied atomic orbitals in the ground state of the atoms involved (see Supplementary Information). The basis set was created based on previous experience 16,27 , showing that diffuse functions are necessary for this property to gain a faithful description of the interaction with light. The extended def2-SVP basis set was found to present a good compromise between the computational feasibility and accuracy for systems of the present size, when benchmarked against the correlation-consistent d-aug-cc-pCVDZ basis set used for NSCD in earlier work 16 . The BHandHLYP functional was presently used for all the calculations as its performance for the excitation energies and oscillator strengths 16 , as well as NSOR 7 has been found satisfactory in comparison to coupled cluster results. The NSCD spectra were calculated with DALTON using the complex polarization propagator (CPP) method, also termed the damped response function formalism 16,39,40 . CPP allows convergent response calculations even in the region of the spectrum where electronic transitions take place. The implementation makes use of the efficient damped response theory solver developed by Kauczor et al. 41 , involving an empirical linewidth parameter Γ . This parameter can be adjusted to reproduce the experimental bandwidth resulting from various factors such as solvent and vibrational effects. As there are no experimental NSCD data available yet, the damping parameter Γ = 1000 cm −1 was chosen to represent a reasonable situation for UV-VIS absorption bands in the aqueous phase. The NSCD signal was calculated in the wavelength range of 200-280 nm with the step of 2 nm, which covers the lowest optical excitation energies of the target systems and should be experimentally available to NSCD measurements in the future. For reasons of practicality of the forthcoming experiments, we were mainly interested in the NSCD response corresponding to the 13 C (and 1 H) nuclei as it is more difficult to maintain the macroscopic nuclear spin polarization for quadrupolar 17 O and 14 N nuclei due to their rapid relaxation, or the spin 1 2 isotope 15 N due to its low abundance. The residues of the linear and quadratic response functions for the involved electric dipole and orbital magnetic hyperfine operators required for the sum-over-states (SOS) approach were obtained from calculation performed by the adapted DALTON code. The excited-state differential densities were obtained using Turbomole 6.5 35 . Molecular graphics were created using the UCSF Chimera package 42 .

chemsum

{"title": "Relation between molecular electronic structure and nuclear spin-induced circular dichroism", "journal": "Scientific Reports - Nature"}

site-selective_aqueous_c–h_acylation_of_tyrosine-containing_oligopeptides_with_aldehydes

## Abstract: The development of useful synthetic tools to label amino acids within a peptide framework for the ultimate modification of proteins in a late-stage fashion is a challenging task of utmost importance within chemical biology. Herein, we report the first Pd-catalyzed C-H acylation of a collection of Tyr-containing peptides with aldehydes. This water-compatible tagging technique is distinguished by its site-specificity, scalability and full tolerance of sensitive functional groups. Remarkably, it provides straightforward access to a high number of oligopeptides with altered side-chain topology including mimetics of endomorphin-2 and neuromedin N, thus illustrating its promising perspectives toward the diversification of structurally complex peptides and chemical ligation. ## Introduction Non-natural amino acids and peptides derived thereof are highly coveted compounds in proteomics and drug discovery due to their often enhanced biological activities and improved metabolic stability compared with their native analogues. 1 As a result, there is an urgent demand to increase the available synthetic toolbox to perform chemical labelling processes of peptides and proteins in a late-stage fashion. Innovation is occurring at a rapid pace and a myriad of reliable methods have emerged within the last decade in the burgeoning area of bioconjugation. 2 Metal catalysis has recently unlocked new tactics in the feld, 3 thereby enabling the development of a sheer number of metal-catalyzed modifcation techniques featuring the manipulation of otherwise unreactive C-H bonds embedded within the amino acid backbone 4 and the corresponding side-chains. 5 The latter have streamlined the straightforward assembly of biomolecules of paramount signifcance in a more sustainable manner by avoiding the use of pre-functionalized substrates. Despite the wealth of reports in the feld, several challenges still remain: (a) most of the protocols entailed the diversifcation of a limited number of amino acid residues including tryptophan (Trp), glycine (Gly), alanine (Ala), cysteine (Cys) or phenylalanine (Phe), among others, 6 and (b) toxic halide-counterparts and organic solvents are usually required. Accordingly, the selective tagging of other canonical amino acid residues while implementing atomeconomical C-H coupling partners represents an unmet challenge of capital signifcance within peptide chemistry and protein engineering. Tyrosine (Tyr), the 4-hydroxylated congener of Phe, constitutes an abundant proteinogenic amino acid, which is a privileged core in a vast array of relevant compounds such as neurotransmitters and hormones, as well as a versatile precursor to numerous alkaloids with potent antibiotic activity such as vancomycin, among others (Fig. 1). 7 Likewise, Tyr-containing compounds are of widespread use as dietary supplements or food additives, and play a pivotal role in biological processes such as photosynthesis. Its inherent chemical reactivity is dictated by the pH-tunable and electronrich phenol-containing side-chain. In this regard, several transformations performed in simple phenols have been translated into elegant tagging techniques of Tyr-containing peptides. Whereas O-functionalization reactions including arylation, 8 alkylation 9 and glycosylation 10 reactions generally occur under basic conditions, C-targeted reactions at the ortho-C-H bond to the phenol motif usually proceed in acidic or neutral conditions. The introduction of highly reactive electrophiles upon Mannich-type reactions, 11 diazonium couplings 12 or enetype reactions 13 as well as N-centered radicals derived from phenothiazines 14 has been achieved under metal-free reaction conditions. Conversely, metal catalysis has been crucial to install other coupling partners such as aryl halides, 15 aryl trilfuoroborate salts, 16 acrylates 17 and the trifluoromethyl group 18 (Scheme 1, route a). The latter have been rarely applied within a challenging peptide framework and hence the site-selective modifcation of tyrosine unit still remains elusive. As part of our interest in the radical modifcation of peptides, 19 we have recently reported unprecedented Pdcatalyzed C-H acylation reactions for the efficient labelling of Phe-containing peptides. 20 While conceptually innovative, the protocol suffered from certain downsides such as the requirement of stoichiometric amounts of silver salts to ensure high yields, 21 the use of DMF as toxic organic solvent and it was found just applicable to Phe derivatives housing a picolinamide as directing group (DG) at the N-terminal position. Given that the use of aldehydes has been overlooked in the realm of bioconjugation, we sought to unveil their full synthetic potential in the virtually unexplored ortho-C-H acylation of Tyr residues within complex peptide settings. Building on precedents in C-H acylation reactions, 22 we predicted that the use of a DG would be determinant for achieving site-selectivity 23 through the formation of a 6-membered palladacycle prone to undergo further oxidative addition of the transient acyl radical species (Scheme 1, route b). In particular, we envisioned that the conversion of the native Tyr unit into the corresponding 2-pyridyl ether compound 24 would be crucial to enable the site-selective modifcation of Tyr residues at any position within the peptide sequence. Herein we report on the frst Pd-catalyzed C-H acylation of Tyr-containing peptides with aldehydes. This scalable method avoids the undesired use of toxic and expensive silver salts, and features the use of water as a non-flammable and environmentally-friendly solvent, thus resulting in a powerful diversifcation technique of paramount importance in peptide chemistry. ## Results and discussion We commenced our studies by exploring the radical acylation of Boc-Tyr(OPy)-Leu-OMe (1a) with p-anisaldehyde (2a) as the model reaction. When submitting dipeptide 1a to the previously reported reaction conditions for the acylation of Phe-containing peptides 20 just traces of 3aa were obtained, hence showing the subtleties of the modifcation of the Tyr scaffold. Initial exploratory solvent screening with tert-butyl hydroperoxide (TBHP) as oxidant showed the feasibility of our approach and moderate to good yields were obtained in solvents such as toluene, acetonitrile, chlorobenzene and even water (Table S1 †). 25 Driven by its clear benefts in the modifcation of biomolecules, we focused on the development of a practical acylation under an aqueous environment. 26 After considerable experimentation, 25 we eventually found that the combination of Pd(OAc) 2 (10 mol%), an aqueous solution of inexpensive TBHP (Luperox®) as oxidant in neat water as solvent at 90 C provided 3aa in 78% yield as a mixture of mono-and diacylated products (8 : 2 ratio) (entry 1). Blank experiments underpinned the crucial role of both Pd catalyst and oxidant in the acylation reaction as not even traces of 3aa were detected in their absence (entries 2 and 3, respectively). The performance of the reaction under air resulted in lower yields of 3aa (entry 4) and the process was entirely inhibited under an oxygen atmosphere (entry 5). The use of variable amounts of oxidant and p-anisaldehyde led to an optimal balance between yield and mono-selectivity when using 4.0 and 3.0 equivalents, respectively. For example, when using 2.0 equivalents of TBHP, 3aa was obtained in a comparatively lower yield (entry 6) and when increasing the amount of aldehyde to 6.0 equivalents the higher yield was due to a loss in regioselectivity (entry 7). Other reaction parameters such as palladium source, supporting ligands and reaction temperature were analyzed but lower yields were obtained; indeed, higher selectivity toward the monoacylation product was only achieved at the expense of obtaining lower overall yields. 25 As depicted on Table 1, an aqueous solution of TBHP provided much better results than other related peroxides or persulfates (entries 8-12), which together with the use of water as solvent constitutes an additional bonus of the method in terms of economics and sustainability. Importantly, unlike other Pd-catalyzed C-H peptide modifcation reactions, silver additives were found to be unnecessary. Moreover, despite the highly oxidizing reaction system, undesired N-acylation of the peptide backbone with p-anisaldehyde was never observed. 27 With the optimized conditions in hand, we next investigated the scope of the C(sp 2 )-H acylation protocol with Tyr-containing dipeptide 1a. Notably, a wide variety of electronically diverse aldehydes smoothly underwent the target oxidative coupling, thus enabling the rapid access to a variety of unknown acylated Tyr-containing dipeptides in a late-stage manner. In general, a variety of benzaldehydes regardless of their electronic nature provided the corresponding acylated products 3 in good yields as mixtures of mono-and diacylated compounds, with a preferential selectivity toward the monoacylated product (up to 8 : 2 ratio). The method was compatible with the presence of ethers (3aa-ad), halides (3ae-ah), acetamide (3ai) and naphthyl system (3ak). Conversely, the highly electron-withdrawing nitro group resulted in no conversion of dipeptide 1a. Interestingly, the lower tendency to oxidation of aliphatic aldehydes such as heptanal and cyclohexanecarboxaldehyde ushered in the exclusive formation of mono-acylated products 3am and 3an, respectively, in moderate yields under the standard reaction conditions featuring the use of water as the sole solvent. However, the performance of the process in chlorobenzene with a higher excess of the corresponding aldehydes resulted in slightly higher yields (up to 57%). Likewise, other challenging aldehydes housing pharmaceutically relevant heterocyclic scaffolds could be also employed as reaction partners, although the use of toluene as solvent at 100 C was found determinant for the process to occur. 25 Accordingly, 2-thiophene (2o), 2-furan (2p), N-methyl-2-pyrrole (2m) and N-methyl-3-indolyl carboxaldehydes (2n) selectively afforded the corresponding monoacetylated products 3ao-ar. It is noteworthy that a gram-scale acylation with p-(trifluoromethyl)benzaldehyde (2h) was successfully performed, which verifed the robustness and synthetic utility of our peptide tagging manifold. Collectively, the aqueous acylation of Tyr derivatives reported herein underscores related protocols on simple 2-phenoxypyridine derivatives, which occurred at higher reaction temperatures (up to 140 C), in chlorinated solvents and are restricted to the use of benzaldehydes (Table 2). 24b,g We next explored the synthetic scope in the challenging setting of short-to-medium size peptides (Table 3). Notably, peptides bearing Phe (1b), Val (1c), Gly (1d), Asn (1e), Ser (1f), Asp (1g), non-protected Tyr (1h), Pro (1i), Ala (1j), and Lys (1k) boded well with the reaction conditions and provided the corresponding acylated dipeptides (3b-k) in good yields. The success of the method did not rely on a specifc situation of the Tyr along the peptide sequence, and was applicable to Tyr residues located both at the N-and C-terminal positions. We next evaluated the viability of the Pd-catalyzed acylation for the late-stage diversifcation of more complex oligopeptides. In this respect, the process efficiently occurred in a selective manner in Tyr-containing tri-and tetrapeptides (1l-o), including even peptides housing the Tyr unit in inner positions 1, entry 1. b Yield of isolated product after column chromatography, average of at least two independent runs. c Ratio of mono-and diacylated product (3 : 3 0 ). d Using toluene as solvent. (1m,n). Importantly, pentapeptide 1p and hexapeptide 1q bearing the amino acid sequence of biologically relevant Endomorphin-2 and Neuromedin N, respectively, were also acylated in a late-stage fashion. The latter examples illustrate the high synthetic potential of this acylation technique toward the site-selective labelling of biomolecules of high structural complexity. In all cases, the incorporation of the radical acyl species was biased by the 2-pyridyloxy group attached to the phenol ring within the Tyr residue and other sensitive functional groups such as carboxamides within Gln (3nh) and Asn (3eh) as well as oxidizable protic free-hydroxyl groups of Ser (3) and Tyr (3ha) remained intact under the reaction conditions. The free-amino group of Lys residue (3ka, 3qh) as well as the N-terminal residue of the peptide sequence was conveniently protected to avoid undesired oxidative aminations with the corresponding aldehyde. 28 Peptides bearing other electronrich aromatic residues such as His and Trp or guanidinecontaining Arg unit were not tolerated. 25 Likewise, thiolcontaining Cys and Met residues were not accommodated, which could be due to competitive oxidative reactions with the corresponding aldehydes (Table S8, ESI †). 29 With the aim to create molecular diversity, we extended the scope beyond the introduction of one single aldehyde into the Tyr unit, and found that the performance of two consecutive aroylation reactions with two distinct benzaldehydes enabled the assembly of fully decorated Tyr-containing dipeptide 3as in a selective manner (Scheme 2). Likewise, the aldehyde unit could be tethered within a Tyrcontaining peptide and efficiently coupled with another Tyrcontaining short-to-medium peptide to deliver unprecedented oligopeptides featuring a unique diaryl ketone cross-linking (Table 4). Notably, the functionalization always occurred in a selective fashion toward the mono-acylated compounds. The practicality of the method within the realm of chemical ligation was verifed by the selective mono-acylation of Neuromedin N analogue 1q, thereby enabling the assembly of octapeptide 3qt of high structural complexity. It must be noted that all the experiments were run at least twice with a variable yield by no more than 5% between runs, thus showing the reliability of the protocol. Although the facile removal of the OPy group has been customarily described in simple aryl systems upon a two-step sequence entailing N-methylation with methyl triflate followed by the cleavage of the resulting C(pyridinium)-O bond by treatment with an alcoholic solution of sodium, 24 its application in a peptide sequence resulted in mixtures of products and racemization of the existing chiral centers. Accordingly, we studied the use of modern metal-catalyzed borylation reactions for the ultimate, yet milder removal of the directing group. 30 After careful evaluation of a number of Rh-, Ni-and Fe-catalyzed borylation reactions, the targeted borylative cleavage was never achieved within our Tyr-containing peptides and the removal of the OPy group occurred in moderate yield to produce the corresponding reduced product 4. 30b Although it poses a limitation at frst sight and remains an issue to be improved, further optimization of the process could provide a complementary avenue for the assembly of meta-aroylated Phe-containing peptides, thereby resulting in the direct conversion of a Tyr residue into the Phe analogue featuring the use of the OPy as a traceless directing group (Scheme 3). Likewise, the fully decorated Tyr(OPy)-containing peptides assembled herein could offer interesting possibilities within drug discovery. ## Conclusions In summary, we have developed a broadly applicable method for the site-selective tagging of Tyr-containing oligopeptides featuring a novel Pd-catalyzed C(sp 2 )-H acylation reaction with abundant aldehydes. This labelling platform represents a reliable, yet innovative, means for the radical diversifcation of a wide variety of Tyr-containing compounds, thus providing access to novel peptidomimetics of high structural complexity. Although one might anticipate that site-selectivity issues might come into play in the presence of multiple C-H bonds, our work meets this challenge, offering an unrecognized opportunity within the radical labelling of peptides. Salient features of our protocol are the compatibility with an aqueous environment, the widespread availability and low-cost of the aldehydes, the broad functional group tolerance, the preferential siteselectivity toward the functionalization of the Tyr unit and the facile installation and removal of the required 2-pyridyl ether group. Accordingly, we anticipate that this Pd-catalyzed oxidative acylation manifold could become a useful synthetic tool for the late-stage and rapid modifcation of a virtually unlimited set of Tyr-containing lead drug candidates.

chemsum

{"title": "Site-selective aqueous C\u2013H acylation of tyrosine-containing oligopeptides with aldehydes", "journal": "Royal Society of Chemistry (RSC)"}

the_hidden_transition_paths_during_the_unfolding_of_individual_peptides_with_a_confined_nanopore

## Abstract: A fundamental question in peptide folding/unfolding is how the peptide fleets through a set of transition states which dominate the dynamics of biomolecular folding path. Owing to their rapid duration and sub-nm structure difference, however, they have always been oversimplified because of limited instrumental resolution. [1][2][3] Moreover, the most experiments indicate a single fold pathway while the simulations suggest peptides owns the preference in multiple pathways. Using the electrochemical confined effect of a solid-state nanopore, we measured the multiple transit paths of peptide inside nanopores. Combining with Markov chain modelling, this new singlemolecule technique is applied to clarify the 5 transition paths of the β-hairpin peptide which shows 4 nonequilibrium fluctuating stages. These results enable experimental access to previously obscured peptide dynamics which are essential to understand the misfolding in peptides. The statistical analysis of each peptide from high throughput shows that 78.5% of the peptide adopts the Pathways I during their folding/unfolding in a nanopore while 21.5% of the peptide undergoes the hidden folding/unfolding of transit Pathways II-IV. The frequency of the ionic fluctuation reveals a harmonic structure difference of the metastable peptide. Our results suggest the folding/unfolding of βhairpin undergo four major structure vibrations which agree well with the theoretical expectation. These measurements provide a first look at the critical experiment picture of the mechanical folding/unfolding of a peptide, opening exciting avenues for the high throughput investigation of transition paths. The polypeptide chain adopts specific conformations to realize unique biological functions. The central problem of protein science is how the dynamic folding/unfolding process resolves the conformation heterogeneity of peptides. 1,2 Proper identification of the transition conformation is essential to accurately describe the dynamic and hidden folding/unfolding pathway, which is highly relevant to understanding the structure-function relationship of the protein. Although molecular dynamics (MD) simulations provide richly detailed information regarding to the intermediate states along with the multiple folding/unfolding pathway 3,4 , there are few experiment approaches that can directly recognize three or more transition states for elucidating a clear folding/unfolding pathway. The ensemble techniques have led to interpret of the peptide conformation into two simple folding and unfolding state. 5,6 However, the transition path is very short-lived, and are moreover inherently a single molecule phenomenon. Single-molecule techniques including fluorescence, IR spectroscopy and force spectroscopy, have proven effective in detecting populated intermediates. Recently, the optical tweezer has been used to observe the microscopic diffusion motion of single protein transition paths 10 . Despite the advances, all these experimental results only determine the average transition pathway at single molecule level instead of high throughput single-molecule characterization, which oversimplifies the heterogeneous folding/unfolding pathways while the simulations suggest peptides owns the preference in multiple pathways 13 . Until now, it is still difficult to measure every individual folding pathway for each single molecule at a high throughput. Therefore, the hidden folding/unfolding pathway along with the kinetic information remains challenge. To achieve the high throughput identifying the nature of folding/unfolding pathways for each single molecule, several technical hurdles must be overcome. First, a high temporal resolution on microsecond-to-millisecond is required for tracing the subtle conformational changes of each peptide. Second, a rapid and facile readouts should be considered for analyzing hundreds of individual peptides at a high throughput. Third, the peptide should be remained their nature dynamic features without any interferences from the labeling. To overcome these hurdles, we electrochemically confine a single peptide once at a time into a SiNx nanopore. The single peptide undergoes spontaneous folding/unfolding process inside a confined space, which disturbs the ionic distribution and mobility inside the nanopore. By directly monitoring the time-series ionic current of individual molecules, the numerous of characteristic folding/unfolding trajectory could be rapidly readout, which clearly reveals the hidden transition pathways. Here, we confined β-hairpin peptide consisting of 16 amino-acid from B1 domain of protein G, as a model peptide, with a ~4 nm SiNx nanopore of temporal resolution of 100 μs (Figure 1) and thereby uncovered every folding/unfolding pathway for each peptide. Combining with Markov chain model, this new single-molecule technique reveals that each peptide owns its unique ionic current pattern, but exhibits the similarity in the number of nonequilibrium fluctuating stages. The dynamic simulation results suggested that the current stage 1-4 corresponds to four major conformations of the peptide which are compact, stable β-hairpin, partially unfolding and unfolding, respectively. The statistical analysis from each peptide show that 78.5% of the peptide adopts Pathways I during their folding/unfolding in a nanopore, however the whole peptides could be classified into 4 distinctive pathways. These hidden folding pathways with low probability of occurrence suggests the order of barrier height for each intermediate. ## Figure 1| Reading the folding/unfolding pathway of a β-hairpin peptide using a SiNx nanopore. (a) a single β-hairpin peptide binding with monovalent streptavidin is electrochemically confining into a SiNx nanopore with diameter of ~ 4 nm at applied voltage of +150 mV. The dynamic conformation of a peptide perturbated the ionic current flowing through the nanopore, resulting a multi-stage current trace. The peptide sequence is GEWTYDDATKTFTVTE with biotin tag at C-terminal where allowed to form complex with monovalent streptavidin. A hydrophobic cluster is made up of W43, V54, Y45 and F52. 14 The SiNx nanopore is fabricated in thin (10 nm) freestanding SiNx membrane. The detailed structure of β-hairpin is shown in Figure S1. The monovalent streptavidin is a gift from Prof. Mark Howarth at Oxford University. 15 The two compartments of the nanopore chamber are termed as cis and trans. The I-V curve (b) and TEM (c) characterization of a ~4 nm SiNx nanopore. (d) The continuous ionic-current trace of Event 1 without (i) and with (ii) the confining single peptide. The current fluctuations originate from a single peptide captured inside a SiNx nanopore. The current histogram illustrates the current distribution which could be divided into four stages as stage 1 (dark green), stage 2 (light blue), stage 3 (light green) and stage 4 (yellow), respectively. (e) A magnified view of the ionic-current blockades shown in (d). As shown in Figure S2, the raw experimental trace (black color) was processed with a home-designed analysis software to extract the current amplitudes and durations (red color). All the data is required at + 150 mV in 1 M KCl, 10 mM Tris-HCL, pH 8.0. The current was filtered by a low-pass filter of 5 kHz and recorded at sampling rate of 100 kHz. (f) The Markov chain model generated from the current trace of Event 1 revealing the 16 folding/unfolding pathways for the 4conformational state of β-hairpin peptide. The number in the arrow is the probability for each path transition. Each state (n= 1, 2, 3, and 4) shows the highest probability to transit to state 2, which is defined as Pathways I. ## Results and Discussions Confining a single β-hairpin peptide in to a SiNx nanopore. The β-hairpin is the minimal β-structure element, which is also the element component of antiparallel βhairpins. The sequence of β-hairpin that we used is from the C-terminal fragment (41-56) of protein G B1 in the intact protein (PDB code: 1GB1). 16 Previous NMR and molecular dynamic study show that this peptide consists of a hydrophobic cluster to stabilize the β-strand region and four residues to form main-chain H-bonds in the turn region. 16,17 Similar to the previous studies , translocation of a short β-hairpin through the SiNx nanopore leads to an ultra-fast duration which is beyond the bandwidth of the current amplifier (Figure S3). Moreover, the translocation of untethered protein is complicated by the unknown orientation of the protein and the interaction between the pore and nanopore interface. To prolong the observation time and reduce the orientation complicity, here, the β-hairpin peptide is immobilized within the SiNx nanopore by using biotinmonovalent streptavidin complex. The β-hairpin peptide with biotin tags at the C terminus was allowed to form complexes with monovalent streptavidin (Figure 1a and Supplementary Information). Note that the control experiment for the translocation of monovalent streptavidin through a SiNx nanopore did not generate any blockages in ionic current (Figure S4). To provide a confined space for this β-hairpin peptide with 2 nm × 3 nm cross-section, the controlled dielectric breakdown 22,23 is used to fabricate SiNx nanopores with diameter of ~4 nm (Figure 1b-c). In this state, the peptide was captured and immobilized within the SiNx nanopore at an applied potential of +150 mV. We quote the blockage current (ΔI) as a percentage of the open pore current (I0). A previous study demonstrated that the folded state of the maltose binding protein does not change the shape at the 50-250 mV as it translocation through the SiNx nanopore. 24 To eliminate electric field effect on stretching the conformation of the peptide, we chose a low potential of +150 mV in our experiments. At + 150 mV in 1 M KCl, 10 mM Tris-HCL, pH 8.0 (the conditions for all the experiments in this article), the peptide blocked ~4 nm SiNx nanopores generating a time-series current fluctuation at ΔI/I0 of 0.1 to 0.9. Note that the baseline for the open SiNx nanopore shows a low noise of RMS = 19 pA (peak-to-peak noise of 45 pA), which ensures the high signal-to-noise ratio in the experiments. We evaluated the blockage currents and durations for all the spike within the entire current signature via a home designed data analysis program based on our previous data recovery methods 25,26 (Figure 1d-e, Supplementary Information). The current histogram shows the well discriminated current stages for all of the spikes from one capturing event of β-hairpin peptide. For example, the captured Event 1 in Figure 1d exhibits four current stages at + 150 mV which are assigned as stage 1 to 4, respectively. We combined all the spikes in each event show that the current distribution still falls into four stages with peak currents of I1/I0 = 0.18, I2/I0 = 0.40, I3/I0 = 0.53, and I4/I0 = 0.70, respectively (Figure 1d and 2a-b). Base on the current distributions, we divided all the spikes into four current stages which are stage 1 of I1/I0 = 0.1 -0.24, stage 2 of I2/I0 = 0.24 -0.46, stage 3 of I3/I0 = 0.46 -0.58 and I4/I0 = 0.58 -0.8, respectively. Therefore, every single peptide gave the distinct current traces inside the nanopore whose spikes were all assigned to these four current stages (Figure S5). These results suggested that each single peptide exhibits the similar behaviors inside the nanopore, however illustrates its fingerprinted current pattern. Previous studies demonstrated that the conformational change of the peptide produces the current signatures with nanopore techniques. 27 The characteristic current fluctuations during each event suggest the dynamic conformational changes of a single peptide within a confined nanopore. The recent all-atom MD simulations demonstrated that the folded and unfolded protein produce the distinct current values as a single protein confined inside a nanopore. 27 Although the folded or unfolded protein occupied the same volume inside the nanopore, the differences of the ion mobility in proximity to a protein surface causes the distinguishable current stage for the heterogeneous conformation of the protein. 27 Similarly, in our experimental current traces from each peptides, we suggest that the four clear current stages corresponds to the four conformations of the β-hairpin peptide as it is confined inside a nanopore. To validate our experimental current stages for the peptide conformations, we proceeded to carried out the pathway model with MD simulations. Figure 2c illustrates the RMSD values of the peptide backbone atoms from initial structure during the MD simulations. The RMSD indicates the similarity between any MD snapshots and the initial state. We analyze the total 60 ns trajectory and construct the free energy landscape for it. There is one minimum in the landscape, which corresponds to the most stable β-hairpin conformation. As shown in the scatter plots of event 1 (Figure 2a), 73% of the fluctuation locates in stage 2, which occupies 74% of the whole event duration (Figure 2b). Since the peptide adopts β-hairpin as its stable structure as shown in previous study 17 and proved in Figures 2c-d, the stage 2 was suggested to be the most probably β-hairpin conformation and the related small nonequilibrium conformation of the β-hairpin. Stages 1, 3 and 4 may be related to the short-lived metastable conformations which occupies 12% and 6% of the whole event durations, respectively. The simulation results show that the folded β-hairpin presents in 60% simulation time during 60 ns of MD simulations, which further confirms that the folded β-hairpin contributes to the current fluctuation in stage 2. A previous simulation study demonstrated that the I/I0 value increase proportionally with the radius of gyration from the folded state to the extended unfolded state. 27 Simulation results in Figure 3d show that the peptide adopts compact, βhairpin and partially unfolded conformations which own the radius of gyration centered at 7.2, 7.6, 7.9 and 8.4 , respectively. Therefore, in our experiments, the deeper current stage 3 and 4 in event could be attributed to the unformed hydrophobic cluster and/or the twisted of loop formation while the less blockage stage 1 could be related to the compact conformation. Moreover, the interactions between the metastable conformation of the peptide and the nanopore surface may also contribute to current fluctuations. As reported in previous studies, the charge and dipole moment induce current fluctuations 28 . Therefore, the non-stabilizing, intra-strand salt bridge occurring at charged G41...E42 and D47…K50 might give the rise to the current due to the dynamic changes of the charge and dipole moment. Consequentially, the four possible current stages 1, 2, 3 and 4 correspond to the four-different state of the mSA-peptide system, which induces by the four peptide conformational states as compact, stable β-hairpin, partially unfolding and unfolding, respectively. By successively reading every current spike from one blockage event in time-series scale, we suggested that each folding/unfolding pathway of a single peptides could be traced in real time. The transition between the four states of the peptide can be describe by a Markov chain models as shown in Figure 1f, which could build a discrete-state stochastic model capable of describing long-time statistical dynamics. For any short period of time, δt, there is a set of probabilities that this state will transition to another permitted state or will remain unchanged. As a result, for the transit pathways for Event 1 (Figure 1f), all the transitions essentially reached to State 2. For example, there are four possible exit paths for State 1, either 1→2, 1→3, 1→4 or 1→1, whereas 1→2 owns the probability of as high as 0.83. As a contrast, the transition from State 1 to other states (1, 3, or 4) exhibits extremely low possibility of below 0.05. Similarly, the transitions of 2→2, 3→2 and 4→2 reveals the high probability of 0.81, 0.93 and 0.84, respectively. These results suggest that the compact and unfolded peptide may form before the transition to the stable State 2, then the β-hairpin peptide with the lowest energy landscape remains unchanged or undergoes slight dynamic changes within a confined nanopore. In this respect, State 1, 3, or 4 does not appear to be necessarily on each folding pathway, since in the cases where the β-hairpin is independent and directly formed through either State 1, 3 or 4. As shown in Figure 2b, the life time in each state is well fitted by an exponential function and thus could be described by the first order rate law kinetics 29 : where xn is the number of transitions from State n (of normalized current In/I0) in time t, xn(T) is the total number of measured transitions from state n, and k is the rate constant (s -1 ) depicting transition kinetics. The time constants of each state, τ, are given by the inverse of the rate constant k. The results of the calculations are shown in Supplementary Information. As for Event 1, the rate constants of 1→2, 2→2, 3→2 and 4→2 are closed to 10 5 s -1 which is 1-2 orders larger than that of other transitions (Table S1). These results suggest that the metastable conformation of peptide favored to turn to the stable conformation of β-hairpin. The ability of the presented method to identify the folding/unfolding pathway and the related kinetics from the visible current traces of a single capture event as demonstrate here provides a significant advantage to high throughput screening every single peptide for the hidden pathways. ## High throughput elucidating the multiple folding/unfolding pathways Since every capture of a peptide within a confined nanopore produces the unique current traces for the folding/unfolding pathways, this method permits the high throughput analysis of every individual peptides. Here, we have carefully examined more than 300 hundred peptides. Each current trace displays the characteristic folding/unfolding pathways for each peptide. The selected events and their related pathway are shown in Figure 3. We defined the pathways with the transition probability of State n (n = 1-4) → 2 is higher than 50% as Pathways I for the peptide. Therefore, the Event number of 1-33 could be assigned to Pathways I (Figure 4). The statistical analysis reveals that approximate 82.5% of peptides adopt Pathways I as its folding/unfolding pathway, which suggests the peptide highly prones to stay in its most stable conformation as β-hairpin. Interestingly, 5% peptides choose Pathways II as its folding/unfolding features whose transition probabilities of State 1→2 and 2→2 are lower than 50% while 3→2 and 4→2 are higher than 50% (Figure 3ac and 4). For example, the molecule of Event 2 starting from State 1 or 2 takes high probability to experience State 3 before reach to the stable conformation at State 2 (Figure 3a-c). The rate constant of State n (n = 1-2) → 3 in Event 2 is approximately 3-4 times higher than that of State n (n = 1-2)→3 in Event 1 (Table S1- S3). Therefore, this result suggest that this single peptide exhibits the competitive unfolded conformation as its dominant conformation when it stay in the confined nanopore. Moreover, the results indicate that the free energy required to fold the hydrophobic cluster is higher than other interactions for the folding of β-hairpin. This finding is consistent with previous model which predicts that forming the tetrad of hydrophobic residues will speed up the β-hairpin formation. 17 The Pathways IV also further prove this folding mechanism. As shown in Pathways IV (Figure 3f-h), the transition probability of State n (n = 1, 2 or 4) →3 is higher than 50%, however that of State 3→2 is higher than 50%. The Pathways IV occupies 7.5% of the total peptides. This result suggests that the unfolded conformation of state 3 in Pathways IV shows the high possibility to further fold into the stable β-hairpin after the nonequilibrium between state 1, 3, and/or 4. Although the pathways could be sorted into four types, every single peptide exhibits the fingerprinting pathways which resolves the conformation heterogeneity of peptides (Figure 4). ## Conclusions By confining a single peptide inside a SiNx nanopore, we have developed a highly detailed view of β-hairpin hairpin folding/unfolding with high spatial resolution of sub-nanometers, high time resolution of submilliseconds and high throughput. The nanopore experiment has revealed a multiplicity of folding/unfolding pathways with four dominant type for the model β-hairpin peptide from Protein G B1. Combining with the Markov chain model and first order law kinetics, the kinetic and mechanistic details of the folding/unfolding mechanism have been revealed, which enables the discovery of previous unknown pathways and substrates. We have found three major intermediates as compact, unfolding of hydrophobic cluster along with the loop vibration and unfolded ones correspond to states 1, 3, and 4, respectively. The hidden pathways of Pathways III and IV suggest that the unfold conformation of state 3 requires the high free energy to reform into the βhairpin. These nonobligate folding/unfolding pathway of Pathways II-IV may suggest that most possible metastable conformation of the misfolding peptide although with lower probability of occurrence. Therefore, our methods using an electrochemically confined space has shed new light on the complexity of a peptide and protein's folding/unfolding pathways. Future research aimed at increasing the bandwidth of the current recording would benefit a higher temporal resolution for solving every dynamic conformation. The thin solidstate nanopore also will provides a high spatial resolution for determination of metastable intermediates. After establishing a library for thousands of the single events, one may expect to predict folding state with the help of machine learning.

chemsum

{"title": "The hidden transition paths during the unfolding of individual peptides with a confined nanopore", "journal": "ChemRxiv"}

controlled_hierarchical_self-assembly_of_networked_coordination_nanocapsules_<i>via</i>_the_use_of_m

## Abstract: Supramolecular chaperones play an important role in directing the assembly of multiple protein subunits and redox-active metal ions into precise, complex and functional quaternary structures. Here we report that hydroxyl tailed C-alkylpyrogallol[4]arene ligands and redox-active Mn II ions, with the assistance of proline chaperone molecules, can assemble into two-dimensional (2D) and/or three-dimensional (3D) networked Mn II 24 L 6 nanocapsules. Dimensionality is controlled by coordination between the exterior of nanocapsule subunits, and endohedral functionalization within the 2D system is achieved via chaperone guest encapsulation. The tailoring of surface properties of nanocapsules via coordination chemistry is also shown as an effective method for the fine-tuning magnetic properties, and electrochemical and spectroscopic studies support that the Mn II 24 L 6 nanocapsule is an effective homogeneous wateroxidation electrocatalyst, operating at pH 6.07 with an exceptionally low overpotential of 368 mV. ## Introduction Hierarchical self-assembly via metal coordination is a ubiquitous process for constructing sophisticated supramolecular structures in nature. 1 As an example of its use in biological systems, metal coordination or bridging plays a crucial role in folding and assembling multiple protein subunits into precise, complex and functional quaternary structures (such as viral metalloproteins). 2,3 Metallosupramolecular assemblies such as metal-organic nanocapsules (MONCs) and/or nanocages are potentially useful models for such complex biological processes, 4,5 and are also promising with regard to energy storage, molecular encapsulation, catalytic, and biomedical applications. 19,20 To date, synthetic chemists have been able to isolate discrete cages consisting of more than 100 precisely designed units through metal coordination. 21 A longstanding challenge, however, is the rational combination of simple components to form hierarchical superstructures with a similar level of assembly complexity as proteins. 22,23 Another challenge that has seldom been addressed in the literature is redox-controlled metal-directed assembly. Albeit at a higher level of complexity, living organisms are able to rapidly select the oxidation state of metal ions such as Cu, Mn and Fe, with regard to protein subunit folding and assembly of quaternary structure, often with the aid of supramolecular chaperones. These metallochaperones are typically employed to capture, protect and insert the highly active metal ions into the specifc coordination sites before elements of the quaternary structure have formed through subunit self-assembly. 26 The powerful selfassembly approach utilised by biological systems may thus provide access to new hierarchical superstructures (HSSs) with unique properties. Our group (and others) have used C-alkyl-pyrogallol arenes (PgC n , where n is the number of carbon atoms in the pendant alkyl chains), bowl-shaped polydentate macrocycles, to synthesise MONCs via metal insertion. 13,27,28 This approach gives rise to large, discrete cages which typically have one of two highly conserved structures: a dimeric cage composed of 2 PgC n s seamed/bridged by 8 metal ions, or a hexameric cuboctahedral analog comprising 6 PgC n s and 24 metal ions (the latter of which form 6 triangular faces). These MONCs are readily accessible via ambient or solvothermal syntheses using redox stable metal ions such as Zn II , Ni II , Ga III . Variations in structure are also possible, for instance by replacing some pyrogallol rings with resorcinol in the PgC n framework, giving mixed macrocycles that cause 'defects' in the perfect MONC structure. 29 Despite the fact that these two general supramolecular architectures accommodate metals of different size and charge, the controlled assembly of redox-active transition metals has proven difficult. For instance, it has been shown that the reaction of Fe II or Mn II ions with PgC n s rapidly yielded MONCs with metal ions in mixed oxidation states. 31,32 Indeed, the assembly of mixed-valence MONCs, such as Mn II /Mn III , should be more kinetically favored than solely Mn II -based analogs since Mn II is more thermodynamically stable and kinetically labile than Mn III for coordination. 26 We only recently achieved the assembly of Co II hexameric MONCs by using a route inspired by zinc-fnger proteins (ZNFs). 33 In that case the Zn II ion was used to direct assembly of hexameric MONCs that were spontaneously transmetallated with Co II ions to afford the target assembly. Such results indicate that new MONCs with redox-active functionality may (as can be the case with biological systems) require additional templates or chaperones to control their assembly into the correct state. In this context, we are encouraged to challenge the synthesis of HSSs constructed from C-propan-3-ol-pyrogallol arene (PgC 3 OH) and coordination-inert but redox-active Mn II ions; the hydroxyl group on PgC 3 OH can link MONCs to obtain HSSs. 30 This may not only help to develop a better understanding of the redox-based self-assembly of metalloproteins, but also the construction of HSSs with emergent properties, such as magnetism and catalysis, based on the oxidation state distribution of the metal ions. Several reaction conditions and methodologies have been investigated to this end, yet all failed to deliver the selective assembly of any anticipated HSSs (see ESI † for details). We hypothesised that in situ redox reactions may prevent the formation of such highly intricate structures. Herein, we present a design strategy for the construction of such otherwise unobtainable HSSs that uses a reaction system consisting of PgC 3 OH, Mn II ions, and proline. The use of proline was inspired by the Mn II coordination sphere in manganese-based proteins, which may effectively capture and stabilise the free metal ion, as well as modulating its weak coordination ability with regard to metal insertion. 26,37,38 We propose a system in which PgC 3 OH is assembled into hexameric hydrogen-bonded nanocapsules (MONCs), whilst proline molecules act as the molecular chaperones to capture, protect and insert the Mn II ions into the framework (Scheme 1). Once formed, the thermodynamically and kinetically very stable MONCs serve as subunits (secondary structures) and organise into more complex HSSs through the formation of intermolecular metal-hydroxyl coordination bonds. Using this approach, we obtained 2D and 3D HSSs consisting of Mn II -seamed MONC subunits ( ## Results and discussions Compound 1 has been studied and characterised by scanning electron microscopy images (Fig. S1 †), single-crystal X-ray diffraction (SC-XRD, Fig. S2 †), FT-IR (Fig. S3 †), elemental analysis (EA), MALDI-TOF MS (Fig. S4 †), thermogravimetric analysis (TGA, Fig. S5 †) and differential scanning calorimetry (DSC, Fig. S6 †), details of which can be found in the ESI. † Compound 1 crystallises in the monoclinic space group P2 1 /n. The crystal structure of 1 shows a 2D framework constructed from infnite MONCs subunits, with each MONC being assembled from 30 components: six PgC 3 OH molecules and 24 metal ions (Fig. 1). The overall geometry of the MONC subunit corresponds to that of a truncated octahedron, which is similar to the previously reported hexameric MONCs. 28 Each hexagonal face of the MONC is capped by one [Mn 3 O 3 ] trimetallic cluster with Mn-O distances in the range of 2.03-2.11 A, Mn-O-Mn angles in the range of 133.17-137.50 , and O-Mn-O angles in the range of 99.34-105.66 . All Mn II ions adopt an octahedral ligand feld, where the equatorial positions are coordinated with oxygen atoms from the upper-rim of PgC 3 OH units. Bondvalence sum (BVS) analysis, coupled with examination of bonding energy reveals that all Mn ions are in the +2 oxidation state (Table S1 and Fig. S7 †). Interestingly, each MONC encapsulates two proline chaperone ligands that coordinate in Scheme 1 Pre-assembly strategy of Mn II -seamed MONC subunits used in this study. Color codes: carbon, grey; oxygen, red; Mn II , purple. a bridging mode between two Mn II ions (metal-carboxyl distances in the range of 2.24-2.27 A). This suggests that the proline molecules perform the critical function of a molecular chaperone, capturing, protecting and inserting Mn II ions into MONCs via ligand exchange during the assembly process. The extended view of 1 shows that each MONC is connected to four adjacent symmetry equivalents via double manganese-hydroxyl coordination (M-O distances: 2.26-2.35 A). One MONC provides a hydroxyl tail and a metal coordination site for another, and the other axial positions are occupied by water molecules. Introduction of a greater amount of water to similar reaction conditions as those used in the synthesis of 1 changed both the internal and external properties of the Mn II -seamed MONCs, resulting in the formation of a 3D HSSs which crystallises in an orthorhombic system (structure solution in space group Pccn, 2, Fig. 2, Table S2 and Fig. S8-S10 †). On the internal surface, all axial positions at the metal centres are occupied by water molecules, whilst inspection of the exterior reveals that each MONC subunit is linked to eight symmetry equivalents via single manganese-hydroxyl coordination bonds (two crystallographic M-O distances: 2.276 and 2.279 A, respectively), the result being assembly into a cubic tertiary structure (Fig. 2b). This supramolecular nanocube is assembled from 216 Mn II ions and 54 PgC 3 OH macrocyclic ligands and has an edge of 4.5 nm. Within the nanocube there are two types of MONC subunits with different orientation in the solid lattice, highlighted by the disparate colours in Fig. 2. This structural motif is similar to the unit cell of CsCl (Fig. 2c), and the extended view exhibits a hierarchical CsCl-like superstructure (Fig. 2d). Magnetic susceptibility data for 1 and 2 were recorded in the temperature range of 2.0-300 K in an applied magnetic feld of 1000 Oe. The c m , c m T vs. T plots for the complexes are shown in Fig. 3, where c m is the molar magnetic susceptibility. For supramolecular assemblies 1 and 2, the values of c m T at 300 K are 81.8 and 92.2 cm 3 mol 1 K, respectively, but lower than that of expected for the sum of the Curie constants for 24 noninteracting Mn II (s ¼ 5/2) ions, with g ¼ 2.00 (105.0 cm 3 mol 1 K). Upon cooling, c m T frst gradually decreases to a value of 76.1 cm 3 K mol 1 at 100 K, and then decreases more rapidly on further cooling to 27.3 cm 3 K mol 1 at 2.0 K for 1, however, c m T decreases to the minimum value of 29.4 cm 3 mol 1 K at 2.0 K for 2, indicating antiferromagnetic coupling within the Mn II ions. Above 50 K, the temperature dependence of c m 1 obeys the Curie-Weiss law with C ¼ 90.91 cm 3 K mol 1 and q ¼ 12.8 K above 2.0 K for 1 and C ¼ 102.88 cm 3 mol 1 K and q ¼ 41.8 K for 2 (see Fig. 3, inset). The negative q values confrm the antiferromagnetic coupling within the Mn II ions and the antiferromagnetic coupling in 2 is stronger than that in 1. Furthermore, the shapes of the M/H plots are quite like that of the antiferromagnet, in which the M values increase rapidly at low felds, with no obvious saturation observed up to 70 kOe (Fig. S11 and 12 †). Water oxidation (WO, 2H 2 O / O 2 + 4H + + 4e ) is regarded as a key half-reaction for solar fuel production. 39 The rational design and synthesis of cheap, efficient and stable wateroxidising catalysts are signifcant challenges in science and technology. 40 In nature, the oxygen-evolving complex (OEC, a CaMn 4 O 5 cluster) in photosystem II (PS II) can efficiently oxidize water. 41 It has been shown that the Mn IV -O-Mn III -H 2 O motif plays a crucial role in the activity of the OEC and its mimics. 42 Inspired by the OEC, several Mn clusters have been used as structural mimics. In particular, the presence of high oxidation state +3 and +4 Mn ions and four water binding sites have been applied for electrocatalytic oxidation of water, examples such as Mn 12 O 12 (OAc) 16x L x (H 2 O) 4 (L ¼ acetate, benzoate, benzenesulfonate, diphenylphosphonate, and dichloroacetate). However, the catalytic activity of these biomimetic Mn-based clusters for water oxidation was shown to be hindered by either high overpotentials (ranging from 640-820 mV) or low structural stability. 40 Kinetically and thermodynamically very stable Mn clusters assembled with exclusively Mn II ions may solve one of such problem even though a series of mononuclear manganese complexes [(Py 2 NR 2 )Mn II (H 2 O) 2 ] 2+ (R ¼ H, Me, tBu) were reported to be active in electrocatalytic water oxidation with an relatively high overpotential of approximately 800 mV (FTO working electrode). 34 However, to the best of our knowledge it remains unknown whether polynuclear Mn II clusters are capable of being highly active with respect to water oxidation. This long-standing question has been examined with 1 and 2 using electrochemical techniques. Crystals of 1 and 2 were dissolved in 0.1 M aqueous acetate buffer at pH 6.07 via sonochemistry, the pH at which the OEC within PSII shows optimal catalytic performance. 46 The resulting solutions of 1 and 2 were subjected to UV-Vis spectroscopy and showed two broad absorption bands at around l max ¼ 262 and 315 nm for 1 and l max ¼ 260 and 312 nm for 2, which can be assigned to the p-p* transition and ligand-to-metal charge transfer transition, respectively (Fig. S13 †). The redox peaks associated with manganese of 1 and 2 in aqueous acetate buffer have been detected via cyclic voltammetry (Fig. 4a, b and S14 †). These corresponded to the oxidation of Mn 2+ to Mn 4+ (E ¼ 0.87 V) and the reduction of Mn 4+ to Mn 3+ (E ¼ 0.83 V), Mn 4+ to Mn 2+ (E ¼ 0.55 V), and Mn 3+ to Mn 2+ (E ¼ 0.26 V). 47 The solution stability of the coordination structures was investigated using dynamic light scattering (DLS) techniques. It was shown that sonication of these solutions resulted in the formation of species in the size range of 2-3 nm, corresponding to the molecular hydrodynamic diameter of discrete MONCs (Fig. S15 †), 13 and implying that HSSs converted into discrete MONCs; we envisage that some metal-coordinated hydroxyl groups of PgC 3 OH moieties on axial positions may be displaced by water molecules. Interestingly, upon evaporation of an aqueous acetate buffer solution of 1 and 2, spherical, micron-scale metallosuperstructures were observed by SEM (Fig. 4c, d and S16 †). FT-IR and small angle X-ray scattering studies further supported that they were composed of many MONC subunits (Fig. S17 and 18 †). We propose that the hierarchical metal-organic micron spheriods (MOMSs) may be stabilized by a large number of van der Waals interactions between neighboring alkyl chains and hydrophilic regions of the discrete MONCs. Furthermore, cyclic voltammograms (CVs) clearly indicated that water oxidation can be catalyzed by both 1 and 2 (Fig. 5). 43,47 Water oxidation occurs at an exceptionally low overpotential of only 368 mV. This is higher than that of the current state-of-art Ru-bda complex (bda ¼ 2,2 0 -bipyridine-6,6 0 -dicarboxylate, 180 mV at pH 7), illustrating that there is still room for further improvements. 48 Continuous CV scan experiments and bulk electrolysis of 1 and 2 demonstrated that these electrocatalysts have high catalytic activity and stability toward water oxidation (Fig. S19 and 20 †). UV-Vis and DLS measurements taken after electrolysis of 1 and 2 showed that the waves and particle size are retained (Fig. S21 and 22 †). Moreover, the MOMSs re-formed and could be detected upon evaporation of the catalyst solution in subsequent SEM studies (Fig. S23 †). Collectively, these measurements suggest that the MONC subunit is a homogeneous water oxidation electrocatalyst. This result may thus provide a new strategy for the design and synthesis of cheap, efficient and stable water-oxidizing catalysts since it frst suggests that soluble Mn II clusters may be used to effectively facilitate the oxidation of water, despite the enormous efforts made to mimic the CaMn 4 O 5 cluster to date. We further envision that improvements of catalyst stability and activity may be possible. This may be achieved through (for example) attaching appropriate axial ligands to the constituent metal ions, or functionalizing alkyl chains on the MONC surface. In addition, other soluble metal-seamed dimeric or hexameric MONCs, such as those formed with Fe II , Co II and Cu II ions, are also promising with regard to electrocatalytic water oxidation. 36 Fig. 4 Cyclic voltammograms (CVs) of (a) 1 and (b) 2 (0.5 mM) in 0.1 M acetate buffer at pH 6.07 using an FTO (S ¼ 1 cm 2 ) working electrode. Scan rate is 50 mV s 1 . SEM images of hierarchical micron spheroids formed from an aqueous acetate buffer of (c) 1 and (d) 2. Fig. 5 CV scans of (a) 1 and (b) 2 (0.5 mM, 50 mV s 1 scan rate) in 0.1 M acetate buffer at pH 6.07. For comparison, CVs of the blank buffer are also shown. FTO (S ¼ 1 cm 2 ) was used as the working electrode.

chemsum

{"title": "Controlled hierarchical self-assembly of networked coordination nanocapsules <i>via</i> the use of molecular chaperones", "journal": "Royal Society of Chemistry (RSC)"}

interachem:_virtual_reality_visualizer_for_reactive_interactive_molecular_dynamics

## Abstract: Interactive molecular dynamics in virtual reality (IMD-VR) simulations provide a digital molecular playground for students as an alternative or complement to traditional molecular modelling kits or 2D illustrations. Previous IMD-VR studies have used molecular mechanics to enable simulations of macromolecules such as proteins and nanostructures for the classroom setting with considerable success. Here, we present the INTERACHEM molecular visualizer, intended for reactive IMD-VR simulation using semiempirical and ab initio methods.INTERACHEM visualizes not only the molecular geometry, but also 1) isosurfaces such as molecular orbitals and electrostatic potentials, and 2) two-dimensional graphs of time-varying simulation quantities such as kinetic/potential energy, internal coordinates, and user-applied force. Additionally, INTERACHEM employs speech recognition to facilitate user interaction and introduces a novel "atom happiness" visualization using emojis to indicate the energetic feasibility of a particular bonding arrangement. We include a set of accompanying exercises that we have used to teach chemical reactivity in small molecular systems. ## Introduction Traditional molecular representations perform well for pedagogical purposes on small model systems but may struggle to accurately depict the full breadth of possible chemical reactions or the dynamic nature of molecules. For example, Kekule-Lewis structures and arrowpushing mechanisms are widely used to conceptualize organic chemistry reactions, but fail to emphasize the three-dimensional nature of the structures. Conversely, modeling kits provide a spatial representation of molecules, but physical limitations (such as the number and placement of holes in each atom) can make it difficult to describe some motions or chemical reactions and impossible to describe exotic species (for example, CH5 + ). Students can overcome these obstacles with training, but advances in modern technology suggest more direct approaches such as the use of virtual reality (VR). The rise of relatively affordable commodity VR headsets in the last few years has seen the development of a plethora of VR-enabled molecular visualizers. This has in turn fueled the growth of VR applications and activities for chemical education. Several studies have featured interactive molecular dynamics (IMD) simulations, where the user is immersed in a molecular movie with the ability to grab and pull molecules and observe how the system responds in real time. Haptic controllers allow control over translations and rotations and additionally have force-feedback capabilities, making them the traditional choice of peripheral for IMD. However, these controllers have a limited range of motion, are prohibitively costly for most classrooms, and do not have the same widespread familiarity and versatility as VR headsets. Therefore, IMD in VR (IMD-VR) simulations seem to be a promising alternative for increasing the accessibility and impact of these simulations throughout chemical education. Smooth real-time simulations necessitate updates to the molecular geometry at least several times a second; therefore, evaluating the forces on the atoms must be done on the order of a tenth of second or faster. As a result, molecular mechanics (MM) are the most popular choice, enabling IMD to be applied to large macromolecules such as proteins and carbon nanotubes. 15,6,4 However, MM force fields are not well-suited for studying chemical reactivity as bonds cannot be created or broken; therefore, semiempirical and ab initio methods are needed for reactive IMD simulations. The density functional tight binding 16 (DFTB) method has been used to facilitate nanostructure construction and study chemical reactivity in systems with over 100 atoms. The use of consumer-grade graphical processing units (GPUs) enabled ab initio IMD (AI-IMD) for systems with dozens of atoms. 19 Reactive IMD-VR simulations have been reported previously, 4 but their potential impact for chemical education has not been fully explored. Here, we present the INTERACHEM molecular visualizer and a set of exercises developed with reactive IMD-VR simulations for both high school and undergraduate chemistry students. The exercises cover a variety of topics, including molecular geometry and structure, molecular bonding and orbitals, conformational changes, and acid-base and organic reactivity. A large focus was placed on the visualization of quantum mechanical quantities (e.g. bond order, molecular orbitals, electron density) as these quantities are not well-described with traditional molecular modelling kits or MM-based IMD-VR simulations. The INTERACHEM visualizer and an accompanying worksheet that we have used in previous classroom demos are both available for free online. 20 ## INTERACHEM -Virtual Reality Interactive Reactive Molecular Dynamics INTERACHEM is built with the UNITY game engine 21 and can be deployed to a wide variety of peripherals. Our primary development was done with the Oculus Rift S, although the Oculus Rift and Quest (in tethered mode via Oculus Link) have also been used. 22 Several stereoscopic (i.e. 3D) displays and projectors have been tested as well. The Rift S boasts "inside-out" tracking, meaning that no additional cameras or sensors are needed. This makes the INTERACHEM system fairly portable, which was useful as all three of our demonstrations required moving our simulation setup to a classroom. Additionally, haptic controllers such as the GEOMAGIC Touch 23 are supported through the OPENHAPTICS UNITY plugin and can be used alongside a VR headset. This gives the advantages of both systems (i.e. 3D perception and force feedback), but the haptic controller's limited range of motion is still a major disadvantage compared to the freedom of the VR controllers. A minimal control scheme was implemented with three functionalities: opening the menu (which pauses the simulation), selecting menu items or atoms, and panning/rotating the scene. The mappings of this scheme onto the VR handsets and haptic controllers are shown in Figure 1. Menu management (particularly drop-down menus for molecule and orbital selection) have been found to be challenging for users in VR. Table 1 lists a series of available voice commands using UNITY's built-in voice recognition module which has been added to circumvent some of these issues. Most notably, the "[load | get] molecule" command (where "molecule" is a specific molecular common name such as benzene or phenol) fetches a geometry via PubChem's Power User Gateway (PUG) REST API, 26 avoiding the onerous task of navigating a file explorer menu to start new simulations. The built-in recognition has some difficulty recognizing chemical names, and further work to resolve these issues and expand the capabilities of voice commands in INTERACHEM is ongoing. Molecules are visualized using standard ball-and-stick representations, with the bond order dynamically updated by rounding the corresponding entries of the Mayer bond order matrix 27 to the nearest half-integer. Atoms are highlighted when the controller hovers on them, and a yellow tether is displayed to indicate which atom is currently selected by the user. Isosurfaces for molecular orbitals, the full electron density, or the electrostatic potential can be displayed in realtime as shown in Figure 2. The user can request plots for kinetic and potential energy, temperature, or any internal coordinate (e.g. bond distance, angle, or torsion), as shown in Figure 3. These plots can be moved to any location in the VR scene, which we have found to be an excellent use for the additional space in the user's peripheral vision. As an additional heuristic to aid and engage students, a "happiness" mode was added where happy and angry emojis were attached to each atom based on the local bonding environment, as demonstrated in Figure 4. The target number of bonds for first and second row elements is the difference between the number of valence electrons and the nearest full shell of electrons. For third row elements, hypervalency is possible and the set of allowed bonds ranges from the ideal number described above to the total number of valence electrons in increments of 2 (as each electron in a lone pair forms a new bond). We can define the difference between the target number of bonds and the current number of bonds as (1) where is the closest ideal number of bonds and is the Mayer bond order between atoms i and j. The "mood" of an atom is then determined by ( 2) The partial atomic charges (determined via Mulliken population analysis) and total molecular dipole are also shown in this mode, which can be activated by holding both triggers on the two VR controllers simultaneously. The INTERACHEM visualizer is currently interfaced with two electronic structure packages. The xTB package 28 is linked into the executable and provides the GFN2-xTB semiempirical method. 29 Ab initio methods such as Hartree-Fock (HF), density functional theory (DFT), configuration interaction singles (CIS), and complete active space configuration interaction (CASCI) are available from the TERACHEM electronic structure package via the recently developed TERACHEM Protocol Buffer (TCPB) interface. The TERACHEM server currently needs to be run on a separate machine from the INTERACHEM visualizer as TERACHEM does not currently support Windows, but further work is ongoing to resolve this technical issue. The majority of the exercises below can be performed with either backend; however, the GFN2-xTB method from the xTB package is used by default as it is faster and therefore provides a consistently smooth user experience. The details of these interfaces and benchmarking of the electronic structure methods for reactive IMD-VR purposes will be described in an upcoming publication. The reactive IMD simulations use velocity Verlet integration, a Bussi-Parrinello Langevin thermostat, 32 and a multiple timestep scheme incorporating the user force as in the AI-IMD work. 19 We intentionally use a fast frictional relaxation time of 50 fs in the thermostat (i.e., this is a "strong" thermostat) order to compensate for the large forces (on the molecular scale) that are often induced by user interaction. By promoting fast energy dissipation, the possibility of molecular rupture due to unintentionally large user forces is minimized. A spherical reflecting boundary ensures that atoms are not lost when molecules do rupture after being subjected to high forces. All simulations are carried out in the gas phase, although implementations for the generalized Born/surface area (GB/SA) and several conductor-like polarization models are available in xTB and TERACHEM, respectively, paving the way for solvent phase reactive IMD in the future. The user-applied force is implemented as a spring between the selected atom and the controller. The default spring constant is set to comfortably manipulate internal motions (e.g. angles, dihedrals), and the user can scale the spring constant by a multiplicative factor to break bonds or move more massive atoms. In addition to reactive IMD, one can also perform geometry optimization via an internal BFGS optimizer. Normal modes can be visualized from an optimized structure. Atoms can be frozen in space to run constrained dynamics or optimizations, or directly moved while the simulation is paused. ## Teaching Chemistry with INTERACHEM -Overview of Sample Exercises We have developed a series of exercises for reactive IMD-VR designed to be taught by one instructor with a single INTERACHEM-enabled station. Our demonstrations used a gaming laptop, the Oculus Rift S headset, and a few modelling kits. This portable solution allowed us to visit classrooms easily. Students take turns completing exercises while the remaining students observe the simulation on a second display or projector (usually already available in the classroom). The exercises begin with several basic molecular geometry tasks, such as attempting to distort tetrahedral methane, the umbrella flip motion in ammonia, or pseudorotations in trigonal bipyramidal complexes such as PF5. These serve as a general introduction to VR and IMD, and also provide a launching pad for instructor-led discussions of chemical bonding concepts like Valence Shell Electron Pair Repulsion (VSEPR) theory 43 and the pitfalls of Kekule-Lewis structures and physical modelling kits. The next theme is reactivity and molecular orbitals, involving challenges such as comparing C-C torsion for ethane vs. ethene, visualizing s and p orbitals in ethene and benzene, and promoting the ring opening of cyclobutene (as showcased in Figure 5). Several exercises focus on benzene and cyclohexane, including a discussion of molecular planarity, aromaticity and the p system in benzene (in which students attempt to throw a hydrogen atom through the ring), and converting between the chair and boat conformers of cyclohexane. Finally, chemical reactivity is first shown through proton transfers in several acidbase systems and then through standard organic reactions such as nucleophilic substitution and elimination reactions. Links for videos depicting each exercise can be found listed in Table 2 and a full discussion of the exercise goals, typical outcomes, and talking points can be found in the Supporting Information. We first presented these exercises to 15 undergraduate students in a section of Chem 173 at Stanford University, the physical chemistry class introducing quantum mechanics. Together, the students completed the majority of the exercises in the 50-minute section and spent roughly equal time becoming comfortable with the VR setup as performing the exercises. The discussion was tailored to build on their chemical and mathematical background. For example, harmonic oscillators as models of vibration were included in the context of normal mode analysis and molecular mechanics. Additionally, wavefunction-related concepts such as the nodes and energy ordering of molecular orbitals and the electron density were highlighted. We also demonstrated reactive IMD-VR to high school students on two occasions. A total of 46 high school students (primarily 11 th and 12 th graders) were split over four 50-minute sessions as class C7382 at Stanford SPLASH Fall 2019. These sessions had two INTERACHEM stations and three instructors, giving small groups with five or six students per setup. Only half the exercises were run with a higher focus on molecular geometry (e.g. ammonia umbrella flip, cyclohexane conformers) and single vs. double bonds (e.g. ethane vs ethene torsion). The final challenge was setup as competition between the two teams with the goal to perform an SN2 reaction in a hydroxide and methyl fluoride system. Most groups were able to discover the mechanism in reactive IMD-VR after three or four tries despite having no organic chemistry instruction. Finally, an extended 1.5-hour workshop was run for 17 high school students from 9 th to 12 th grade with one INTERACHEM setup. Four of these students evidently had previous exposure to VR headsets and were immediately comfortable in the simulation. In general, the high school students quickly acclimated to VR environment but spent longer completing each exercise compared to the undergraduates; from our experiences, high school students need at least two hours to complete all exercises that the undergraduates completed in 50 minutes. We found the energy plots and molecular modeling kits to be valuable complementary tools for the instructor and external students. The instructor can point out energy barriers and relative stability of the conformers in real time, while the modeling kits provided reference structures to anchor students to their previous lessons. For students who completed exercises quickly, the challenges were further "gamified" by adding additional constraints, such as only allowing them to touch certain atoms or reducing the user-applied force. In the future, these concepts could be quantified to provide a numerical score for each challenge, where the lowest score represents the discovery of the minimum energy mechanism with the smallest amount of external force. Conversely, students also greatly enjoyed using high forces to easily pull apart molecules. In addition to increasing student interest, relevant discussion can be generated by focusing on the recombination products in these scenarios. almost 80 students across both high school and undergraduate levels. In general, both groups of students responded well to the VR simulations, although high school students took longer to complete the exercises. However, this does not mean the high school students had a less successful experience; on the contrary, the high school students had animated discussions and provided creative (even if unintended) solutions. Although not quantified here, the concept of "gamification" for reactive IMD-VR simulations in the form of minimizing the user-applied force or constraining which atoms can be manipulation emerged naturally during our demonstrations. Gamification has seen significant success for both increasing participation and providing creative solutions to scientific problems such as protein folding. Further study is warranted on whether the gamification of reactive IMD-VR simulations leads to increased student participation or greater completion of learning objectives. ## Supporting Information Document including in-depth discussions of each sample exercises, as well as the worksheet with sample exercises discussed in the text. ## Section S1. Curriculum Exercises The following sections provided additional discussion for each exercise, including the goals for students, talking points for instructors, and typical outcomes. ## Section S1.1. Distorting Tetrahedral Methane This first exercise is intended to give students the time to familiarize themselves with both virtual reality (VR) and interactive molecular dynamics (IMD) simulations, while providing an opportunity for the instructor to gauge the students' level of chemical knowledge. The goal is for students to drag all four hydrogens into a planar conformation, typically by freezing two hydrogens in-plane with the carbon and then attempting to pull the final hydrogens into position. Instructors can discuss the 3D nature of molecular structure, ask about Valence Shell Electron Pair Repulsion (VSEPR) theory or orbital hybridization, and introduce the potential energy graphs as a quantitative measure of molecular stability (or lack thereof, in the case of planar methane). Students typically only see a planar methane for a split second, reinforcing the stability of tetrahedral methane. ## Section S1.2. Umbrella Flip in Ammonia The umbrella flip in ammonia is designed to introduce some of the limitations of chemical models the students may be familiar with. The goal is to invert the hydrogens in ammonia without simply rotating the molecule, a motion that is typically not possible with physical molecular modeling kits. Further discussion may build upon VSEPR and hybridization from Section S1.1, but with the introduction of lone pairs. Since applying only two forces in ammonia necessarily applies a torque, students often rotate ammonia several times before a successful inversion event. As an additional challenge, the instructor can impose the restriction that only the nitrogen atom can be touched (usually also turning up the force multiplier to 4x as a concession). This secondary exercise is also typically completed by building up momentum in the symmetric bend vibrational mode, providing an opportunity for instructors to talk about normal modes. ## Section S1.3. Psuedorotation of PF5 and PH3F2 The goal of this exercise is to introduce students to basic inorganic chemistry concepts such as hypervalency and pseudorotations of trigonal bipyramidal complexes. During our three outreach sessions, this section was located later in the curriculum and significantly less time was spent discussing these concepts as we often ran out of time. This is the best chance for an instructor to launch discussions surrounding inorganic chemistry, or heavier elements such as transition metals, as the other exercises feature solely organic molecules. ## Section S1.4. Ethane C-C Torsion The students are tasked with performing a torsional scan around the carbon-carbon single bond in ethane, which is fairly easily completed. If students have difficulty twisting the two halves of the molecule independently, one half can be frozen to help prevent the torsion of the entire molecule. Instructors can focus on the potential energy graph during the scan, as well as introduce organic chemistry concepts such as eclipsed vs. staggered conformers and gauche interactions. ## Section S1.5. Ethene C=C Torsion This is a partner exercise to the ethane torsion in Section S1.4; this time, students are attempting to perform a torsional scan around the carbon-carbon double bond in ethene. Unlike the previous exercise, students are not typically successful with this torsion. Typical outcomes either involve a temporary success followed by the ethene untwisting itself, or a different mechanism where one hydrogen is almost completely removed from the ethene and then reattached to the other side. Instructors can turn on orbital isosurface rendering for the highest occupied molecular orbital (HOMO) to discuss the difference between sigma and pi bonds as an explanation for the difference between the two exercises. ## Section S1.6. Ring opening of cyclobutene In the first reactive exercise, students are tasked with performing a ring-opening reaction in cyclobutene. As with the ethane torsion in Section S1.4, students usually have no problems performing this task, but the exercise serves as a great launching pad for discussion. The HOMO can be visualized and there are several key differences when comparing to the orbitals seen in Section 1.5 with the ethene pi bond. During the course of the reaction, the molecular orbitals fluctuate greatly and may not always correspond to chemically intuitive ideas of orbitals. For more advanced students, this can lead to discussions of the properties and origins of molecular orbitals and caution against using blindly orbitals as a quantitative tool. The overall trend of evolution from carbon-carbon sigma bond, to carbon p orbitals, to the formation of the new conjugated pi bond network depicted in Figure 5 can also serve as a lead-in to orbital symmetry rules (i.e. the Woodward-Hoffman rules) and mechanochemical concepts such as conrotatory vs disrotatory pathways. ## Section S1.7. Hydrogen atom through benzene The goal of this exercise is for students to push a hydrogen atom through the middle of a benzene ring. The starting geometry has been set such that with a significant amount of force, students can often achieve this on their first try; however, they will have much more difficulty on subsequent attempt as it is harder to achieve the proper alignment. Typically, these subsequent attempts will result in either complete repulsion or the occasional attachment/detachment of the hydrogen to the ring rather than pass through the center of the ring. Instructors can visualize the pi cloud above and below the benzene ring, showcasing how the apparent hole is not as large due to the significant electron density. As an aside, benzene also serves as an excellent reference for discussing aromaticity. The non-traditional nature of this exercise means that most students will not have a preexisting intuition for the outcome, making it an interesting testbed for learning new phenomena in VR-IMD simulations (compared to molecular modeling kits and Kekule-Lewis structures, which do not have much predictive power in this case). ## Section S1.8. Cyclohexane chair/boat conformer flip This exercise tasks the students with performing a chair-to-boat conformation flip in cyclobutene. There are typically two major blocking points for students, usually depending on whether they have seen this system before. Students who have never been seen the two structures may have some understanding the two conformers. Molecular modeling kits are extremely helpful in giving reference structures for those not in the VR simulation. The second issue is actually performing the conformer flip. Students usually only attempt to force the flip by pulling only on the carbons, which only works with an increased force modifier. A more successful strategy is to alternate pulling on the carbons and also adjusting the positions of the hydrogens in the half of the cyclohexane that is flipping up. As an instructor, providing these suggestions to the students leads naturally into discussion about equatorial vs. axial hydrogens and steric interactions. Section S1.9. Proton transfer in carbonic acid + water This is the first exercise designed for a more open-ended reactive experience. Students are placed in a simulation with a carbonic acid (H2CO3) and four water molecules and tasked with transferring a proton from one molecule to another. Students can easily try multiple mechanisms but the most successful is pulling one proton from the carbonic acid onto a water. After several seconds, the proton will either hop back directly or a small proton relay will occur, typically ending with transfer back to the carbonic acid. Instructors can use this to discuss acidbase reactions, equilibrium constants, and proton wires (and also allude to the biochemical implications). One can also mention the role of solvent in tautomerization reactions, as these are usually drawn as the same hydrogen when that is often not the case. ## Section S1.10. SN2 of methyl fluoride + hydroxide The goal is for students to carry out the prototypical organic reaction of SN2 nucleophilic substitution. As with the hydrogen through benzene exercise in Section S1.7, the initial arrangement of the molecules makes it simple to succeed with the correct starting momentum, but difficult to realign for subsequent reactions. The alignment is mostly complicated by hydrogen bonding between the hydroxide and methyl group (for the forward reaction) and the fluoride ion and hydroxy group (for the reverse reaction). Given the strong connections to the standard organic chemistry curriculum, instructors can discuss the strength of nucleophiles and leaving groups while comparing the forward and reverse reactions, or the hydrogen inversion and reaction mechanism in general. ## Section S1.11. E1 of t-butyl fluoride + hydroxide This is a partner exercise to the SN2 reaction in Section S1.10, where instead the system uses t-butyl fluoride instead of methyl fluoride. This is nominally listed as the E1 reaction, but in reality both SN1 and E1 reactions occur frequently while the SN2 and E2 reactions can be triggered by a careful user. As with Section S1.7 and Section S1.10, the initial alignment favors the SN1 reaction if the student is quick to apply force to the fluoride leaving group and hydroxide nucleophile; otherwise, students typically perform the E1 reaction first as the alignment for the substitution reaction quickly degrades. This exercise is an excellent starting point for discussion about reaction competition and selectivity, as well as comparing the sterics of the t-butyl and methyl groups. Section S1.12. High Force Decomposition In our experience, students always become excited when they realize that applying high forces to completely destroy the molecule is an option. We've found the t-butyl system from Section S1.11 to be the best candidate system since it is the last exercise in the planned curriculum and also features the richest recombination products. It is also worth mentioning that high forces in the ethene torsion of Section S1.5 can sometimes see the formation of acetylene and hydrogen gas, which can also contribute to discussions about the relative strengths of bonds (i.e. it is easier to break carbon-hydrogen single bonds than it is to break a carbon-carbon double bond). ## VR Instruction Hardware Instruction: 1. General: Press the menu button, hit the "Molecule" using the trigger button, in the "Molecule" dropdown list select the one you are interested. Exit the menu by hit the menu button again. Reach your controller INTO one atom, and drag it using the trigger button. You can use two controllers to drag two atoms simultaneously. Move the whole system using the grab button. Rotate the whole system by holding both grab buttons. Discover other functions yourself! ## Add a new molecule: Generate an xyz file of your molecule or system. It is recommended that your system contains less than 20 atoms. Open "VROption.txt", under the line "#ifdef Molecule", add the complete absolute path of you xyz file. It will occur in the "Molecule" dropdown list. If your molecule or system has a net charge (-1 for example), in the second line of your xyz file, which is intended for a description, write "charge=-1.0". Note: Current backend is xtb 6.3 (semiempirical, https://github.com/grimme-lab/xtb ). The TeraChem backend (HF) wil come soon.

chemsum

{"title": "InteraChem: Virtual Reality Visualizer for Reactive Interactive Molecular Dynamics", "journal": "ChemRxiv"}

a_ratiometric_nmr_ph_sensing_strategy_based_on_a_slow-proton-exchange_(spe)_mechanism

## Abstract: Real time and non-invasive detection of pH in live biological systems is crucial for understanding the physiological role of acid-base homeostasis and for detecting pathological conditions associated with pH imbalance. One method to achieve in vivo pH monitoring is NMR. Conventional NMR methods, however, mainly utilize molecular sensors displaying pH-dependent chemical shift changes, which are vulnerable to multiple pH-independent factors. Here, we present a novel ratiometric strategy for sensitive and accurate pH sensing based on a small synthetic molecule, SPE1, which exhibits exceptionally slow proton exchange on the NMR time scale. Each protonation state of the sensor displays distinct NMR signals and the ratio of these signals affords precise pH values. In contrast to standard NMR methods, this ratiometric mechanism is not based on a chemical shift change, and SPE1 binds protons with high selectivity, resulting in accurate measurements. SPE1 was used to measure the pH in a single oocyte as well as in bacterial cultures, demonstrating the versatility of this method and establishing the foundation for broad biological applications. ## Introduction As a measure of proton activity, pH is a universally important parameter of our aqueous environment and biological milieu. 1 In living organisms, acid-base homeostasis is essential for maintaining physiological functions and therefore requires tight regulation. 2 A disrupted pH balance is associated with various abnormal states in biological systems. For example, low pH in humans has been linked to pathological conditions such as cystic fbrosis, ischemia and cancer, 3 whereas elevated pH (alkalosis) may lead to hyperphosphatemia and hypocalcemia. 4 The development of in vivo pH detection methods is currently of great importance for understanding the physiological roles of pH homeostasis as well as for disease diagnosis and therapeutic monitoring in cases where pH variation is a hallmark of the abnormality. It is possible to measure the pH in tissues by using conventional pH microelectrodes 5 but the invasiveness and lack of spatial resolution is a major limitation. In contrast, fluorescence and bioluminescence imaging with optical pH sensors can report on pH with high spatial and temporal resolution, 6 but they are restricted to superfcial imaging depths due to light scattering and absorption. Although elegantly designed proofof-principle methods based on other detecting techniques have emerged, 7 noninvasive, accurate, and sensitive methods to measure the pH of living organisms remains an urgent challenge. Magnetic resonance (MR) based techniques can offer unlimited tissue penetration in a truly non-invasive manner, and versatile MR read-out methods are established for both spectroscopic and imaging purposes. 8 The recent development of MRI contrast agents based on pH-dependent relaxivity 9 and chemical exchange saturation transfer (CEST), 10 which uses the saturation transfer of exchangeable protons to water, offer promise for in vivo pH mapping. These methods, however, often require specifc calibration or external standards and high accuracy is difficult to achieve. The conventional and most widely used NMR and MR spectroscopic imaging (MRSI) methods for measuring pH rely on sensors that exhibit pHdependent chemical shift changes, which can be monitored by 1 H, 13 C, 19 F or 31 P NMR signals. 11 These pH sensors are typically small molecule acids or bases, such as phosphate 12 or imidazole 13 derivatives, with a pK a compatible with physiological conditions. They exist as a mixture of protonation states in vivo but exhibit only one set of NMR signals because the chemical exchange between these states is faster than the NMR time scale. 14 The protons are highly mobile and rapid (de) protonation is facilitated by the hydrogen bond network of hydrated H + in aqueous media, such that it exceeds the speed of diffusion. 15 The observed average chemical shift of the conventional pH sensors is determined by the relative population of the protonated and unprotonated states and thus reflects the pH in solution. However, chemical shift is susceptible to artifacts caused by variations in ionic strength, local magnetic susceptibility, etc. In addition, the proton binding site (lone pair) of regular pH sensors will unavoidably be involved in interactions with metal ions, which will also induce pH-independent chemical shift changes and therefore experimental errors. 16 An innovative strategy involving hyperpolarized 13 C NMR techniques based on the pH-dependent equilibrium between carbon dioxide (CO 2 ) and bicarbonate (HCO 3 ), which are in slow exchange in vivo, was recently explored. 17 This approach, however, relies heavily on the carbonic anhydrase enzyme that catalyzes the interconversion between CO 2 and HCO 3 ## À . These species are also components of pH-independent biomolecular processes, and the CO 2 partial pressure is affected by the gas/solute equilibrium. 18 In another strategy, a pilot study showed the possibility of 19 F NMR pH sensing by ratio, when fast proton exchange is coupled with slow dissociation of intramolecular metal-ligand binding. 19 The interaction of metal with other coordinative species in the aqueous media, such as HCO 3 , however, perturbs the equilibrium between different protonation states. 20 An ideal ratiometric MR pH sensor should have a slow proton exchange (SPE) on the NMR time scale, but still be fast enough for real time pH monitoring, and more importantly, its protonation equilibrium should not be affected by any factor other than pH. In this paper, we report the frst ratiometric 1 H NMR pH sensing strategy to meet these criteria, based on a synthetic pH sensor, SPE1. This novel sensor is a cage-shaped urea cryptand with high proton selectivity and exhibits unusually slow interconversion rates between the different protonation states, which produce distinct NMR signals, allowing highly accurate ratiometric pH measurements. We demonstrate that this novel pH sensor is biocompatible and can be applied to monitor the pH in living biological systems, including fsh oocytes and bacterial cultures. ## Principle and design of the SPE pH sensing strategy The rapid chemical exchange between the non-protonated (B) and protonated (BH + ) states of conventional pH probes makes it difficult to accurately measure the ratio of [B]/[BH + ] directly by NMR, which is needed to calculate the pH value with the Henderson-Hasselbalch equation: pH ¼ pK a + log[B]/[BH + ]. 21 In contrast, SPE in protein structures is well documented. 22 While amide or alcohol protons on the surface of a protein are in fast exchange with the surrounding aqueous solution, protons from similar groups deep in the protein core have restricted mobility due to the hydrophobicity of the local environment as well as their involvement in intramolecular hydrogen bonds. 22 It is in principle possible to slow down proton exchange in synthetic molecules by introducing a sterically hindered hydrophobic environment and neighboring hydrogen bond acceptor groups that mimic protein structures. Small molecules with slow proton exchange however, are rare and have only been sporadically reported in the literature as unexpected fndings. 23 No systematic study has been conducted on exploring this unusual phenomenon. One molecule that displays such slow proton exchange properties is a tris-urea cryptand (1,4,6,9,12,14,19,21-octaazabicyclo[7.7.7] tricosane-5,13,20-trione) 23b,c which we named SPE1 (Fig. 1). Both the bridgehead N-atoms in SPE1 adopt an endo conformation with the lone pair electrons pointing inside the molecular cavity. Upon protonation, the incoming protons are trapped inside the cage and stabilized in this position through intramolecular hydrogen bonding with the ureido oxygen atoms (Fig. 1). 23c The proton transfer is sufficiently slow to allow direct NMR observation of both the protonated and the neutral forms of SPE1. The ratio between these two forms can therefore be used for accurate pH sensing. In addition, the size of the cryptand cavity is too small to bind any ions larger than H + , including Li + , the smallest metal cation. 23b This minimizes the interaction with ions, which can perturb the chemical shift of conventional NMR pH sensors aforementioned. Other advantageous features of SPE1 include a pK a close to physiological pH and good water solubility. Moreover, because the molecule exhibits a mirror plane and C 3 symmetry, the NMR spectrum is simple and unambiguous for peak assignment. Only 3 peaks are detected in the 1 H NMR spectrum of neutral SPE1 in aqueous solution, one peak corresponding to the 6 urea protons and two peaks for 12 methylene protons each. Having more chemicallyidentical protons contributing to the intensity of a single peak in the spectrum increases sensitivity, which is one of the most common limitations of NMR. ## Synthesis of pH sensor To test the applicability of the SPE strategy for pH sensing, a novel synthetic route was implemented to generate SPE1 in 3 steps, with a 38% overall yield (Scheme 1). SPE1 was synthesized from a tripodal amine, tris(2-aminoethyl)amine (tren), which was readily converted into an isothiocyanate derivative (3) upon treatment with carbon disulfde and N,N 0 -dicyclohexylcarbodiimide (DCC) in 78% yield, according to our previously published procedure. 24 Rapid coupling of the isothiocyanate compound 3 with the trivalent amino counter partner, tren, under high dilution conditions generated the C 3 symmetrical thiourea compound 2 in nearly quantitative yield. The thioureido groups in 2 were then converted to more water-soluble ureido analogs based on a reaction reported by Mikolajczyk in 1972, 25 in which DMSO acts as the oxidant and solvent, in the presence of an acid catalyst. This convenient synthesis allows production of SPE1 in large scale, facilitating the following pH sensing studies. ## Measurement of pH Both bridgehead nitrogen atoms of SPE1 can be protonated under acidic conditions. Due to slow chemical exchange, the neutral SPE1 and its bis-protonated form (SPE1H 2 2+ ) are simultaneously detected by 1 H NMR as distinct species in aqueous solution. Notably, the mono-protonated form of SPE1 (SPE1H + ) was not observed by NMR, due to the strong positive cooperativity in protonation (pK a2 > pK a1 ). 23c This property greatly simplifes the NMR spectrum, as both neutral and bisprotonated SPE1 are highly symmetrical, enhancing the sensitivity of NMR signal detection. For pH calculations, a modifed Henderson-Hasselbalch equation, which takes into account both protonation steps of SPE1, was used based on the ratio of neutral and bis-protonated SPE1: In order to determine the apparent pK a , (pK 0 a ¼ 1/2(pK a1 + pK a2 )) and further demonstrate that SPE1 can be applied for accurate ratiometric pH sensing, a series of 1 H NMR spectra of SPE1 dissolved in phosphate buffer at several pH values between 6.5 and 9.5 were collected (Fig. 2a). At room temperature, under basic conditions (pH $ 9), SPE1 is predominantly in the neutral form, producing two 1 H NMR peaks at 2.58 and 3.13 ppm for both of the bridge methylene units (-CH 2 -CH 2 -). ]), obtained from the NMR integrals (Fig. 2b and ESI †). A similar titration curve was obtained at 37 C. The apparent pK a for SPE1 was 8.00 AE 0.06 at 25 C and 7.72 AE 0.07 at 37 C. These numbers are in agreement with the pK a determined by potentiometric titration, 23b confrming that the pH measured by the current ratiometric approach is comparable to the conventional pH electrode. At body temperature, SPE1 can operate as a pH sensor between pH 6.7 to 8.7, covering slightly acidifed to mildly basic conditions. In addition, this method is very sensitive. Within a pH window close to the pK a of SPE1, differences as small as 0.02 pH units could be experimentally observed (Fig. 3). ## Biological applications of SPE1 In-cell pH detection. To demonstrate that this novel ratiometric approach is suitable for measuring pH in living cells, SPE1 was applied to measure the intracellular pH in a Belonidae oocyte. Live oocytes are widely used as model organisms for drug screening and to study reproduction and development. 26 The popular platform, Xenopus laevis oocytes, require hundreds of cells for NMR acquisition. 27 The Belonidae oocyte chosen for this study has an average diameter of 3 mm. Using a 4 mm MAS NMR probe, the pH was measured in a single oocyte microinjected with a low mM solution of sensor. From the peak ratio, an intracellular pH of 7.50 was obtained (Fig. S2 †), which was confrmed by using a pH electrode in cell lysates. This intracellular pH is in line with previous measurements acquired by different methods on oocytes of other species. 28 Monitoring pH change in Escherichia coli cultures. To demonstrate that the novel ratiometric pH sensing strategy can be applied to different biological systems, we used SPE1 to monitor real time pH changes in live bacterial culture. The biocompatibility of SPE1 was frst tested on Escherichia coli (E. coli). A 1.8 mM solution of SPE1 in phosphate buffer was added to cultured E. coli MC4100 cells (OD 600 ¼ 1) and incubated at 37 C for 12 hours. The viability of the sensor-treated cells was not signifcantly different compared to non-treated cells (ca. 7.3 10 7 CFU ml 1 for both samples). The cells incubated with SPE1 were concentrated, washed and placed into a 4 mm rotor and subjected to NMR experiment. The observed NMR signals of the pH sensor indicate that SPE1 is cell permeable (Fig. S3a †). After the NMR experiment, the cells were washed again with PBS. They were then subjected to further NMR experimentation and the NMR signals of SPE1 disappeared, suggesting that SPE1 can readily come out of the MC4100 cells (Fig. S3b †). The diffusion editing method revealed that the sensor is freely diffusing after cell uptake (Fig. S4 †), suggesting no specifc binding of SPE1 to bio-macromolecules in E. coli. Overall SPE1 causes no observable toxicity in E. coli cells. Various microorganisms, including E. coli cells can grow in both aerobic and anaerobic culture, and are known to increase production of acidic metabolites in response to low oxygen stress. 29 To monitor this process in real time by NMR, we conducted a kinetic study of concentrated E. coli culture (1 ml aliquot at OD 600 ¼ 1) in a sealed 4 mm NMR rotor at 37 C and recorded the change in pH over time using SPE1 (1.8 mM, Fig. 4). An initial pH of 7.55 was determined from the intensity ratio 31/69 (31% for neutral SPE1). The solid NMR rotor insert remained sealed in the spectrometer and new 1 H NMR spectra were acquired every 15 minutes. A continuous slow increase in the intensities of the SPE1H 2 2+ peaks with a diminution of the peak intensities of SPE1 was observed. The high accuracy of the SPE-based allowed precise measurements of small pH changes over 3 hours from pH 7.55 to 6.95 (Fig. 4). Interestingly, in conjunction with the gradual decrease of pH, two new sharp peaks appeared in the 1 H NMR spectra and increased in intensity over the course of the experiment. The chemical shifts of 2.40 and 1.92 ppm of these singlet peaks are consistent with succinate and acetate, which are common metabolites observed in bacterial cultures growing with limited oxygen availability. 30 It is known that bacteria modify their metabolism upon switching from aerobic to micro-aerobic or anaerobic conditions, by increasing the glycolysis rate with a concomitant decrease of acetyl-CoA degradation by the citric acid cycle. 31 This adjustment causes an overall increase in proton concentration as well as other acidic metabolites such as acetate and succinate. 32 Therefore our experiments confrmed that SPE1 was able to accurately monitor pH changes in real time in a biocompatible and reproducible manner and recorded the alteration of metabolism in live bacterial cultures deprived of oxygen. The current setup does not allow determination of the precise location of SPE1 within cells. Future work will involve the development of new SPE-based pH sensors with controllable cell-permeability and subcellular localization. ## Experimental Details for general experimental procedures, syntheses and characterization of all compounds can be found in the ESI. † All 1 H NMR spectra were manually corrected for phase and ## Measurement of intracellular pH of Belonidae oocytes Sample preparation. Freshly produced, unfertilized Belonidae oocytes ($3 mm in diameter) were washed with OR-2 buffer 33 and used within 2 days. Each oocyte was microinjected with 2 ml of a solution of 0.7 M SPE1 with 0.05% phenol red. A single oocyte was used for each measurement. One oocyte without sensor injection was scanned as a control. NMR experiments. The 1 H NMR experiments were performed on a Bruker Avance III 500 MHz spectrometer, using a prototype CMP MAS 4 mm 1 H-13 C-19 F-2 H probe ftted with an actively shielded Z gradient (Bruker BioSpin) at a spinning speed of 1000 Hz. The oocyte was placed into a 4 mm o.d. zirconium rotor with 10 ml D 2 O and the experiments were locked using D 2 O solvent. Water suppression was achieved using the purge pulse sequence. 34 All spectra were recorded with 256 scans, recycle delay set at 5 T 1 and $4 ms 90 pulse widths for the blank and injected oocyte experiment respectively. 32 768 time domain points were acquired for each spectrum with a spectral width of 20 ppm. Data were zero flled and multiplied by an exponential window function corresponding to a 1 Hz line broadening in the transformed spectrum. Real time pH monitoring of E. coli culture Sample preparation. E. coli MC4100 cells transformed with a pBAD24 plasmid (to confer ampicillin resistance) were plated on solid LB-agar medium supplemented with ampicillin and grown overnight at 37 C. LB media and agar were purchased from Bioshop Inc. and used as received. One colony was transferred from the plate into a 50 ml culture of LB-Amp liquid medium and grown for approximately 16 h. The overnight cultures were used to inoculate fresh liquid cultures, which were grown at 37 C to an OD 600 < 1. A 1 ml aliquot of the culture was centrifuged at 10 000g and re-suspended in 30 ml of a 1.8 mM solution of SPE1 in 10 mM phosphate buffer pH 8.0 containing 10% D 2 O. Another aliquot was collected and re-suspended in phosphate buffer to act as a blank for the NMR experiment and a control for cell viability over the course of the experiment. The sample was transferred to an NMR top insert made from Kel-F, sealed with a Kel-F sealing screw and cap, then inserted into a 4 mm o.d. zirconium rotor for the NMR experiment. To test for viability of the sensor-free and sensor-treated cells, the cells were serially diluted 10 3 to 10 9 times in NMR experiments. The 1 H NMR experiments were performed on a Bruker Avance III 500 MHz spectrometer, using a prototype CMP MAS 4 mm 1 H-13 C-19 F-2 H probe ftted with an actively shielded Z gradient (Bruker BioSpin) at 37 C. The samples were all spun at a spinning speed of 6666 Hz and all experiments were locked using D 2 O solvent. Water suppression was achieved using water suppression by gradient-tailored excitation (WATERGATE) and was carried out using a W5 pulse train. 35 All spectra were recorded with 256 scans, recycle delay set at 5 T 1 , 5.8 ms 90 pulse widths and collected using 32 768 time domain points with spectral widths of 20 ppm. Data were zero flled and multiplied by an exponential window function corresponding to a 1 Hz line broadening in the transformed spectrum. ## Conclusions We reported a novel and versatile strategy for ratiometric 1 H NMR pH sensing based on a slow proton exchange (SPE) mechanism. A water-soluble small molecule cryptand SPE1 was prepared through a new synthetic route and was evaluated in vitro and in live cells for ratiometric NMR pH sensing. Slow chemical exchange between different protonation states and high proton selectivity of SPE1 were achieved by shielding the incoming protons inside the small molecular cavity and trapping them with intramolecular hydrogen bonding. Unlike typical small molecule acids or bases, which exhibit a single set of average NMR signals, SPE1 displays distinct peaks for the neutral and protonated forms due to unusual slow chemical exchange. It is therefore possible to use the ratio of NMR peak intensities to provide highly precise pH values of the aqueous media. The new approach is more robust, sensitive and accurate than conventional chemical-shift based methods, which are vulnerable to many pH-independent factors. SPE1 exhibits an apparent pK a value suitable for biological applications and shows no toxicity effects on cell cultures. Therefore the new method was applied to measure the pH in a single live fsh oocyte, and to monitor the real time pH changes of a bacterial culture. Overall, SPE1 has great potential for measuring and mapping pH and pH changes in living systems. Next generation pH sensors based on the SPE mechanism are currently under development to cover different pH windows, which can further expand the scope of biological applications of this new strategy.

chemsum

{"title": "A ratiometric NMR pH sensing strategy based on a slow-proton-exchange (SPE) mechanism", "journal": "Royal Society of Chemistry (RSC)"}

selective_chemical_functionalization_at_n_6_mda_residues_in_dna

## Abstract: Selective chemistry that modifies the structure of DNA and RNA is essential to understanding the role of epigenetic modifications. We report a visible-light-activated photocatalytic process that introduces a covalent modification at a C(sp 3 )-H bond in the methyl group of N6-methyl-adenosine-an epigenetic modification of emerging importance. A carefully orchestrated reaction combines reduction of a nitropyridine to form a nitrosopyridine spin-trapping reagent and an exquisitely selective tertiary aminemediated hydrogen-atom abstraction at the N6-methyl group to form an a-amino radical. Cross-coupling of the putative a-amino radical with nitrosopyridine leads to a stable conjugate, installing a label at N6methyl-adenosine. We show that N6-methyl-adenosine-containing oligonucleotides can be enriched from complex mixtures, paving the way for applications to identify this modification in genomic DNA and RNA. Nucleic acids displays several types of C(sp 3 )-H bonds within its canonical nucleotides, each with subtly different intrinsic reactivities that are influenced by steric, inductive and conjugative effects imparted by the proximal chemical environment. Beyond the core genetic information stored the four-letter nucleotide sequences of A, C, G and T(U), nucleic acids contains a second layer of molecular programming in the form of reversible chemical modifications to its nucleobases -the, so called, epigenetic code -which introduces further diversity to the catalogue of C(sp 3 )-H bonds present in these biomacromolecules. Methylation at specific positions of the nucleobases provides the most common means by which genomic DNA is reversibly marked (1,2). Bacteria can methylate A and C in their own genome (3), while in eukaryotes, DNA methylation was thought to only occur at C (2), which has been linked to gene regulation. Importantly, the discovery of novel chemical methods for the selective modification of 5-methylcytosine (5mC) and its derivatives in DNA have been unequivocally responsible for a better understanding of its function (4). Recently, N6-methyl deoxyadenosine (N 6 mdA, Figure 1A) has been reported (somewhat controversially) in various eukaryote genomes (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) and comprehension of its biological roles remains nascent (5,10). In contrast, the presence of N6-methylation of adenine (m 6 A) in the RNA world is well-established and has been implicated in a wide range of cellular processes (17). Even though methyl groups within N-methylamines are not traditionally reactive, the exclusivity of this motif might underpin a site-selective chemical approach with which to covalently modify and manipulate N6-methyl-adenosine; there is currently no method to directly modify nucleic acids at N6-methyl-adenosine and a selective chemical transformation must target the C(sp 3 )-H bonds of the N-methylamine motif amidst thousands of similarly reactive entities. We were inspired by the cellular processing of methylation state in oligonucleotide sequences, wherein a class of dioxygenases, which include ALKBH1, have been proposed to demethylate N 6 mdA in mammalian DNA (9,11) through the action of electrophilic enzyme-bound Fe(IV)=O intermediates (Figure 1b) (18). By analogy with the corresponding demethylation in RNA (19), the C-H bond of the N6-methyl group would project towards the promiscuous Fe-bound oxygen-centred radical facilitating selective hydrogen atom abstraction (HAA), which is followed by an oxygen-rebound process with the resulting Fe(III)-OH species en route to formation of N 6 -hydroxymethyl-adenosine and spontaneous expulsion of formaldehyde. We questioned whether a synthetic process could generate a discrete a-amino radical intermediate on the N6-methyl group of N 6 mdA, mimicking the enzymatic demethylation pathway, but also intercept the incipient radical with a modular reagent, installing a covalently-bound label directly onto N 6 mdA-containing DNA sequences (Figure 1c). Proposed mechanism for biochemical demethylation of N 6 mA via Fe-dependent enzyme-controlled hydrogen atom abstraction and oxygen-rebound. c. Plan for covalent modification at N 6 mdA via trapping of an 'on-DNA' a-amino radical intermediate. The C-H bonds in N 6 mdA's methyl group have fairly high bond dissociation energies (BDE, ~92-94 kcal/mol) (20). A selective reagent will need to target a strong C-H bond that is present at extremely low effective concentration (reported N 6 mdA levels in eukaryotes have been reported from 5ppm to 0.05% for N 6 mdA with respect to A) (9,11,12), amongst a plethora of similar strength or weaker C-H bonds: for example, deoxyribose units contain many different C-H bonds, each with similar BDE's (21); and the methyl group in thymidine (and 5mC) displays activated C-H bonds with lower BDE (89-90 kcal/mol) (22). Moreover, such a N 6 mdA functionalization strategy also necessitates a method to productively intercept the 'on-DNA' a-amino-radical to fashion a stable covalent linkage to the oligonucleotide. The challenges associated with addressing these problems are multifaceted: firstly, use of a proximity-driven rebound mechanism thought to facilitate enzymatic demethylation is unlikely to be feasible in a synthetic scenario and so the coupling step will need to be fast in order to accommodate the likely short lifetime of the DNA-derived a-amino-radical; and secondly, the HAA and covalent functionalization steps need to operate in concert while minimizing deleterious and non-selective reactivity. These difficulties place additional constraints on potential chemical solutions that must already operate at low concentrations, in aqueous solutions and avoiding acidic or oxidative conditions, which might damage the nucleic acid architecture. The C-H bonds in the methyl group of N 6 mdA will be partially polarized as a result of their interaction with the lone pair on the N6-atom and display 'hydridic' character (23,24) (Figure 2A). In contrast, the C-H bonds of in the methyl group of thymine, although weaker, are relatively neutral as a result of being adjacent to the less polarizing pyrimidine heterocycle. This subtle electronic effect may provide a sufficiently distinct reactivity profile to enable a kinetically-controlled polarity match between an electrophilic hydrogen atom abstracting agent and the more hydridic C-H bonds in the N6-methylamine motif in N 6 mdA. The resulting C-H bond cleavage would lead to formation of an a-amino radical on the modified nucleotide. The intercepting reagent must react quickly with the incipient N 6 mdA-derived a-amino radical and form an open shell species more stable than its precursor. We reasoned that deployment of a spin trapping reagent could provide a potential solution to this challenge. Spin trapping reagents (STRs) are highly reactive molecules, frequently used in excess quantities to capture radicals in the form of persistent radical products, which can enable the identification of short-lived species in complex systems (25). Among commonly used STRs, nitrosoarenes are particularly suitable for the interception of nucleophilic carbon-centred radicals, the properties of which should be inherent to a N 6 mdAderived a-amino radical. Nitrosoarene-derived STRs are, however, highly electrophilic and often display promiscuous non-radical reactivity with nucleophiles, can undergo facile dimerization and readily decompose to non-productive products. To compound these problems, a nitrosoarene must also be compatible with the HAA step, itself a radical reaction, without displaying deleterious reactivity. We questioned whether these problems might be circumvented if a process could be designed wherein the STR was generated in situ, consequentially linked to the chemistry required to facilitate the HAA step, thereby closely linking the proximity of this reactive species to the incipient N 6 mdA-derived radical. Accordingly, we hypothesized that a mild method to reduce a water-soluble nitroarene might be leveraged alongside the HAA step for the in situ generation of a nitrosoarene STR and lead to a selective functionalization process (Figure 2B). Our design plan for a C(sp 3 )-H functionalization of N 6 mdA focused on the visible light-mediated reduction of 3-nitropyridine 1a, a water soluble oxidant that could serve as a precursor for the nitrosoarene STR (Figure 2C). We speculated that Ru(II)(phen)3Cl2 could function as a photocatalyst because it displays adequate aqueous solubility and the oxidative quenching cycle of its triplet excited-state (E[Ru(II)*/Ru(III)]= -0.87 V) (26), Our initial studies focussed on establishing a cross coupling protocol on a representative oligonucleotide 5 (CTTGACAG[N 6 mdA]CTAG, Figure 3A). Following an extensive assessment of the reaction parameters, the exploratory experiments revealed that irradiation of a solution of 5a, 3-nitropyridine 1a, quinuclidine 3 and [Ru(phen)3]Cl2 with a 60 W CFL bulb for just 10 minutes at room temperature, led to the formation of Nhydroxyformamidine-N 6 mdA oligonucleotide conjugate 6 with 16% conversion to product, as determined by LC-MS analysis (Figure 3a) (36). A 30% conversion to the de-methylated oligonucleotide 7 was also observed. It is important to stress that both conjugate 6 and demethylated oligonucleotide 7 arise from the same putative a-amino radical, int-III, which reflects a 46% conversion of the N6-methyl group via the new hydrogen atom abstraction process and is surprisingly high given the highly complex molecular framework upon which this transformation is affected. We believe that the constant oxidative quenching of triplet excited state of the metallophotocatalyst by 3-nitropyridine and high concentration of quinuclidine prevents oxidative damage of DNA, especially at G nucleobases. This is reflected in the observation that the transformation using the Ru(phen)3Cl2 as catalyst produces a cleaner reaction profile compared to the use of other, more oxidizing photocatalysts, [Ru(II)(bpz)3](PF6)2 (Figure S18, supplementary information). The oligonucleotide conjugate (6) has a half-life of approximately 12 hours at room temperature in neutral or basic solutions (pH = 7-11). We next sought to incorporate a latently reactive functionality capable of downstream elaboration to tailored nucleic acid fragments. A design-augmentation process revealed an alkyne-containing, amide-linked nitropyridine 1b could be coupled with 5a upon treatment with the Ru(phen)3Cl2, quinuclidine and irradiation for 10 minutes (Figure 3B), forming the desired alkyne-containing oligonucleotide 8 (11% conversion to product, identified by HRMS). It is notable that 1b is devoid of potentially competitive hydridic C-H bonds and the amide substituent does not seem to affect the oxidative reactivity of the 3-nitropyridine core. Exploiting the newly installed alkyne functionality, we found that a 'click' Huisgen-cycloaddition between 8 and PEG3 biotin-derived azide 9 necessitated specific conditions for an effective reaction; a solution of copper sulfate and sodium ascorbate required the addition of quinuclidine (presumably to act as a ligand for the copper-catalyst) to facilitate cycloaddition to the biotin-conjugated oligonucleotide 10 with 92% conversion to product. A series of control experiments showed that the photoredox coupling reaction on an oligonucleotide without a N 6 mdA residue (CTTGACAGACTAG, 7) formed no N-hydroxyformamidine-containing products arising from the incorporation of a 3-nitrosopyridine unit, indicating, as expected, that hydrogen atom abstraction does not take place in the oligonucleotide unless the N6-methylation is present. It is remarkable that the hydrogen atom abstraction step is so exquisitely selective for the N6-methyl group in spite of the vast number of similar C-H bonds in oligonucleotides (termed 'HAA selectivity', Figure 3C). We were, however, able to detect trace levels of oligonucleotides that had a mass ion reflecting the inclusion of an intact 3-nitropyridine (16 mass units higher than N-hydroxyformamidine-derived oligonucleotide 8). Although we were not able to elucidate the structure of this trace-level modification, a series of control experiments revealed that the addition of 3-nitropyridine was taking place at G residues (Figure S20-21, supplementary information). We were able to calculate that selectivity for the formation of the desired N 6 mdA-derived N-hydroxyformamidine linkage compared to the inclusion of 3-nitropyridine at G was 50:1 for N 6 mdA per G nucleobase (termed 'Probe selectivity', Figure 3C), a ratio which is, again, quite remarkable given the proclivity of G nucleobases to undergo oxidative side reactions. Figure 3D shows a preliminary scope of the N 6 mdA functionalization tactic, with the conversions to conjugates in line with those observed in the optimization studies. Not only did the reaction work on oligonucleotides in combination with nitropyridines 1a and 1b but it also converted longer and self-complimentary DNA sequences (42 & 49mers) to the desired products. Importantly, a reaction with an RNA oligonucleotide was successful converted to the corresponding conjugate although the yields were lower than for the DNA congeners; partial strand decomposition was observed in the RNA oligonucleotide, which will require further optimization of the reaction conditions. Nevertheless, the success of this methodology on RNA oligonucleotides has many potential applications in the emerging field of m 6 A-focussed epitranscriptomics (17). The versatile biochemical properties inherent to the biotin motif provide a means to isolate the modified N 6 mdAderived oligonucleotide from other nucleic acid fragments via a streptavidin-based pull-down procedure (37), which could enable us to enrich N 6 mdA-containing oligonucleotides in complex mixtures. Classical methods for substrate retrieval from streptavidin pull-down protocols involve relatively harsh reaction conditions, which are designed to denature the protein scaffold. However, our photoredox conjugation procedure installs the Nhydroxyformamidine linkage, a more labile functional group, which we believed would permit the use of significantly milder, nucleophile-mediated, cleavage conditions in the retrieval of the labelled oligonucleotide. This is important because mass analysis of the photoredox reaction mixtures had suggested that the traces products arising from unselective functionalization at G do not contain an electrophilic N-hydroxyformamidine linkage. Consequently, we speculated that these off-target products of functionalization at G could be retained on the streptavidin beads during cleavage, thereby enhancing the selectivity observed in the photoredox step and enrichment of the N 6 mdA-derived oligonucleotide. Guided by this hypothesis, we began the pull-down and enrichment procedure by conducting the photo-conjugation with oligonucleotide 5a and nitropyridine 1b in the presence of a distinct, but importantly non-methylated oligonucleotide CGTACTAGACG 11, as a means to test whether our method could be used to enrich N 6 mdA-containing oligonucleotides (Figure 5a). The Nhydroxyformamidine product 8 was formed with 10% conversion and observed by LC-MS alongside unreacted 5a, the demethylated oligonucleotide 7, the control oligonucleotide 11 and traces of the G-nitropyridine functionalized oligonucleotide (50:1 probe selectivity, N 6 mdA/G). Subsequent cycloaddition with biotin-azide 9 afforded the selective N 6 mdA biotin-conjugated oligonucleotide 10. Treatment of the oligonucleotide mixture with streptavidin-coated magnetic beads allowed immobilisation of all the species containing biotin (specifically, 12 plus the trace levels of product arising from unselective reaction at G residues) and permitted the removal of unlabelled oligonucleotides (7 and 11) via successive washing procedures. Following this, we found that the electrophilic nature of the N-hydroxyformamide linkage made it susceptible to reaction with aqueous hydrazine and led to the release of an N 6 -(hydrazonomethyl)dA-containing oligonucleotide 13 with a small amount of 7 (arising from the hydrolysis of 13) and trace quantities of the other oligonucleotides that had been indiscriminately retained by the streptavidin-coated beads. The recovery of both 13 and 7 provides direct evidence for the presence of N 6 mdA in the starting sequence and their ratio to all other oligonucleotides gives rise to an enrichment greater than 50:1. It is particularly important to note that the maximum theoretical enrichment value that can be obtained as a result of the observed photoredox probe selectivity is ~17:1, since oligonucleotide 13 contains 3 G residues (probe selectivity of 50:1 N 6 mdA per G residue). Therefore, the observed enrichment of >50:1 clearly demonstrates that the hydrazine cleavage procedure is selective for Nhydroxyformamidine linkage in the N 6 mdA-derived oligonucleotide conjugates versus products of reaction at G (that are presumably retained on the beads), leading to the observed enhanced enrichment. To simulate the complex matrix of a cellular DNA sample, where the concentration of N 6 mdA with respect to dA will be very low, we combined longer single-stranded (ss) DNA fragments (99 nucleotides) with a 10-fold excess of salmon sperm DNA, to create complex DNA mixtures with N 6 mdA/dA ratios of 1:383 (0.26%). Applying the photo-conjugation and pull-down procedure to these samples and now using quantitative PCR (qPCR) to analyse the enriched fractions and determine the amplifiable amount of both initially methylated and unmethylated DNA sequences after the pull-down, we found that the N 6 mdA-containing ssDNA fragment (yellow strand, Figure 5b, experiment i) was enriched to a level of approximately 10:1. Importantly, a parallel experiment (ii) using the same nucleic acid sequences but having the N 6 mdA residue in the other sequence (grey) showed a similar level of enrichment (11:1). These results clearly demonstrate that N 6 mdA is required for the enrichment, the observed enrichment is not dependent on the oligonucleotide sequence and that the protocol is functional in complex DNA mixtures. In double stranded DNA, the methyl group of N 6 mdA is thought to project into the major groove of the double helix, providing an additional challenge for the photoredox functionalization of complex nucleic acid samples due to potentially adverse steric and electronic effects that arise from the local chemical environment. Despite this congestion, we found that on applying the sequential photoredox conjugation, click reaction and pull-down procedure to a mixture of N 6 mdA-containing 99 base pair dsDNA (blue double strand, iii) and a non-methylated double stranded fragment (green double strand) resulted in an enrichment of 4:1. A parallel experiment (iv) using the same nucleic acid sequences but having two N 6 mdA residues in the second template showed that the N 6 mdAcontaining dsDNA was recovered with an enrichment of greater than 9:1, indicating a positive cumulative effect. Furthermore, when these dsDNA fragments were combined with salmon sperm DNA (N 6 mdA/dA ratio is 1:3433, 0.03%), the N 6 mdA-containing oligonucleotide was again recovered with an enrichment of almost 4:1 (experiment v). The corresponding experiment (vi) using the same nucleic acid sequences but with two N 6 mdA residues in the second template (N 6 mdA/dA ratio is 1:1717, 0.06%) showed an increased enrichment of greater than 8:1 for the N 6 mdA-derived ds-oligonucleotide, again highlighting that the presence of additional N 6 mdA residues enhances the output. Taken together, this set of experiments demonstrates the applicability of the developed chemistry on longer ssDNA, dsDNA, as well as in complex samples with excess DNA, providing a proof of concept for the enrichment of N 6 mdA-DNA strands and showcasing the potential for future application with this underexplored methylated nucleotide. While the cell's biochemical machinery is capable of regulating the methylation state of A in nucleic acids, we have developed a selective chemical transformation that generates and intercepts an 'on-DNA' a-amino radical to form a stable covalent modification at N 6 mdA residues. Orchestrated by a visible light-activated photoredox catalyst, a polarity-matched hydrogen atom abstraction step at the N6-methyl group of N 6 mdA generates an aamino radical and dovetails with a distinct, in situ, reaction to form a nitrosopyridine spin trapping reagent, which together lead to radical cross coupling process that introduces a modular functional handle into oligonucleotide sequences. This strategy is underpinned by a previously unknown transformation founded on a mechanistically unique and remarkably selective photoredox cross coupling reaction, which targets a traditionally unreactive and scarce motif amidst the complex scaffold of nucleic acids. To set this work in context, the evolution of this synthetic method towards a basic technology upon which a chemical method for locating N 6 mdA in genomic DNA can be founded, will require a number of further challenges to be addressed, despite the remarkable selectivity's observed in this functionalization process. Firstly, while the photocatalytic hydrogen atom abstraction process proceeds with competent efficiency (46% conversion-combination of conjugation and demethylation-with respect to the starting oligonucleotide), the spin trapping of the N 6 mdA-dervied oligonucleotide radical requires improvement in order to increase the yield of the desired conjugation product. Secondly, the already high (50:1) probe selectivity for reaction at N 6 mdA per G residue could to be further increased because the effective discrimination of the process becomes diminished as the length of the oligonucleotide increases (due to the increased concentration of G residues in complex DNA). Finally, further investigation of the unexplored reactivity of N-hydroxyformamidines should lead to improved chemoselectivity in the release step of the pull-down procedure and higher final enrichment. Together, the next phase of these studies will focus on the development of a detection protocol for N 6 mdA-containing oligonucleotides that already returns an enrichment of 4:1 for a single N6-methylation in a 99 base pair double stranded fragment that is part of a complex DNA matrix. Not only might this technology provide an unequivocal answer to the controversy surrounding the presence of N 6 mdA in genomic mammalian DNA but could also lead to sequencing methods that will further unravel the role of this epigenetic modification and should also be amenable to targeting methylated nucleobases in the many forms of RNA that regulate cellular function (17,38). 33) Another plausible pathway for the formation of the nitroso derivative is via radical-radical coupling of the QRC and the persistent nitroarene radical anion to furnish the corresponding intermediate, which spontaneously fragments to deliver the desired nitroso compound. For a similar mechanism, see 34 34) White, N. A. & Rovis, T. Enantioselective N-heterocyclic carbene catalyzed β-hydroxylation of enals using nitroarenes: an atom transfer reaction that proceeds via single electron transfer. J. Am. Chem. Soc. Tables S1-S2 ## General Remarks General procedures ## Supplementary Text and Figures Important remarks on General Remarks Proton nuclear magnetic resonance (1H NMR) spectra were recorded at ambient temperature on a 400 MHz Bruker Avance III HD spectrometer (400 MHz) or a 500 MHz Bruker Avance III HD Smart Probe spectrometer (500 MHz). Chemical shifts (δ) were reported in ppm and quoted to the nearest 0.01 ppm relative to the residual protons in CDCl3 (7.26 ppm), DMSO-d6 (2.50 ppm), methanol-d4 (3.31 ppm) and coupling constants (J) were quoted in Hertz (Hz). Coupling constants were quoted to the nearest 0.1 Hz and multiplicity reported according to the following convention: s = singlet, d = doublet, t = triplet, q = quartet, qnt = quintet, sxt = sextet, spt = septet, oct = octet, m = multiplet, br = broad and associated combinations, e.g. dd = doublet of doublets. Where coincident coupling constants have been observed, the apparent (app) multiplicity of the proton resonance has been reported. Data were reported as follows: chemical shift (multiplicity, coupling constants, number of protons and molecular assignment). LCMS spectra were recorded on an Amazon X ESI-MS (Bruker) connected to an Ultimate 3000 LC (Dionex). Oligodeoxyribonucleotidess were analysed using a gradient of 5-30% or 5-40% methanol vs. an aqueous solution of 10 mM triethylamine and 100 mM hexafluoro-2-propanol on a XTerra MS C18 column (125, 2.5 µm, 2.1x50mm) with TMS endcapping or an XBridge Oligonucleotide BEH C18 column (130, 2.5 µm, 2.1x50mm). Small molecules were analysed with a gradient of 0-100 % acetonitrile with 0.1% formic acid vs. water with 0.1% formic acid on a Kinetex® C18 column (100 , 2.6 µm, 50 x 2.1 mm). Mass chromatograms shown are base peak chromatograms, UV absorption was recorded at 260 nm. High-resolution mass spectra (HRMS) of small molecules were conducted using Shimadzu LC-MS 9030 QToF. Oligodeoxyribonucleotidess were analysed using a gradient of 5-30% methanol vs. an aqueous solution of 10 mM triethylamine and 100 mM hexafluoro-2-propanol on a XTerra MS C18 column (125, 2.5 µm, 2.1x50mm) with TMS endcapping. Small molecules were analysed with a gradient of 0-100 % acetonitrile with 0.1% formic acid vs. water with 0.1% formic acid on a Kinetex® C18 column (100 , 2.6 µm, 50 x 2.1 mm). Analytical thin layer chromatography (TLC) was performed using pre-coated Merck glass-backed silica gel plates (Silicagel 60 F254 0.2 mm). Visualization was achieved using ultraviolet light (254 nm) and chemical staining with basic potassium permanganate solution as appropriate. Flash column chromatography was undertaken on Fluka or Material Harvest silica gel (230-400 mesh) under a positive pressure of air or on a CombiFlash Rf 200 (Teledyne Isco) system using 50 μm Si-HP PuriFlash columns. Primer elongation reactions were performed on a T100 Thermocycler (BioRad) or on a 96 Universal Gradient (PeqStar) machine. Automated gel electrophoresis was performed using an Agilent Technologies 2200 Tapestation and D1000 ScreenTapes and sample buffer. qPCR was performed using a CFX96 Real-TimeSystem (BioRad), and data was processed using CFX software manager 3.1(BioRad). qPCR reactions (volume: 10 μL) contained DNA calibration or sample mixtures (1 μL), the corresponding forward and reverse primers (1 μM each), and Brilliant III ultra-Fast SYBR green qPCR mastermix (Agilent Technologies, 5 μL). Reactions were run according to the manufacturer's protocol. Calibration curves were made to determine the amounts of target DNA in the analysed samples. Oligodeoxyribonucleotides (ODNs), including short ONDs for reactions, 99nt ssDNA strands and template and primers for the synthesis of 99nt dsDNA strands were custom synthesised and HPLCpurified by ATDBio or Sigma-Aldrich and used without further purification after dissolution into milliQ H2O. Reagents were obtained from Sigma-Aldrich, Acros, Alfa Aesar, TCI, or Jena Bioscience and used without further purification. Enzyme solutions were obtained from Zymo, New England BioLabs, and Sigma and used directly. Dichloromethane, ethyl acetate, tetrahydrofuran, toluene, and petroleum ether (40-60) were dried and distilled using standard methods. Water was purified on a milliQ system. Other solvents used were purchased anhydrous and used without further purification unless otherwise stated. Reactions were carried out under nitrogen atmosphere unless otherwise stated. Reactions were monitored by LCMS. ## General procedures General procedure A: functionalization of N General procedure E: functionalization of N 6 mdA in 99nt ssDNA with the 3-nitropyridine probe for enrichment A 2 mL microwave vial was charged with a mixture of two ODN substrates (methylated and unmethylated, 2 μM in milliQ H2O, 12.5 µL each). In a separate Eppendorf tube, quinuclidine (8.25 mg, 75 µmol) and the 3-nitropyridine probe (0.3 mg, 1.3 µmol) were dissolved using a stock solution of [Ru(phen)3]Cl2 (3 mM in 1 mL of 30% MeCN in milliQ H2O, 12.5 µL). The latter mixture was then added to the ODN mixture. For the reactions with added salmon sperm DNA, 2 ul of a salmon sperm ssDNA solution (5 mg/ml, abcam ab229278) were added before the photoredox reaction. Final concentrations: ODNs: 0.67 μM each, quinuclidine: 2 M, 3nitropyridine probe: 33 mM, [Ru(phen)3]Cl2: 1 mM, 10 % MeCN in H2O. The microwave vial was sealed under nitrogen atmosphere after flushing for 15-20 seconds. The vial was then placed inside a 55W CFL bulb. The mixture was irradiated for 5 min while the temperature was maintained with a fan and subsequently filtered through a prewashed Micro BioSpin 6 column (BioRad) after the reaction. ## General procedure F: click reaction on the functionalized N 6 mdA in 99nt ssDNA To the resulting ODN mixture from general procedure E in a 1.5 ml Eppendorf tube was added quinuclidine (8.25 mg, 75 µmole in 10 µL H2O), Azide-PEG3-biotin conjugate (20 mM, 7.5 µL), sodium ascorbate (40 mM, 7.5 µL) and CuSO4 (2 mM, 7.5 µL). Final concentrations: quinuclidine: 1 M, Azide-PEG3-biotin conjugate: 2 mM, sodium ascorbate: 4 mM, CuSO4: 0.2 mM. After 30 minutes reaction at r.t., the mixture was filtered twice through a prewashed Micro BioSpin 6 column (BioRad). General procedure G: functionalization of N 6 mdA in 99bp dsDNA with the 3-nitropyridine probe for enrichment A 2 mL microwave vial was charged with a mixture of two dsODN substrates (methylated and unmethylated, 0.25-0.35 μM in milliQ H2O, 8.3 µL each). In a separate Eppendorf tube, quinuclidine (5.5 mg, 50 µmol) and the 3-nitropyridine probe (0.2 mg) were dissolved using a stock solution of [Ru(phen)3]Cl2 (3 mM in 1 mL of 30% MeCN in milliQ H2O, 8.3 µL). The latter mixture was then added to the ODN mixture. For the reactions with added salmon sperm DNA, 2 ul of a salmon sperm ssDNA solution (5 mg/ml, abcam ab229278) were added before the photoredox reaction Final concentrations: ODNs: 0.08-0.12 μM each, quinuclidine: 2 M, 3nitropyridine probe: 33 mM, [Ru(phen)3]Cl2: 1 mM, 10 % MeCN in H2O. The microwave vial was sealed under nitrogen atmosphere after flushing for 15-20 seconds. The vial was then placed inside a 55W CFL bulb. The mixture was irradiated for 5 min while the temperature was maintained with a fan and subsequently filtered through a prewashed Micro BioSpin 6 column (BioRad) after the reaction. ## General procedure H: click reaction on the functionalized N 6 mdA in 99bp dsDNA To the resulting ODN mixture from general procedure G in a 1.5 ml Eppendorf tube was added quinuclidine (5.5 mg, 50 µmol in 5 µL H2O ## Important remarks on the photocatalyst used While the design plan in Figure 2 of the main manuscript describes the use of Ru(phen)3Cl2 photocatalyst and its oxidative quenching cycle, many of the preliminary results of this project were conducted using similar photocatalyst, [Ru(bpz)3](PF6)2. Due to the difference in the reduction potentials, the operative photoredox cycle for the latter is more likely to be the reductive quenching cycle: oxidation of the quinuclidine 3 as first step and reduction of 3-nitropyridine 1a to restore the metal complex ground state (Fig. S1). It is important to note that while the photoredox cycles are different, the radical species formed are the same (int I and int III); indeed, both the photocatalysts showed very similar yields and reaction profile on short DNA fragments. However, the difference in the quenching cycle was crucial for the application of the reaction to single stranded (ss) and double stranded (ds) long oligonucleotides, as the Ru(phen)3Cl2 showed significantly cleaner reaction profiles due to lower oxidative power compare to [Ru(bpz)3](PF6)2 (Fig. S16). When substituting substrate ODN 5 by its demethylated analogue 7 and applying the same conditions, no reaction could be observed (Fig. S3). This set of experiments led to the conclusion that both observed products in the reaction on ODN 5 are the result of a selective reaction at the N 6 mdA residue. We posit that both are the product of an initial highly selective HAT from the N-methyl group to the well-described quinuclidinium radical cation formed in situ. ## Reaction with quinuclidine derivative 2-(quinuclidin-3-yl)propan-2-ol Further confirmation of the incorporation of quinuclidine on the N 6 mdA residue was shown by developing a quinuclidine derivative that is compatible with the reaction in Fig. S2: 2-(Quinuclidin-3-yl)propan-2-ol. Under similar conditions to the quinuclidine functionalization in Fig. S2, we observed the new product [N 6 mdA-Qder] along with product 7 (Fig. S4). This experiment confirms that the second observed product (as shown in Fig. S4) is a quinuclidine derivative. ## Reactions with [Ru(bpz)3](PF6)2 and reduced concentrations of quinuclidine To understand if excited [Ru(bpz)3](PF6)2 can damage ODNs in the absence of high amounts of quinuclidine, we run control reaction on ODN 5 (as depicted in the scheme of Fig. S2) with decreasing concentrations of quinuclidine 3 (Fig. S5). Considerable broadening of the LCMS peaks when using lower quinuclidine concentrations indicates that decreasing quinuclidine concentration is detrimental to the selectivity of the reaction. LCMS studies on the formation of the N-hydroxyformamidine product Encouraged by the validation of the HAT step, we next investigated the in situ generation of the nitrosoarene spin trapping reagent 2a from 3-nitropyridine 1a and its interception of the N 6 mdAderived a-amino radical. We found that irradiation of a solution of oligonucleotide 5, quinuclidine 3, [Ru(bpz)3](PF6)2 and 3-nitropyridine 1a for 10 minutes at room temperature produced Nhydroxyformamidine conjugate [N 6 mdA-P] 6 in 14% yield (Fig. S6), along with quinuclidine product [N 6 mdA-Q] (3% yield) and demethylation product [dA] 7 (26% yield). Given the fact that the observed mass of N-hydroxyformamidine conjugate [N 6 mdA-P] 6 (m/z = 1358) is very similar to the mass of quinuclidine product [N 6 mdA-Q] (m/z = 1359), we conducted a control experiment using 5-Phenyl-3-nitropyridine (Fig. S7) to confirm that the new product 6 is indeed a derivative of 3-nitropyridine 1a and not the oxidized version of quinuclidine product 6. When 5-phenyl-3-nitropyridine was used under similar reaction conditions showed in Fig. S5 Additional evidences are provided by the following experiment, where we used 3-nitropyridine 1a and 2-(Quinuclidin-3-yl)propan-2-ol instead of quinuclidine 3 (Fig. S8). In this case, we could detect the formation of product [N 6 mdA-P] 6 with m/z = 1358 (5% yield) and the 2-(Quinuclidin-3-yl)propan-2-ol product [N 6 mdA-Qder], but we could not observe the formation of the previously observed quinuclidine product [N 6 mdA-Q] as expected. This experiment clearly confirms that product [N 6 mdA-P] 6 is not the mere oxidized version of quinuclidine product. Selective functionalization of N 6 mdA ODN 5 with derived 3-nitropyridines Various 3-nitropyridine derivatives have been tested under the N 6 mdA selective reaction conditions described in Fig. S6. The aim of this study was to provide information regarding the solubility, reactivity and compatibility of modified 3-nitropyridine compounds toward the development of a probe that can contain an alkyne group for future enrichment studies. Table S1. Conditions and yields for the selective functionalization reactions of N 6 mdA ODN 5 with different 3nitropyridine derivatives. ## 3-nitropyridine der. [c] reaction time a We first started our exploration using 3-nitropyridine derivatives containing a phenyl group as substituent or linker for the alkyne moiety. The choice of the phenyl group was done based on the small changes in the reduction potential compare to unsubstituted 3-nitropyridine. Indeed, 3nitropyridine needs to be reduced to generate the desired spin trapping reagent 3-nitrosopyridine; therefore, a phenyl substituent would not compromise the key reduction step. Unfortunately, all the substrates containing an aromatic substituent were poorly soluble in the aqueous media and showed low yields for the selective functionalization of N 6 mdA. Later, we found that the amide linkage was well tolerated in position 4 and 5 of the 3-nitropyridine, furnishing the desired functionalisation in better yields. Finally, we could improve the initial result of 3-nitropyridine 1b (8% yield) using a more powerful source of light, such as 60W CFL (17% yield). ## Comparison of photoredox reaction outcome with [Ru(bpz)3](PF6) and [Ru(phen)3]Cl2 Based on the proposed reaction mechanism described in Fig. 2 of the main text, excited [Ru(bpz)3](PF6)2 undergoes reductive quenching through the oxidation of quinuclidine 3. Instead, excited [Ru(phen)3]Cl2 directly reduces the 3-nitropyridine through the oxidative quenching cycle. Even though both photocatalysts are able to promote the desired chemistry, which quenching cycle is operative under the reaction conditions could be important for the stability of the starting oligonucleotide and the selectivity of the process. Below, we compared the UV traces of reactions using either [Ru(bpz)3](PF6)2 or [Ru(phen)3]Cl2 in otherwise identical conditions (Fig. S16). HRMS studies on Ru(phen)3Cl2 system and photoredox selectivity Whereas the LCMS traces of the photoredox reaction with [Ru(bpz)3](PF6)2 indicate the occurrence of trace side reactions, we thoroughly analysed the products of the reaction with [Ru(phen)3]Cl2 by HRMS to confirm identity of the compounds and understand whether any other 'on-DNA' N-hydroxyformamidine product was formed (Fig. S17). Fig. S17. All the products have been characterized by HMRS using a Shimadzu LC-MS (Q-Tof) 9030. Control experiment using the corresponding oligonucleotide without N 6 mdA did not show any N-hydroxyformamidine adducts, searching for the calculated exact masses within a range of 100 ppm. These results indicate >100:1 selectivity (detection limit: 0.00007%) for the HAT process and addition of the α-amino radical to the nitroso intermediate. Identity of the demethylation product 7, the quinuclidine derivative [N 6 mdA-Q], and the Nhydroxyformamidine product 9 were confirmed by HRMS. We further observed metal complexes of the N-hydroxyformamidine products (with Ni(II) and Zn(II)) as confirmed by the observed high-resolution mass and isotope pattern (Fig. S17). Furthermore, in the reaction mixture from a control reaction with unmethylated ODN 7 as substrate under identical conditions, we could not detect any N-hydroxyformamidine product (with a detection limit of 0.00007%), indicating a >100:1 selectivity of the HAT process and the formation of the N-hydroxyformamidine on N 6 mdA. Moreover, HMRS analysis of the photoredox crude reaction mixture showed a peak with the exact mass of the nitroso intermediate 2b (Pred. C11H11N3O2+H = 218.0924, found 218.0926), supporting the proposed mechanism. ## HRMS studies on unselective incorporation of intact 3-nitropyridine 1b at dG residues To further understand the selectivity of the developed N 6 mdA functionalisation process, we thoroughly analysed the reaction mixture of the photoredox conjugation (according to the scheme in Fig. S17) by HRMS. The only apparent trace products arising from a non-selective reaction were identified as incorporation of the unmodified 3-nitropyridine probe 1b into the ODN substrate (Fig. S18). No trace of any other side reaction was detected. Importantly, HRMS-based comparison of the trace impurities with the N 6 mdA-specific Nhydroxyformamidine formation shows that the overall selectivity for the incorporation of 1b onto N 6 mdA is 98% (50:1). To understand if the reaction was specific to one of the canonical deoxynucleosides, we applied the conditions on a set of four unmethylated 13nt ODNs with a 4:4:4:1 distribution of the four canonical nucleosides. Whereas for the ODNS containing only one dA, dC, or dT residue, several products were observed (as for ODN 5 in Fig S18 ), only two new peaks were found for the ODN containing only one dG residue, indicating that the unselective incoporation of 1b is specific to dG (Fig. S19). Overall, these studies show that the selectivity of N 6 mdA functionalisation with the 3-nitropyridine 1b versus any other nucleoside is 98% (50:1) and indicate that only dG residues might undergo trace functionalisation, although via a different pathway not resulting in a hydrolysable Nhydroxyformamidine linkage. ## Collection of evidences for the characterisation of N-hydroxyformamidine products Evidence for the selective functionalisation of the N6mdA residue In the proposed mechanism, we envision the abstraction of an H atom from the N 6 methyl group of N 6 mdA to obtain the corresponding α-amino radical species (Fig. 2, main text). To confirm the selective HAT and therefore the position of the new label in the observed products, we conducted several control experiments where oligonucleotide 7 (containing the same sequence of oligonucleotide 5 but having dA instead of N 6 mdA) was subjected to our photoredox conditions with various 3-nitropyridine alkyne-derivatives 1 (Fig. S20). In all cases, LCMS of the reactions showed recovery of pure unreacted oligonucleotide 7, with no products apparent (such as the N-hydroxyformamidine products when the methylated substrate ODN 5 was used, see Fig. S13, S14 and S15). This confirms that no visible reaction takes place on ODNs without N 6 mdA. These results not only indicate an exquisite selectivity for the functionalisation of N 6 mdA, but also unambiguously confirm the position of the new label on the N 6 methyl group (HAT selectivity > 100:1). ## Evidence for the N-hydroxyformamidine structural linkage The N-formamidine group follows a distinct hydrolysis process which provides two sets of hydrolysed products containing N-formyl derivatives (J. Org. Chem. 1999, 64, 991−997). Given this unique feature, we conducted hydrolysis studies on the labelled ODN 8 to confirm the identity of the N-hydroxyformamidine structural linkage (Fig. S21). As expected, treatment of the reaction mixture acquired from the photoredox conjugation (containing product 8) with acetate buffer (pH = 4.7) for 60 minutes fully consumed ODN 8 under formation of three new products: N-formyl ODN 14, 3-N-hydroxypyridine derivative 15 and 3-N-formyl-N-hydroxypyridine derivative 16. The three products 14, 15, and 16 were characterized by HRMS (Fig. S22 and S23). Importantly, a control experiment where oligonucleotide 7 was subjected to the same procedure (photoredox and hydrolysis) did not result in formation of any of the hydrolysis products according to scans of calculated exact masses with error ranges of up to 100 ppm. Together, these results are consistent with the presence of a N-hydroxyformamidine linkage in the photoredox N 6 mdA product. Additional evidence for the N-hydroxyformamidine linkage are provided by reactivity studies using hydrazine as nucleophile instead of water (Fig. S24). Treatment of immobilised DNA fragment 12 with 10% hydrazine aqueous solution for 5 minutes delivered N-NH2 formamidine ODN 13 along with small amount of 7 (due to the hydrolysis of 13) as detected by LCMS at the end of the pull-down procedure described in Figure 4 of the main text (see also Fig. S30). The formation of the hydrazine product 13 upon treatment of the immobilised product 12 is a further clear indication of the identity of the Nhydroxyformamidine moiety. Another indication of the presence of the N-hydroxyformamidine moiety is the detection of Ni(II) and Zn(II) adducts of product 8 and the corresponding click product 10 as confirmed by HRMS and isotopic pattern studies (Fig S25, see also Fig. S17). Having demonstrated that several 3-nitropyridine alkyne derivatives are compatible with our N 6 mdA selective photoredox functionalisation, we next evaluated and optimised the reaction conditions for the 'click' Huisgen cycloaddition of azide-PEG3-biotin 9 into various alkynefunctionalized N 6 mdA oligonucleotides. The selective incorporation of a biotin moiety into N 6 mdA residues is fundamental for the next enrichment studies, which rely on the immobilisation of the N 6 mdA DNA fragments into streptavidin-coated magnetic beads. Table S2. Conditions and yields for the Huisgen cycloaddition of alkyne-functionalized N 6 mdA ODN with azide-PEG3-biotin 10. THTPA: Tris(benzyltriazolylmethyl)amine, TCEP: tris(2-carboxyethyl)phosphine, Na-asc.: sodium ascorbate. ## 3-nitropyridine der. [c] photoredox conditions a The results showed that the alkyne substrate, copper ligand, reductant and time are key parameters for the successful outcome of the reaction. To our surprise, quinuclidine proved to be fundamental as copper ligand compare to well-established water-soluble ligands such as Tris((1-benzyl-4triazolyl)methyl)amine (TBTA) and 3-[tris(3-hydroxypropyltriazolylmethyl)amine] (THPTA). Sodium ascorbate (Na-asc) was superior to tris(2-carboxyethyl)phosphine (TCEP) as reducing agent and 30 minutes was found to be the best time in term of yield. ## LCMS monitoring of 13nt ssDNA enrichment experiments In order to demonstrate the high selectivity of the N 6 mdA functionalisation reaction, we devised an enrichment experiment, consisting of immobilising N 6 mdA-containing ssDNA, washing off unmethylated strands, then cleavage and retrieval of the initially methylated sequence, as depicted in Fig. S29. Importantly, selective nucleophile-mediated cleavage of the N-hydroxyformamidine using an aqueous hydrazine solution was expected to afford high enrichment factors. The enrichment experiment was monitored by LCMS (Fig. S30) ## Enrichment studies of 99nt ssDNA To test if the developed N 6 mdA-specific DNA enrichment protocol can also be applied on longer ssDNA strands, and if enriched DNA can be amplified, we applied it on a set of two 99nt ssDNA strands and quantified both sequences by qPCR after enrichment (Fig. S31). We also added excess salmon sperm DNA (SS-DNA) in an additional set of experiments to show that the presence of additional DNA does not interfere with the N 6 mdA-selective chemistry. Enrichment factors of 6:1 and 10:1 in the presence or absence of SS-DNA demonstrate that the developed chemistry also allows for an N6mdA-specific enrichment of 99nt DNA strands. Furthermore, excess background DNA does not interfere with the process. Enrichment studies of 99bp dsDNA ## 2-step biotinylation To test application of the developed N 6 mdA-specific enrichment protocol on dsDNA, we applied it on a set of two 99bp dsDNA strands and quantified the outcome by qPCR (Fig. S32). Again, we added excess SS-DNA in an additional set of experiment to show that the presence of additional DNA does not interfere with the N 6 mdA-selective chemistry. Where SS-DNA was added, the calculated N 6 mdA/A ratio is 1:3433 (0.03%) for dsDNA fragment (blue) containing one N 6 mdA and 1:1716.5 (0.06%) for dsDNA fragment (orange) containing two N 6 mdA. Enrichment factors of 4:1 (for 99bp dsDNA containing one N 6 mdA residue) and of 8:1 resp. 9:1(for 99bp dsDNA containing two N 6 mdA residues) confirmed applicability of the developed protocol on dsDNA, as well as a positive cumulative effect in the enrichment factors for higher N 6 mdA densities. ## 5-nitro-2-phenylpyridine Synthesised according to General Procedure J from 2-bromo-5-nitropyridine (203 mg, 1 mmol) and phenylboronic acid (183 mg, 1.5 mmol) to afford product as a white solid (112 mg, 0.56 mmol, 56%). ## 3-nitro-4-phenylpyridine Synthesised according to General Procedure J from 4-chloro-3-nitropyridine (1.00 g, 6.3 mmol) and phenylboronic acid (0.90 g, 7.4 mmol) to afford product as a pale yellow solid (0.65 g, 3.3 mmol, 52%). ## N-(3-(3-nitropyridin-4-yl)phenyl)hex-5-ynamide Synthesised according to General Procedure J from 4-chloro-3-nitropyridine (150 mg, 0.95 mmol) and (3-(hex-5-ynamido)phenyl)boronic acid (200 mg, 0.87 mmol) to afford the product as an offwhite solid (187 mg, 0.61 mmol, 70%). ## 3-(3-nitropyridin-4-yl)-N-(pent-4-yn-1-yl)benzamide Synthesised according to General Procedure K from 3-(3-nitropyridin-4-yl)benzoic acid (18 mg, 0.074 mmol) and 4-pentyne-1-amine hydrochloride (11 mg, 0.089 mmol) to afford the product as an off-white solid (17 mg, 0.055 mmol, 74 %). ## 5-nitro-N-(pent-4-yn-1-yl)picolinamide Synthesised according to General Procedure K from 5-nitropyridine-2-carboxylic acid (141 mg, 0.84 mmol) and 4-pentyne-1-amine hydrochloride (100 mg, 0.84 mmol) to afford the product as an off-white solid (166 mg, 0.71 mmol, 85 %). ## N-(5-nitropyridin-3-yl)hex-5-ynamide, 1b Synthesised according to General Procedure K from 3-amino-5-nitropyridine (139 mg, 1.00 mmol) and 5-hexynoic acid (0.11 ml, 1.00 mmol) to afford the product as a pale orange solid (96 mg, 0.41 mmol, 41 %). ## 5-nitro-N-(prop-2-yn-1-yl)picolinamide Synthesised according to General Procedure K from 5-nitropyridine-2-carboxylic acid (141 mg, 0.84 mmol) and propargylamine (54 µl, 0.84 mmol) to afford the product as a pale yellow solid (124 mg, 0.60 mmol, 72 %).

chemsum

{"title": "Selective chemical functionalization at N 6 mdA residues in DNA", "journal": "ChemRxiv"}

a_retrospective_cross-sectional_study_to_determine_chirality_status_of_registered_medicines_in_tanza

## Abstract: Medicines with a stereogenic center (asymmetric carbon) are mainly present as racemates with a mixture of equal amounts of enantiomers. One enantiomer may be active while the other inactive, alternatively one may produce side-effects and even toxicity. However, there is lack of information on the chirality status (either racemates, single active enantiomer or achiral) of medicines circulated on the market particularly in African countries. We established the chirality status of registered medicines in Tanzania by conducting a retrospective cross-sectional study. Registration data for the past 15 years from 2003 to 2018 were extracted from TMDA-IMIS database to Microsoft excel for review and analysis. A total of 3,573 human medicines had valid registration. Out of which 2,150 (60%) were chiral and 1,423 (40%) achiral. Out of the chiral medicines, 1,591 (74%) and 559 (26%) were racemates and single active enantiomers, respectively. The proportion of racemates within chiral medicines was considerably higher than single enantiomer medicines. The use of racemates may cause harm to the public and may contribute to antimicrobial resistance due to potential existence of inactive and toxic enantiomers. In order to protect public health, regulatory bodies need to strengthen control of chiral medicines by conducting analysis of enantiomeric impurity.Chirality is very important in the pharmaceutical field [1][2][3] . Chirality (sometimes called stereoisomerism, enantiomerism or dissymmetry) is a property of an object which renders it non-superimposable with its mirror image [1][2][3][4][5][6][7] . Chiral medicines are those medicines with a stereogenic center (often called an asymmetric carbon) and exhibit chiral properties [1][2][3][4] . Most pharmacological processes present a high degree of stereoselectivity which results in differences between the activities of the enantiomeric forms of chiral medicines 1,4,8,9 . It is well known that a racemic mixture consists of an equimolar mixture of two enantiomers of the same chemical structure 8 . Enantiomers in some chiral medicines may exhibit marked differences in biological activities such as pharmacology, toxicology and pharmacokinetics 1,8,10,11 . One enantiomer may be active while the other inactive, and may produce side-effects and/or exhibit toxicity 6,8,12,13 . The use of racemic mixtures may present problems, such as adverse effects or toxicities particularly if they are associated with a less active, or inactive isomer 8,14,15 .There are many examples of chiral medicines whose enantiomers vary drastically in their properties. A well-known example of enantiomer related toxicity is the (R)-and (S)-enantiomers of thalidomide 1,2,4,8 . The (R)enantiomer of thalidomide is an effective sedative agent while, the (S)-enantiomer is known to cause teratogenic birth defects 1,4,8 . These defects included phocomelia, a severe shortening or lack of limb structures 8 . Ibuprofen, a painkiller has (S)-enantiomer with desired pharmacological activity while the (R)-enantiomer is totally inactive 1,3,16 . (R)-Naproxen is used for arthralgia pain while (S)-Naproxen is teratogenic 2 . Ofloxacin has a chiral mixture of levofloxacin [(S)-Ofloxacin] and dextrofloxacin [(R)-Ofloxacin], in which levofloxacin is 8-128 more active than dextrofloxacin 17 . So far, many chiral medicines are still used as racemates. Examples of structures of (R)-and (S)-enantiomers of chiral medicines and their activities have been shown in Fig. 1. Worldwide, there is no mandatory regulatory requirement to enforce the development of new medicines exclusively as pure active single enantiomers 13,18 . Some regulatory agencies leave the decision of a racemate or a single enantiomer formulation of a new medicine to the manufacturers 4,13,19 . However, the choice to make a racemic mixture versus a single enantiomer formulation must be justified based on quality, safety and efficacy together with the risk-benefit ratio between the two forms 4,13,19 . Due to the lack of regulatory requirements and increased technological developments, the number of new chiral entities as a pure active single enantiomers is increasing 5,15 . This contributes to the availability of both single enantiomers and racemic mixtures circulating on the market particularly in developing countries such as Tanzania 5,15 . Nonetheless, it has been reported that some manufacturers are marketing more single enantiomers of the old racemic drugs as a new drug; this is known as chiral switch, and has claims of greater activity, less toxicity or both 1,5 . In addition, chiral switches have also contributed to a number of agents being commercially marketed as both single enantiomer and racemic mixture 4,20 . Few studies conducted in some developed countries such as United States of America (USA) and Japan revealed that more single enantiomers are approved compared to racemates 18,21 . In the period between 2001 and 2011, the United States Food and Drugs Administration (USFDA) approved registration of 15 single enantiomeric medicines. These medicines showed improved therapeutic index through increased potency, selectivity and decreased side-effects compared to its corresponding racemate 21 . A review of chiral medicines approved for use in Japan was conducted between 2001 and 2003 and the results indicated an increased trend towards development of single enantiomer medicines 18 . During that period, Japan approved 3 types of chiral medicines that were classified as single enantiomer (48%), racemic (13%) and achiral (39%). It is notable that single enantiomer medicines were produced three times more than racemic medicines 18 . In developing countries, little attention has been given on the importance of chiral medicines 22 . Some guidelines for submission of technical documents for medicines registration particularly in Africa have included the requirement for submission of evidence of the occurrence of isomers, chirality or polymorphorphism . However, there is no requirement for the development of either single enantiomer medicines or racemic mixtures 6, . The decision regarding the development of a single enantiomer or racemic mixture is left to the manufacturers 24 . Moreover, there is no need for manufacturers to give justification as to why they have decided to formulate single or racemate medicines 23,24 . In addition, National Medicines Regulatory Agencies (NMRAs) in African countries have not set systems to conduct assessment of registered medicines to establish their chirality. Consequently, there is lack of information on the chirality status of registered medicines circulating on the market within the African region including Tanzania. One of the functions of the Tanzania Medicines and Medical Devices Authority (TMDA) is to register quality, safe and effective medicines 25 . Since its establishment many human medicines have been approved of which some are chiral medicines 26 . However, their chirality status (single enantiomer or racemic mixtures) is not known. It is important to know the chirality status of medicines circulating on the market as such products may cause adverse effects to the users 1,4,7,8,14 . Therefore, this study aimed to establish the chirality status of registered human medicines in Tanzania. Additionally, it was considered valuable to know the pharmacological groups of registered chiral medicines and their availability in the essential medicine list. The findings can provide common understanding of the status of chiral medicines between manufacturers and regulatory authorities. This will facilitate improvement during development of chiral medicines and also the regulatory requirements. ## Registered chiral medicines. A total of 3,573 human medicines had valid registration. Out of registered human medicines, 2,150 (60%) were chiral and 1,423 (40%) achiral. Of 2,150 registered chiral medicines, 1,591 (74%) were racemates while 559 (26%) single enantiomers. Moreover, out of all 2,150 registered chiral medicines, 550 were fixed-dose combinations of either chiral-chiral 332 (60%) or chiral-achiral 218 (40%). Medicines circulating on the market were registered by manufacturers from 45 countries. India registered more medicines 1,860 (52%) followed by Kenya 262 (7%), Germany 156 (4%), China 106 (3%), United Kingdom 109 (3%), Cyprus 100 (2.8%), Tanzania mainland 94 (2.6%), France 75 (2%), Pakistan 69 (1.9%), Switzerland 1.8%, Italy 1.8%, Malaysia 1.5%, Egypt and Belgium 1.4%. Other countries contributed less than 1.4% of the registered medicines. Details on chirality of medicines registered from specific countries have been indicated in Table 1. Out of 45 countries, 14 had registered more than 50 human medicines. Results indicated that all countries have registered both chiral and achiral medicines, in which more chiral medicines were registered compared to achiral medicines with exception of three (3) countries. These countries namely Cyprus, Tanzania and France had registered more achiral medicines than chiral medicines. Results further revealed that all countries registered chiral medicines in both racemic and single enantiomeric forms whilst racemic mixtures predominate over single enantiomers (Table 1). ## Anatomical therapeutic chemical (ATC) classification system. In the first level of ATC codes, results indicated that, the number of registered anti-infective human medicines for systemic use was significantly higher 941 (26%) than medicines used for alimentary tract and metabolism 438 (12.3%, p < 0.0001) and cardiovascular system 390 (11%, p < 0.0001). However, the percentage of registered chiral anti-infective (35%) and cardiovascular (12%) medicines were high compared to chiral medicines in alimentary tract and metabolism (7.8%) and other pharmacological groups. High proportion of chiral medicines within pharmacological groups were observed in systemic hormonal preparations (excluding reproductive hormones and insulin) 97.3% (36/37) and anti-infective 80% (753/941). The proportion of registered chiral medicines in each ATC level is indicated in Table 2. In each of the 14 main ATC groups, both racemates and single enantiomers were registered. However, more racemates were registered compared to single enantiomers with the exception of various ATC level (V) which had equal number (50%) of racemate and single enantiomer chiral medicines. Only nervous system (N) had more single enantiomers (64.7%) compared to racemates (35.3%) as indicated in Fig. 2. Chiral anti-infective medicines which have been registered in the Tanzania as per level 2 ATC classification were antibacterials for systemic use, antimycobacterials, antimycotics and antivirals for systemic use as indicated in Fig. 3. More chiral medicines had been registered in pharmacological group of antibacterials and antivirals for Vol:.( 1234567890 ## Chiral medicines listed in the National Essential Medicines List in Tanzania (NEMLIT). Out of 3,573 human medicines registered by TMDA, 2,450 (68.6%) were listed in the NEMLIT in which 1,507 (61.5%) were chiral medicines; 1,197 (79%) racemates and 310 (21%) single enantiomers. As per ATC code classification, number of anti-infective medicines for systemic use listed was 782 (32%) which was more as compared to other pharmacological groups as indicated in Fig. 5. Out of 1,123 (31.4%) medicines which were not in the NEMLIT, 481 were achiral and 642 chiral of which 394 were racemates and 248 single enantiomers. ## Most commonly used chiral medicines in Tanzania. The importation records for three years (from 2015/16-2017/18) indicated that Amoxicillin capsules and powder for suspension were the most imported Ceftriaxone injection which is also a chiral medicine was among the top ten imported medicines. These medicines were among the reported medicines with some adverse drug reactions as per Uppsala Monitoring Centre (UMC) Vigiflow/VigiLyze database. The top ten (10) imported chiral medicines for human use are indicated in Table 3. ## Discussion In this study, we conducted retrospective assessment of all registered human medicines to establish chirality status by determining ratio and types of chiral medicines circulating on the Tanzanian market. The results revealed existence of high percentage of registered chiral medicines (60%) with both racemates (45%) and single enantiomers medicines (15%) available. Within the classification of chiral medicines, racemates (74%) predominated over single enantiomer medicines (26%). These findings contrast most of studies that indicate an increased number of single enantiomer drugs and only 25% to 40% of medicines are used as racemic medicines 7,27 . A review on chirality done in Japan 17 and in United Kingdom (UK) 28 reported the tendency of registration approval of medicines has been observed to be towards chiral switching or development of the pure enantiomer 28 . The trend indicated development of 60% single enantiomers with only 5-10% of racemic mixtures and 30-35% achiral www.nature.com/scientificreports/ medicines 17 . A study conducted in Argentina on the pharmacology of chiral 2-arylpropionic acid derivatives indicated about a quarter of all therapeutic agents were marketed and administered as racemic mixtures 29 while the other findings for a study on drug chirality in anesthesia stated one-third of all synthetic drugs are chiral and marketed as racemates 30 . Some studies indicated that more than half of the medicines were chiral; however, single enantiomers were more registered than racemates 31 . On the other hand, a study conducted on side effects and toxic reactions of chiral drugs 32 and a review study on chiral drugs conducted in 2006 reported more than half of the medicines on the market were chiral with more approved racemates 1,32 . These earlier studies are in-line with our findings in which more than 50% of the medicines were chiral with more racemates registered and marketed for the past 15 years compared to single enantiomers. It has been reported that regulatory authorities in countries like U.S, Canada, Europe, China and Japan are emphasizing only on active enantiomers of chiral medicines be brought into the market. They have also issued guidelines to manufacturers outlining this requirements 19,33 . This means that the trend towards development of single enantiomers and their use depends on how stringent the regulatory authority is, and the existence of guidelines or requirements on registration of such medicines. Moreover, our results have revealed high percentage (60%) of registered chiral medicines formulated as fixed-dose combinations of chiral-chiral medicines either with racemates or single enantiomers or both, and the remaining 40% as chiral-achiral combinations. The use of the fixed-dose combination containing racemic mixtures can create more risk to patients as enantiomers may have different pharmacological activities and different levels of toxicity 11 . There are many examples of either fixed-dose combination or co-administration of chiral medicines with reported adverse events 34 . A study conducted in UK on chirality and its importance reported that senior medical advisors on regulatory bodies emphasized on the use of single chemical entities rather than combination of medicines in safeguarding the patient 11 . It has been reported that single enantiomers have less complex and more selective pharmacodynamic profiles and hence, have lesser adverse drug reactions, improved therapeutic profile and have less chances of drug interactions compared to racemic mixtures 2 . Therefore, in order to protect the public from any harm caused by the use of racemic medicines, regulatory bodies should take measures to strengthen the control of chiral medicines even for those existing in a fixed-dose combination and clear guidelines must be laid down 11 . Our results further indicate low percentage of registered single enantiomers against racemates compared to developed countries such as USA, EU and Japan where the trend of approving single enantiomers is increasing. The increased approval of single enantiomers in developed countries is contributed by existence of chiral separation technology 17,19,20 , the technology which is either minimal or lacking in developing countries. The results also indicate that manufacturers from European countries (such as Germany, UK and France) have been registering more racemic chiral medicines in Tanzania compared to single enantiomers. This was observed within antiretroviral and some of antihypertensive medicines. The reasons for registering low number of single enantiomers could be a result of non-existence of regulatory requirements to compel manufacturers to study each enantiomer and provide justification onto why a single enantiomers or racemates are beneficial 23 . The cheaper price of racemic active pharmaceutical ingredients (APIs) compared to single enantiomers may also explain their higher representation among registered medicines 21,29 . It has been reported in some studies that some API manufacturers import cheaper crude racemic mixtures of chiral medicines 21 which are expensive to separate single enantiomers on a large scale 30,35 and leaving the racemic mixture in the final drug product. An additional reason is that, due to lack of capacity in terms of skills, knowledge and infrastructure within regulatory bodies to conduct enantiomeric purity separation during post marketing surveillance studies and therefore, trigger manufacturers to register and market more racemic chiral medicines. These results support some studies that indicated, necessity of promoting and conducting chiral separation of these medicines especially during quality control stage as chirality plays a key role in clinical therapeutics 1 . It has also been reported that, the existence of problems caused by lack of techniques for separation of chiral enantiomers in medicines needs stricter control by regulatory authorities and detailed consideration during approval of newly-developed medicines 7,18,36 . Most regulatory authorities in developed countries have set regulatory controls for chiral medicines. For example, the USFDA released a policy statement on the development of stereo-isomeric medicines in May 1992 20,33,37 . The policy requires properties of each enantiomer be studied separately before decisions are made to market the medicines as one of the enantiomers or as a racemate 13,14,38 . Results also indicate that the majority of chiral medicines were registered by Indian companies amounting to 52% of all registered chiral medicines in which 44% were racemic mixtures. This indicates that chiral medicines circulating in Tanzania are mainly imported from India. These results are also consistent with number of imported medicines in Tanzania where by 47% of them come from India 39,40 . Most chiral medicines registered in Tanzania are classified in the pharmacological group of anti-infectives (35%). In addition, results revealed that 32% of all medicines listed in NEMLIT were anti-infectives including antibacterials, antivirals, antimycobacterials and antimycotics. This means that most of anti-infective medicines have been distributed at all levels in the health care facilities 41 . Considering its usage at all level and if their enantiomers are either inactive or ineffective, may contribute to the occurrence of antimicrobial resistance. Consequently, if their quality with regards to enantiomeric purity is not known then the risk to the population will be higher due to potential existence of inactive, active or toxic enantiomers. Subsequently, there is a need to conduct enantiomeric purity analysis for anti-infective medicines circulating on the market. In addition, further studies should be conducted to investigate if the inactive or ineffective enantiomer is among contributing factor in the development of antibiotic resistance. Moreover, these results indicate that, out of the top ten (10) commonly used chiral medicines, seven (7) were anti-infectives. In this study, two antibiotic chiral medicines were selected for future monitoring and linking their chirality with occurrence of adverse drug reactions. The first one is Amoxicillin (capsules and suspension) which ranks number one among the top ten and is most commonly used over all achiral and chiral registered medicines. The second one is ceftriaxone powder for injection which is number ten among the most commonly www.nature.com/scientificreports/ used chiral medicines. These two medicines have also been reported to be associated with adverse drug reactions as per UMC/WHO-Vigilyze database. It is not known if these adverse drug reactions are due to chirality of the medicines or not. A signal detection of adverse drug reaction identified by WHO-UMC reported a potential sub-group at risk which suggested that males may be at increased risk for drug induced aseptic meningitis with the use of amoxicillin and amoxicillin in combination with clavulanic acid 42 . Another 12 new signals were detected in Korea on the use of amoxicillin including ineffective amoxicillin 43 , however, the chirality factor was not reported to be associated with the new signals. In addition, a concern on safety of ceftriaxone has been reported for both adults and children 44,45 . Fatal adverse effects of injected ceftriaxone sodium were reported in China, the reason among others being the possibility of poor drug quality 45 . The results of our current study also call for more investigation to be conducted on safety and quality control of medicines especially anti-infective medicines, which are mostly used in the public. Investigations will help to determine enantiomeric purity against safety profile so as to establish if there is any relationship between chirality and reported adverse drug reactions. Enantiomeric chiral separation will prevent occurrence of any hazards to the public and may contribute to reduce risk of antimicrobial resistance in case of existence of inactive enantiomers. Our study has limitations and strengths of which limitations are that, the study reviewed all medicines registered for the past 15 years from when TMDA (previously TFDA) was established (July, 2003). However, during the TMDA establishment time, the TMDA-Integrated Management Information System (IMIS) registration database was not in place and therefore retrieval of some data was conducted by using hardcopy documents in order to capture all information. One of the strengths is that, there was no missing data. Recorded data were verified and validated using common technical documents (CTD), Public Assessment Reports and search engines such as PubMed and Google scholar. In conclusion, the study revealed the existence of both chiral and achiral registered human medicines in Tanzania with chiral medicines predominating. The study further revealed that, the proportion of racemates within chiral medicines group was significantly higher compared to single enantiomers. In addition, anti-infectives racemic mixtures were more registered and listed in NEMLIT than other pharmacological groups. The use of available medicines as racemates may cause harm to the public due to the potential existence of inactive and toxic enantiomers. Moreover, the use of racemic anti-infectives may also have consequences on antimicrobial resistance. In order to protect public health from any harm that might be caused by racemates, it is necessary for developing countries, Tanzania inclusive to develop chiral separation techniques especially during the quality control of these medicines. This is pivotal as chirality plays a key role in clinical management. There is also a need for regulatory bodies to strengthen the regulatory control of medicines to include determination of inactive and/or toxic enantiomers of chiral medicines during post marketing surveillance. ## Methods Study design. This was a retrospective cross-sectional study. All human medicines registered in Tanzania for the past 15 years from July 2003 to June 2018 were reviewed to ascertain their chirality status. Retrospective review of registered medicines. TMDA-IMIS registration database and Chirality review. We used the TMDA-IMIS registration database 26 to obtain data source of all registered medicines for the past 15 years from July 2003 to June, 2018. The ratio and chirality types of registered medicines were determined by reviewing registered medicines listed in the registration database stationed at TMDA, Dar es Salaam, Tanzania. Veterinary medicines were excluded and only human medicines were studied. All human medicines that were withdrawn from the market and those which had no valid registration status were not included in the review. In this study, all human medicines with valid registration were reviewed to identify their chirality. During the review, all information pertaining to brand name, generic name, dosage form, strength, ATC classification system (code), ATC description, manufacturer's name and country of origin were exported to the Microsoft excel data sheet for analysis. Different tools such as registration dossier (common technical document-CTD) which consisted of technical information of the medicines submitted from manufacturers during the application for registration of medicines, TMDA medicines assessment reports; Public Assessment Reports published by Stringent Drug Regulatory Authorities (SDRA) and search engines such as PubMed and Google scholar were used. Data were cleaned, verified and validated to minimize entry errors and missing information. The ATC classification was used during the review of chiral medicines in order to assess their therapeutic classification based on their pharmacological group. Moreover, the review was extended to assess if these registered chiral medicines are among the listed essential medicines and the identification of commonly used chiral medicines were done by reviewing the TMDA importation database. ## ATC classification. ATC Classification is an internationally accepted classification system for medicines that is maintained by the World Health Organization (WHO) 46 . ATC codes have been assigned to all active substances contained in medicines based on the therapeutic indication. In the ATC classification system, the active substances were divided into 14 different groups according to the system on which they act and their therapeutic, pharmacological and chemical properties 46 . Medicines were classified in five (5) different levels and divided into fourteen main groups (1st level), with pharmacological/therapeutic subgroups (2nd level). The 3rd and 4th levels are chemical/pharmacological/therapeutic subgroups and the 5th level is the chemical substance 46 . The 2nd, 3rd and 4th levels are often used to identify pharmacological subgroups and are considered to be more appropriate than therapeutic or chemical subgroups 46 . In this study, two levels "i.e. " 1 st and 2 nd were considered to comprehensively categorize registered chiral medicines into their pharmacological subgroups. www.nature.com/scientificreports/ NEMLIT. Chiral medicines were further reviewed to assess their existence in the NEMLIT to determine if they were among essential medicines. NEMLIT reflects the policy of the Government of Tanzania of ensuring availability of safe and efficacious essential medicines to all its citizens 41 . It is therefore, a key tool which is used effectively to promote access to essential medicines to achieve maximum therapeutic benefit and optimize patient outcomes. The NEMLIT fifth edition (2017) was used during the review. It is in line with the WHO model list of Essential Medicines (EML) 20th edition (2017). NEMLIT gives guidance to healthcare workers at all levels including dispensaries and health centers on the medicines to be prescribed. It also gives restrictions on the use of antibiotics in health facilities to those selected as the most appropriate for use at each level of health care delivery 41 . TMDA-IMIS importation database and WHO VigiLyze database. The most commonly used chiral medicines in Tanzania were assessed by using the importation database located at TMDA. All medicines imported for public and private use were assessed. In addition, the WHO VigiLyze database was also used to identify commonly used chiral medicines with reported adverse drug reactions. Data management and statistical analysis. The obtained data from the reviews were checked for any inconsistencies. The data was entered into a Microsoft Excel spreadsheet (version 2010), verified and exported to STATA Version 15 software for statistical analysis. Descriptive statistics was determined for types of chirality, ATC classification-pharmacological subgroup and those listed in NEMLIT. Moreover, countries of origin of chiral medicines registered in Tanzania and two mostly used chiral medicines in Tanzania were also identified. A one-sample t-test between proportions was performed to determine whether there was a significant difference between the percent of chirality type, country of origin, ATC classification and existence of chiral medicines in NEMLIT. The level of confidence required for significance was set at p < 0.05.

chemsum

{"title": "A retrospective cross-sectional study to determine chirality status of registered medicines in Tanzania", "journal": "Scientific Reports - Nature"}

protonated_glycine_supramolecular_systems:_the_need_for_quantum_dynamics

## Abstract: IR spectroscopy is one of the most commonly employed techniques to study molecular vibrations and interactions. However, characterization of experimental IR spectra is not always straightforward. This is the case of protonated glycine supramolecular systems like Gly 2 H + and (GlyH + nH 2 ), whose IR spectra raise questions which have still to find definitive answers even after theoretical spectroscopy investigations. Specifically, the assignment of the conformer responsible for the spectrum of the protonated glycine dimer (Gly 2 H + ) has led to much controversy and it is still debated, while structural hypotheses formulated to explain the main experimental spectral features of (GlyH + nH 2 ) systems have not been theoretically confirmed. We demonstrate that simulations must account for quantum dynamical effects in order to resolve these open issues. This is achieved by means of our divide-andconquer semiclassical initial value representation technique, which approximates the quantum dynamics of high dimensional systems with remarkable accuracy and outperforms not only the commonly employed but unfit scaled-harmonic approaches, but also pure classical dynamics simulations. Besides the specific insights concerning the two particular cases presented here, the general conclusion is that, due to the widespread presence of protonated systems in chemistry, quantum dynamics may play a prominent role and should not be totally overlooked even when dealing with large systems including biological structures. ## Introduction Protonated systems are ubiquitous in chemistry. 1 They are involved in many different processes and activities, ranging from organic reactions and intermediates to biological and interstellar-medium events. Furthermore, protonation is determinant for the chemical properties of heteroatomic compounds, such as amino acids. For instance, hydrolysis of amides, peptides, and proteins at biological pH is initiated and driven by the process of protonation. In general, the electronic and conformational structure of proteins as well as their dynamics are strongly influenced by protonation with the resulting three-dimensional structure playing a key role in their biological activity. Protonated glycine compounds are pivotal examples of protonated systems because they are the smallest building blocks of more complex biological entities. A full comprehension of their dynamics is indeed essential for a correct understanding of the stability and reactivity of many other protonated systems. For this reason, in the past, protonated glycine compounds have been the subject of extensive experimental and computational studies. However, there are some fundamental questions about these systems which are still open: specifcally, to what extent is the proton shared between the amide and the carboxylic group? Is it a static or a dynamical effect? Do nuclear quantum mechanical contributions play a major or a minor role in determining the properties of these protonated compounds? One should expect very peculiar quantum mechanical effects when the proton is shared between nucleophilic groups either belonging to different molecules, like in the case of the glycine proton-bound dimer, or when they are part of the same molecule, as in protonated glycine. The main reason for this expectation is that the proton is the only ion with basically zero ionic radius and it has the lightest mass. These peculiarities are at the origin of proton mobility and reactivity, and one would expect quantum mechanical contributions to be determinant. This paper aims at providing answers to the open questions illustrated above and at estimating the impact of quantum mechanical effects by comparing quantum and classical simulations versus available experimental results. One popular experimental spectroscopic technique to study the vibrational features of protonated compounds is represented by infrared multiple photon dissociation (IRMPD). IRMPD provides enhanced signals of gaseous molecular ions in the infrared region once they have been trapped in the high vacuum cells of mass spectrometers. 2,10-13 However, when applied to the glycine proton-bound dimer, Gly 2 H + , IRMPD does not permit undisputed identifcation of which Gly 2 H + conformer is more representative of the IRMPD spectrum. This open issue, which has to be resolved in order to properly characterize the structural properties not only of the dimer but also of complex peptide chains, has been the topic of previous joint experimental and computational studies in 2005 by McLafferty et al. 14 and in 2007 by McMahon et al. 15 with conclusions clearly at odds. Specifcally, Gly 2 H + has two low energy gas-phase conformers, named CS01 and CS02, plus a zwitterionic form ZW01 (see the ESI † for more information) which rapidly interconverts to CS01 during the dynamics. In the CS01 conformer, the two moieties making up the dimer are bound by means of an O/H + N interaction, while a N/H + N interaction is peculiar of CS02. According to some studies, including McLafferty's one, 9,14 CS02 is the most representative conformer in the experimental IRMPD spectrum, while for other studies, with McMahon's paper among them, 4,5,15 it is CS01 that deserves recognition, so a defnitive conclusion has not been reached yet. In both studies 14,15 the authors employed a scaled-harmonic approach to interpret the main features of the IRMPD spectrum. In the scaled-harmonic technique, 16 frst the normal mode frequencies (i.e. the purely harmonic frequencies of vibration) at the minimum geometry are calculated by diagonalizing the equilibrium nuclear Hessian matrix and taking the square roots of the Hessian eigenvalues. Then, they are scaled to account for anharmonicity and coupling between modes. Such an approach is widely employed since it is easily doable even for large size molecules. It requires calculation of just a single Hessian matrix. However, the approach remarkably neglects any dynamical and anharmonic effects that may become crucial when interactions such as hydrogen bonds dominate the interaction picture. 17 Even if several research groups have provided full sets of scaling constants for the different levels of theory and electronic basis sets employed 16,18 as well as different scaling constants for calculations of different observables (frequency, zero point energy, enthalpy, entropy, etc.), the scaled harmonic approach is misleading for the interpretation of the glycine proton-bound dimer spectrum. Furthermore, it is generally classifed as an ab initio method in an improper way, since an empirical tuning parameter is enforced. ## Protonated glycine dimer To prove the inaccuracy and ineffectiveness of scaled-harmonic frequency estimates, following ref. 15, we performed geometry optimizations of Gly 2 H + at the DFT-B3LYP level of theory with the 6-311+G(d,p) basis set, followed by a scaled-harmonic analysis. In agreement with the previous studies, we found that conformer CS01 is the lowest energy conformer, while CS02 is just 2.1 kcal mol 1 higher in energy. Given this small difference in energy, the determination of the conformer responsible for the IRMPD spectrum in panel (a) of Fig. 1 implies the assignment and the interpretation of the spectrum in full detail. By scaling all the harmonic frequencies at the CS01 geometry by a factor equal to 0.96, panel (b), the OH and NH stretching region is well reproduced while the mid-range one is not. Likewise, simulations based on the 0.97 factor suggested in ref. 14 would follow a similar destiny. Conversely, if the scaling factor equal to 0.985 suggested in ref. 15 is applied, then the stick spectrum of panel (c) is found. This scaled-harmonic spectrum mimics reasonably well the experimental peaks in the mid-range region between 1000 cm 1 and 2000 cm 1 . However, the same scaling factor performs poorly in the Hstretching region above 3000 cm 1 . Since the anharmonicity parameters for each vibration are not known a priori, we conclude that it is impossible to model IR spectra on the basis of harmonic calculations, at least when different conformers are present. For these reasons, we deem that previous investigations cannot be recognized as conclusive ones. A computational technique able to account for conformational and dynamical effects should be conveniently based on (quantum) molecular dynamics, since the dynamics allows exploration of the actual Potential Energy Surface (PES) even far from the harmonic region. Such a non-local approach may be crucial for a correct interpretation of hydrogen bonding. The quantum dynamical way to spectroscopy and frequency computation (i.e. the power spectrum) is given by the Fourier transform of the autocorrelation of the time-evolved nuclear wavepacket averaged over the quantum density matrix of vibrational states 22 Eqn ( 1) includes all quantum mechanical spectroscopic information like zero point energy (ZPE), fundamental and overtone frequencies, anharmonicities and couplings, tunneling effects and quantum resonances between overtones and fundamentals. The classical equivalent of eqn (1) is the Fourier transform of the velocity autocorrelation function ( where v(t) is the vector of the atomic velocities at time t. Here the average is over a suitable ensemble of initial phase space confgurations. 6, Eqn ( 2) is limited to the calculation of classical fundamental frequencies, mode couplings and resonances. In other words, it accounts for the classical contribution to anharmonicity only. Anyway, both approaches are dynamical and represent a step forward with respect to single point harmonic calculations. Unfortunately, when dealing with high dimensional systems, purely quantum mechanical simulations based on eqn (1) are out of reach because of the so-called curse of the dimensionality problem. Furthermore, accurate and fast-to-evaluate analytical PESs are usually not available and must be replaced by more computationally expensive ab initio "on-the-fly" calculations, whereby the dynamics can be performed (even if at a lower level of electronic theory), and which demand for a theoretical formalism that permits a convenient interface to them. Semiclassical (SC) dynamics can be interfaced to ab initio "on-the-fly" calculations straightforwardly, so we adopted it to calculate I qm (E). In a SC simulation, quantum mechanical effects are reproduced by employing many entangled coherent states, which are time-evolved on top of classical trajectories. The semiclassical initial value representation (SC-IVR) approximation of quantum dynamics has been proved to be reliable and robust. Recently, we developed an SC-IVR method based on a "divide-and-conquer" strategy (DC SCIVR) to undertake the spectroscopic calculations of eqn (1). 48 The method can deal with very high dimensional molecular and supra-molecular systems, and it is very accurate when compared to available exact vibrational quantum mechanical calculations. 49,50 Specifically, our DC-SCIVR method has been tested successfully against systems with up to hundreds of degrees of freedom, 48 and in particular it has been employed to study the vibrational features of the protonated water dimer, the Zundel cation. The results are very accurate (within a few wavenumbers) even for the vibrational bands of the proton doublet in the region of the O-H-O stretching frequency and associated with the proton transfer ($1000 cm 1 ), when compared to exact grid-based quantum dynamics results on the same PES. The DC-SCIVR idea is to calculate the power spectrum as a sum of partial reduced-dimensionality spectra. First, full dimensional ab initio on-the-fly classical trajectories are calculated. Then, the normal modes are divided into vibrational subspaces according to their mutual coupling, 48,49,51 and the partial spectra are calculated by projecting the classical trajectory information according to the following formula 48 Ĩqm ðEÞ ¼ 1 2pħ where F is the dimensionality of the vibrational subspace, and the multi-dimensional phase-space integration is characterized by a positive-defnite time-dependent integrand made of the classical action ( St (p(0),q(0))), the phase of the semiclassical prefactor (ft), 46 and the overlap between the reference state |Jĩ and the coherent state |p(t),q(t)i. More details can be found in the ESI. † Fig. 2 shows the resulting DC-SCIVR power spectra, where the ZPE has been shifted to zero for better comparison with the experiments, which cannot be used to measure it. The reported peaks are those of the vibrational modes which have the highest oscillator strengths. A peak-by-peak comparison between the semiclassical spectra of the two possible conformers shown in panel (a) and (c) in Fig. 2 and the experimental one which appears in the middle panel (b) shows clearly that in panel (a) all the experimental vibrational features are accurately reproduced, differently from the case of panel (c) where the agreement is not as good. By looking at the actual calculated frequencies for each peak, a mean absolute error from the experimental peaks of 14 cm 1 is associated with the spectrum of conformer CS01 in panel (a), while a deviation of 32 cm 1 characterizes the spectrum of conformer CS02 in panel (c). This discrepancy in the mean absolute error values is pretty signifcant and conclusive even upon weighing in the typical accuracy of semiclassical calculations. 49 More specifcally, in the fngerprint region, the two carbonyl stretching frequencies (violet and orange lines) are degenerate in the case of the CS02 conformer, while they are not for CS01, in agreement with the experimental spectrum. Furthermore, the two vibrational peaks at around 1500 cm 1 , corresponding to the fundamentals for the OH hydrogenbonded bending (blue profle) and the umbrella inversion mode (green line), are well reproduced in panel (a), while panel (c) shows more elaborate vibrational coupling features. Similar considerations are valid also for the symmetric NH stretching Starting from the left, the red continuous line is for the free OH bending mode, the blue one is for the OH hydrogen-bonded bending, the green line is for the umbrella inversion mode, the violet and orange ones are both for carbonyl stretchings, the cyan line is for the NH symmetric stretching and, finally, the magenta one is for the free OH stretching. and the free OH stretching in the high frequency region, with the spectrum in panel (a) better resembling the experimental profle. We stress that our assignment of the experimental spectrum to conformer CS01 has been obtained without any tuning parameter and in a fully ab initio way. Furthermore, no ad hoc frequency scaling factor nor ftting procedures have been applied to the calculated spectra reported here. From our simulations we note that there are numerous peaks in the frequency domain between the fngerprint peaks and the NH and OH stretching region which have not been detected by the experiments. These peaks belong to mode overtones and combinations of them, and are consequently experimentally much less intense. Finally, energetics calculations for both conformers have been performed at a quantum level by adding ZPE values to the classical minimum, i.e. the bottom of the well, and CS01 has been found to be about 2.5 kcal mol 1 more stable than CS02, revealing that the conformer that we identifed as the major contributor to the experimental IR spectrum is also the (quantum mechanical) global minimum. Despite the success of the semiclassical simulations presented above, one key methodological question remains open. In fact, if on the one hand quantum effects are hallmarks of spectroscopy, on the other hand for systems of dimensionality similar or bigger than Gly 2 H + it could be argued that a classical picture is enough to describe with sufficient accuracy the spectral features of at least fundamental transitions. This would require much less effort since a semiclassical simulation is signifcantly more computationally intense than a classical one. Eqn (2) requires calculating at each time step the cartesian velocities v(t) of the nuclei only. Instead, in eqn (3), the calculation of I qm (E) in semiclassical approximation implies the evaluation of not only the position and the velocities of the nuclei at each time step, but also the nuclear Hessian (for evolution of the phase term). This is about an order of magnitude more expensive in terms of computational efforts. ## Protonated glycine tagged by hydrogen molecules To point out clearly the importance of quantum mechanical effects in vibrational spectra influenced by proton dynamics, we consider that, recently, Masson, Williams and Rizzo published a series of very interesting IRMPD spectra, where protonated glycine, GlyH + , was tagged by an increasing and controlled number of hydrogen molecules. 3 We focus particularly on two of these investigations. The frst one regards protonated glycine solvated by a single hydrogen molecule (GlyH + H 2 ) + , while in the other instance three H 2 molecules are involved (GlyH + 3H 2 ) + . The minimum geometries of these systems are reported in Fig. 3. This fgure suggests that panel (a) is characterized by a strong hydrogen bond interaction between one of the amide hydrogens and the carbonylic oxygen atom of the carboxylic acid group, while in panel (b) the presence of the three hydrogen molecules may suppress the hydrogen bond interaction by inducing a reorientation of the amide group. In panel (a) the H/O distance at the minimum geometry is about 1.90 , while in panel (b) the distance is equal to 2.52 . This last distance is still shorter than the sum of hydrogen and oxygen van der Waals radii (2.72 ), which is considered, as a rule of thumb, the limit for hydrogen bonding. To check whether the hydrogen bond is lifted or not by virtue of the H 2 tagging process and if quantum mechanical effects play any relevant role in this kind of interaction, we frst performed ab initio "on-the-fly" DC-SCIVR simulations using the DFT-B3LYP level of theory and employing the aVDZ basis set, and then compared the results with the experimental spectra. report it in panel (c) separately for each mode for a better comparison with the experiment. The main spectroscopic features are reproduced, even if the signal corresponding to the NHa and NHb bands is quite broad. Finally, in the semiclassical spectrum of panel (d), calculated with eqn (3), the fundamental bands are accurately reproduced, with the addition of overtones that are too weak to be detected in the experiment and that are missing in the classical and scaled-harmonic simulations. In general, the simulated peaks are broader than the experimental ones because, on the one hand, experiments are performed at very low temperature (a few K) and rotations are hindered or even blocked, while, on the other hand, in the simulations the dynamics is propagated only for a short time (less than 1 ps) before the Fourier transform is undertaken, and every mode (including internal rotators) is given an amount of energy according to its contribution to the ZPE and left free to evolve without any artifcial constraints. Furthermore, the dynamics of the hydrogen bonding may contribute to the broadening of the NH stretching bands as shown by Gaigeot and coworkers by applying fnite temperature classical molecular dynamics to small protonated peptides, such as Ala 2 H + and Ala 3 H + . 6,7,25 We now turn to the other system, i.e. (GlyH + 3H 2 ) + , reported in panel (b) of Fig. 3. Fig. 5 shows the spectra corresponding to those presented in Fig. 4 but this time for this bigger system. One may think that hydrogen molecules do not interact signifcantly with GlyH + and that it is possible to obtain in good approximation the IRMPD signal of the isolated molecule. However, there are clear differences between the experimental spectra of Fig. 4 and 5. One of them is represented by the blue-shifted NHa peak. As usual, scaled-harmonic calculations are shown as a stick spectrum in panel (a) of Fig. 5 and they fail to account correctly for the anharmonicity of the NHa stretching motion. Once more, the scaling of harmonic frequencies brings us to a dead end. A classical approach based on eqn (2) is not helpful in this circumstance as demonstrated by the set of spectra in panel (c) of Fig. 5 that clearly show classical mechanics overestimating the NHa stretching frequency. We believe that this is due to the fact that the intramolecular hydrogen bond and the dynamics of the involved proton have a prevalent quantum nature. In other words, a scaled-harmonic or classical dynamics approach leads to the wrong conclusion that the intramolecular hydrogen bond is broken in the presence of 3H 2 molecules interacting with GlyH + . Conversely, a semiclassical simulation based on eqn (1), reported in panel (d), reproduces accurately all the vibrational features of the IRMPD spectrum also in this case, including the strong anharmonicity of the NHa stretching and the consequent red shift, thus confrming that the hydrogen bond interaction is only weakened and not completely broken, even in the presence of three H 2 molecules coordinated to the amide group. Another key difference between Fig. 4 and 5 lies in the appearance of a second OH stretching band, located at 3491 cm 1 , and labeled as OHr. Masson et al. 3 suggested that this band is the result of a confguration where one of the three H 2 molecules interacts with the carboxylic group. Indeed, the peak is red-shifted by about 55 cm 1 with respect to the free OH stretching (the OHb band). This would mean that the experimental spectrum (b) of Fig. 5 actually originates from two different conformers. To validate the previous conformational hypothesis we consider a confguration with a single H 2 molecule tagging the carboxylic group. This geometry is not stable experimentally (in fact the OHr peak appears only when 3 or more H 2 molecules are involved), but it can be investigated theoretically. The system is reported in panel (a) of Fig. 6, and we focus on the OH stretching. Still in panel (a) the harmonic stick spectrum (at the DFT-B3LYP level of theory with the aVDZ basis set) for the OH stretching is presented after scaling by a factor of 0.96, which is the same coefficient employed in the previous simulations. This estimate is defnitely off the mark. In contrast, both the classical (panel c) and the semiclassical (panel d) peaks are quite accurate for the OH stretching, confrming that the OHr band is indeed due to the interaction between a H 2 molecule and the carboxylic group. The presence of the H 2 molecule weakens the OH bond, leading to the observed red shift. ## Amino group deuteration To further prove that the differences between the classical and the semiclassical spectra reported respectively in panel (c) and (d) of Fig. 5 are due to quantum mechanical effects only, we quenched them by deuterating all the three hydrogen atoms, pictorially represented by the gray atoms of the molecule in Fig. 7. Then, we calculated the spectra for the deuterated GlyH + molecule tagged by the three H 2 molecules. We focus on the amino group modes and selectively plot the NDa (previously NHa) band, both using the classical spectrum formulation of eqn (2) and the quantum formulation of eqn ( 1) and ( 3). The results are reported in Fig. 7. The NDa band is centered around 2350 cm 1 , which is signifcantly red-shifted with respect to the previous NHa band, because of the heavier deuterium mass. However, we note that the classical and semiclassical peaks are almost identical in this case. This proves that the previous discrepancy of about 150 cm 1 between the classical and the semiclassical NHa band location of Fig. 5 was exclusively due to a quantum mechanical effect of the light hydrogen atom. It is quite surprising that this quantum anharmonic effect is so huge. However, when considering the strong anharmonicity of the NHa potential well and the consequent huge delocalization of the quantum mechanical vibrational eigenfunction (as pictorially represented in the ESI †), the prominent quantum mechanical nature of this hydrogen bond interaction appears fully justifed. ## Conclusions We conclude by remarking the importance of employing a quantum dynamical approach in calculating vibrational frequencies and going beyond the scaled-harmonic level. This has been demonstrated by means of divide-and-conquer semiclassical dynamics, which has permitted us to reproduce experimental anharmonicities quite well and to explain some open issues involving the protonated glycine dimer and the tagging of protonated glycine with molecular hydrogen. In particular, for the former, the CS01 conformer has been assigned consistently in the whole frequency range, while, for the latter, some peculiar spectral quantum features due to hydrogen bonding and intermolecular interactions have been rigorously explained, a task that neither scaled-harmonic nor classical approaches were able to accomplish. On this point, we note that the reference experiments were performed at very low temperatures, so we did not run standard thermalized classical simulations (which would have provided just harmonic estimates), but we estimated a classical analog of the quantum mechanical vibrational spectral density. Nevertheless, these classically inspired calculations were not as satisfactory as the semiclassical ones. Interestingly, due to the interaction, vibrational frequency calculations of the tagging H 2 molecules display a red shift ($50 cm 1 ) comparable to that of the OH stretching of glycine. Remarkably, the adopted DFT-B3LYP level of electronic theory is not only suitable for a realistic description of the entire supramolecular system but is also able to provide frequency estimates in quantitative agreement with the experiments. Finally, we are able to answer the three questions with which we introduced the paper by stating that quantum effects certainly play a very important role in these protonated systems; the intramolecular hydrogen bond interaction has a strong impact on the NH stretching revealing an elevated degree of delocalization of the proton shared with the carboxylic group; the very same interaction is influenced by the dynamics with the hydrogen bond being less and less directional as the number of tagging molecules increases. All these fndings point out very clearly the crucial role that quantum dynamics may play, suggesting that it should not be neglected even when dealing with larger systems.

chemsum

{"title": "Protonated glycine supramolecular systems: the need for quantum dynamics", "journal": "Royal Society of Chemistry (RSC)"}

synthesis_and_crossover_reaction_of_tempo_containing_block_copolymer_via_romp

## Abstract: We report on the block copolymerization of two structurally different norbornene monomers (±)-endo,exo-bicyclo[2.2.1]-hept-5ene-2,3-dicarboxylic acid dimethylester (7), and (±)-endo,exo-bicyclo[2.2.1]-hept-5-ene-2,3-dicarboxylic acid bis(1-oxyl-2,2,6,6tetramethyl-piperidin-4-yl) ester (9) using ruthenium based Grubbs' type initiators [(PCy 3 ) 2 Cl 2 Ru(benzylidene)] G1 (PCy 3 = tricyclohexylphosphine), [(H 2 IMes)(PCy 3 )Cl 2 Ru(benzylidene)] G2 (H 2 IMes = 1,3-bis(mesityl)-2-imidazolidinylidene), [(H 2 IMes)(py) 2 Cl 2 Ru(benzylidene)] G3 (py = pyridine or 3-bromopyridine) and Umicore type initiators [(PCy 3 ) 2 Cl 2 Ru(3phenylinden-1-ylidene)] U1 (PCy 3 = tricyclohexylphosphine), [(H 2 IMes)(PCy 3 )Cl 2 Ru(3-phenylinden-1-ylidene)] U2 (H 2 IMes = 1,3-bis(mesityl)-2-imidazolidinylidene), [(H 2 IMes)(py)Cl 2 Ru(3-phenylinden-1-ylidene)] U3 (py = pyridine or 3-bromopyridine) via ring opening polymerization (ROMP). The crossover reaction and the polymerization kinetics were investigated using matrix assisted laser desorption ionization mass spectroscopy (MALDI-TOF) and nuclear magnetic resonance (NMR), respectively. MALDI showed that there was a complete crossover reaction after the addition of 25 equivalents of the second monomer. NMR investigation showed that U3 gave a faster rate of polymerization in comparison to U1. The synthesis of block copolymers with molecular weights up to M n = 31 000 g/mol with low polydispersities (M w /M n = 1.2) is reported. ## Synthesis and crossover reaction of TEMPO containing block copolymer via ROMP Olubummo Adekunle, Susanne Tanner and Wolfgang H. Binder * ## Introduction Block copolymers are macromolecules composed of linear or non-linear arrangements of chemically different polymeric chains. If the different blocks are incompatible, a rich variety of well defined self-assembled structures both in bulk and selective solvents arises . The synthetic approach to block copolymers has been widely discussed and achieved extensively via living polymerization methods. Thus, besides acyclic diene metathesis polymerization (ADMET) , ring opening Scheme 1: Grubbs G1-G3 and Umicore U1-U3 catalyst. metathesis polymerization (ROMP) is another type of olefin metathesis polymerization that can be used for the synthesis of block copolymers. Early examples of catalysts for ROMP were based on molybdenum alkylidene catalysts, however, the true breakthrough of the method was hampered by the restricted functional group tolerance of Schrock initiators due to their sensitivity towards protic solvents and air . With the advent of the Grubbs' catalyst G1 (see Scheme 1) and related complexes as initiators, polymerization reactions can now not only be performed in protic media but also without rigorous exclusion of molecular oxygen. However, these advantages are hampered by the considerable lower activity of catalysts such as G1 when compared with Schrock's molybdenum catalysts . Often, the polydispersity indices of the resulting polymers obtained with initiator G1 are large with values ranging between 1.3 and 1.5 arising from an unfavorable rate of initiation (k i ) relative to propagation (k p ) as well as considerable secondary metathesis (backbiting). Grubbs' second generation catalyst G2 displays an activity comparable to the Schrock type initiators. It exhibits a higher functional group tolerance than G1, but initiation by catalyst G2 is often slow as a result of the slow dissociation of the phosphine group, sometimes limiting its application in polymer synthesis. Alternatively, Grubbs' third generation catalyst G3 introduced by Grubbs et al. has an ultrafast initiating ruthenium benzylidene. The rate of reaction of G3 with ethyl vinyl ether thus is six orders of magnitude higher than for G2 , leading to a faster initiation and often lower polydispersities of the resulting polymers. Recently, structural variations of G1-G3 catalysts generated a new series of catalysts U1-U3 bearing indenyl-carbenes instead of benzylidene-carbenes. These new catalysts are now commercially available and are well known as the Umicore catalysts (NEOLYST ™ ). However, their synthetic profile with respect to the synthesis of block copolymers is largely unexplored. As we recently have reported extensively on the use of ROMP methods in blockcopolymer synthesis , either via direct copolymerization or coupled to postmodification methods via azide/alkyne-"click"-chemistry , the crossover reaction of more complex monomers remains the crucial factor in achieving defined BCP's with low polydispersities. In a recent example, the crossover reaction of various monomers with the Grubbs' type catalysts G1-G3 was studied in detail via MALDI mass spectrometry , revealing a more detailed picture of the crossover reaction (Scheme 2). Thus mass spectrometry could often demonstrate insufficient crossover reactions between monomers of different reactivity such as monomer A 7 and monomer T 9, despite a low polydispersity when the crossover reaction was monitored by conventional GPC methods . A semi-quantification method of the respective spectra allowed a good correlation between the rate-constants of initiation and propagation of the different monomers. The polymer was synthesized for MALDI analysis. b The experiment was performed for kinetic measurements by taking samples every 5 minutes. The current publication describes the synthesis of block copolymers A n T m composed of monomers 7 and 9, initiated via the catalysts U1-U3, as well as mass spectrometric investigations of the crossover reactions via MALDI methods. The incorporation of the free radical 9 into block copolymer is an important contribution in the generation of polymers for reversible charge storage materials, as monomer 9 can accept or donate electrons reversibly. ## Results and Discussion The polymerization of monomer 7 was investigated using catalysts U1-U3 (see Table 1). Basically, the catalyst U3 showed good polymerization behavior, furnishing the homopolymers (entries 2, 3) with excellent control of chain length and low polydispersities (M w /M n = 1.2). The catalysts U1 and U2 gave poor results (see entries 4 and 5) presumably due to slow initiation and fast polymerization, which is in accord with the struc- turally similar catalysts G1 and G2 (see Scheme 1). Similarly, the polymerization of monomer T 9 was investigated, which gave good results with the catalysts G2, G3, and U3 (entries 6, 7, and 8). The other catalysts G1, U1, and U2 did not yield good polymerization results (data not shown) and were therefore not considered further for the synthesis of the respective block copolymers. The 1 H NMR spectrum (Figure 1) of the respective homopolymer (A 50 ) clearly shows the expected resonances together with the resonances of the indenyl-moieties of the initiatorstructure. The spectrum further revealed that the unsaturated polymer exhibited no stereoregularity (cis/trans ~ 50/50 see the peaks at 5.2 and 5.4 ppm in Figure 1) which is in accordance with results reported in the literature . Figure 2 shows the relevant region of the 13 C NMR spectrum, with the approximately equal peak intensities indicating an equal m:r ratio. Thus the polymerization yields the respective polymer, although in poor yields which is underlined when monitoring the kinetics of the polymerization of monomer A 7 using catalysts U1 and U3 (see Figure 3, Figure 4 and Figure 5). As expected in accord-ance with the known polymerization reactions of the respective Grubbs' type catalysts, catalyst U3 polymerizes significantly faster (the polymerization reaction is complete after ~20 seconds) whereas the polymerization initiated with catalyst U1 takes significantly longer and never yields significant amounts of the homopolymer (yield < 10 %). As the polymerization kinetics of 7 using catalyst U3 could not be monitored effectively with GPC because it was too fast (50 units were polymerized in less than 1 minute), the kinetics were monitored by NMR. NMR measurements were conducted every 8 seconds and the result showed that the polymerization was complete within ~20 seconds as shown in Figure 4 and As monomer T 9 is a free stable radical, the progress of its polymerization with catalyst U3 could not be monitored by 1 H NMR spectroscopy. Therefore, the conventional method of following the M n vs. time (t) profile was carried out as shown in Figure 6. Chain growth with a maximum polydispersity of M w /M n ~1.3 was observed, clearly proving the high precision of this type of polymerization reaction. As the polymerization of both, monomer A 7 and monomer T 9 proceeded well with catalyst U3, the synthesis of the BCP was achieved by use of this initiating system to yield the respective BCP-A 10 T 10 , A 20 T 20 , A 25 T 25 and A 50 T 50 with the expected molecular weight and with low polydispersity (see Table 1, entries 18-21). The GPC traces of A 25 block and the A 25 T 25 block copolymer are shown in Figure 7, indicating the expected shift in the retention time after addition of monomer T 9 after all of the monomer A 7 had been consumed. In order to achieve a deeper insight into the exact nature of the crossover reaction when changing from monomer A 7 to monomer T 9 with catalyst U3, the respective reaction was monitored according to our previous methods using MALDI mass spectrometry . Thus homopolymer A 25 was initiated with catalyst U3 and subsequently reacted with 1, 2, 5 and 25 equivalents of monomer T after all of the monomer A 7 had been consumed. The respective samples were then quenched with ethyl vinyl ether, and subsequently analyzed by MALDI-TOF mass spectrometry and GPC. The GPC results are shown in Table 1, entries 9-14 and 18-21, indicating that with increasing amount of added monomer T an equal increase of M n can be observed. However, in order to check for the detailed composition of the reaction mixture, MALDI spectra were measured. As shown in Figure 8, homopolymer A 25 can be desorbed well in MALDI, featuring the respective A n Na + -ions as a pure series. Thus the homopolymer A n can serve as molecular probe for the subsequent desorption of the individual A n T 1, 2, 5 -species. The MALDI spectrum of the crossover reaction of A 25 with exactly one equivalent of monomer T 9 using U3 as initiator is shown in Figure 9. Thus, together with the still present A n -series (visible as A n Na + -series), the respective crossover species A n T 1 , and A n T 2 can be seen as the respective Na + -ions. These results demonstrate that a large amount of A n -species did not participate in the crossover reaction, since due to the fast polymerization of monomer T 9, it was rapidly consumed, leading to A n T 2 -species and its respective higher homologues. The respective MALDI spectrum of the crossover reaction of the homopolymer A 25 with exactly two equivalents of monomer T 9 is shown in Figure 10. Again, a significant amount of homopolymer A n (visible as A n Na + -series) is present in the reaction mixture, the respective crossover-species A n T 1 , and A n T 2 can be seen as the respective Na + -ions. Additionally, the respective series A n T 3 is visible, indicative of the further chain growth process after the crossover reaction. Again, despite the excess of T n -species a large amount of A n -species did not participate in the crossover reaction due to the fast polymerization of monomer T 9. MALDI spectra of a further series of block copolymers A 25 T 5 and A 25 T 25 was carried out in order to check for the presence of residual homopolymer A 25 in the polymer mixture (see Figure 11). We could not detect any residual homopolymer in either these final samples. As it is known from our previous investigations, that the homopolymer A n in MALDI is desorbed preferentially by a factor of 13 with respect to the A 25 T n -species, this now indicates a complete cross-over reaction and thus the successful preparation of block copolymers A n T m via this methodology. Basically, this synthetic approach now allows the synthesis of AT type BCP's with high precision and chain length control up to molecular weights of ~31000 g/mol. ## Conclusion The synthesis of new block copolymers containing free radical centers within one block via ROMP has been described. MALDI analyses especially provide a detailed picture of the crossover reaction. Basically, the NEOLYST ™ catalysts are comparable to the well known Grubbs' catalysts, indicating a similar profile of initiation and propagation. However, the catalyst U3 is especially a highly potent catalyst for ROMP and displays a broad profile of tolerance against functional groups within the monomer, enabling the successful synthesis of block copolymers containing free-radical species in high densities. ## Experimental General Remarks Solvents/Reagents/Materials: Catalysts G1, G2 and G3 were obtained from Sigma-Aldrich. Catalysts U1, U2 and U3 were obtained as gifts from the Umicore chemical company. All reagents used for the synthesis of norbornene monomers 7 and 9 were obtained from Sigma-Aldrich Chemical Co. (Germany) and used as received without further purification unless otherwise indicated. Bicyclopentadiene (100%), fumaric acid (99+%), thionyl chloride (99+%, Fluka), pyridine (99.8%), methanol and 4-hydroxy-2,2,6,6-tetramethyl-piperidin-1-oxyl (TEMPOL) were obtained from Sigma-Aldrich and used without further purification. Dichloromethane (CH 2 Cl 2 ) was freshly distilled over CaH 2 and degassed with argon prior to use. Other solvents such as hexane and ethyl acetate were used after distillation. Instrumentation: 1 H NMR spectra were recorded on a Varian Gemini 400 MHz FT-NMR spectrometer, and MestRec (4.9.9.9) was used for data interpretation. The polymerization kinetics of the polymerization reactions with both catalysts U1 and U3 were measured at 25 °C on a 200 MHz FT-NMR spectrometer using CDCl 3 as a solvent. GPC analysis was performed on a Viscotek VE2001 system with THF as the eluant at a flow rate of 1 ml/min and an injection volume of 100 µL. Polystyrene standards were used for conventional external calibration using a Viscotek VE3580 refractive index detector. Positive ion MALDI-TOF (matrix-assisted laser desorption ionization time-of-flight) measurements were performed on a Bruker Autoflex-III instrument equipped with a smart ion beam laser. Measurements were carried out in linear and reflector mode. Samples were prepared from THF solution by mixing matrix (20 mg/ml), polymer (20 mg/ml), and salt (20 mg/ml solution) in a ratio of 100:10:1. Dithranol (1,8-dihydroxy-9(10H)-anthracetone, Aldrich 97%) was used as the matrix. Sodium trifluoroacetate (Aldrich, 98%), silver trifluoroacetate (Aldrich, 99.99%) or lithium trifluoroacetate (Aldrich, 99.8%) were added for ion formation, with sodium trifluoroacetate as the optimal salt for obtaining the highest S/N ratio. ## Monomer synthesis 5-Norbornene-endo,exo-2,3-dicarboxylic acid dimethylester, monomer A 7 was synthesised according to reference , 5-norbornene-endo, exo-2,3-dicarboxylic acid bis(1-oxyl-2,2,6,6-tetramethyl-piperidin-4-yl) ester, monomer T 9, was prepared according to references . ## Synthesis of homopolymers A 15 and T 20 Monomer A 7 (50.0 mg, 0.23 mmol) dissolved in 1 ml of CH 2 Cl 2 was introduced into a heated and argon flushed glass tube equipped with a magnetic stirring bar. A solution of catalyst U3 (11.8 mg, 0.015 mmol) dissolved in 1 ml of CH 2 Cl 2 was then added. After 5 min of stirring at room temperature, the total consumption of monomer A 7 was confirmed by thin layer chromatography (TLC). The reaction was then quenched with 5 drops of cold ethyl vinyl ether, and the resulting polymer purified by column chromatography (SiO 2 ). The homo-polymerization of monomer T 9 was carried out in the same manner with catalyst G2. Homopolymers (A n ) with different chain lengths (n = 15, 50, 25) with the catalysts G3, U1, U2 and U3 as initiators were also synthesized using the same procedure by adopting the required polymerization times. ## Block copolymer syntheses The synthesis of block copolymers (A n -b-T n ) was carried out analogously to methods developed previously in our laboratory . For example the synthesis of BCP-A 50 T 50 was performed by sequential addition of monomers. Monomer A 7 (15 mg, 0.071 mmol) dissolved in 1 ml of CH 2 Cl 2 was introduced into a heated and argon flushed glass tube equipped with a magnetic stirring bar. To this solution, catalyst G3 (1.26 mg, 0.0014 mmol) dissolved in 1 ml of CH 2 Cl 2 was then added. The mixture was allowed to stir at room temperature for 1 h until all of the monomer A 7 was consumed as confirmed by GPC and TLC. Subsequently, monomer T 9 (35 mg, 0.071 mmol) dissolved in 1 ml of CH 2 Cl 2 was then added to the above reaction mixture and stirred for 2 h at room temperature until all of monomer T 9 was consumed, as confirmed by GPC and TLC. The polymerization was quenched by the addition of cold ethyl vinyl ether. The polymer was isolated by column chromatography (SiO 2 ) (eluent: DCM). ## Kinetic experiments A pyrene stock solution was prepared from 70 mg of pyrene dissolved in 2 ml of CDCl 3 . Monomer A 7 (20.83 mg, 0.099 mmol) dissolved in CDCl 3 (0.2 ml) was first introduced into the NMR tube and then the pyrene stock solution (0.2 ml) was added. Before adding the initiator solution, the ratio of the monomer to the internal standard was determined by NMR. On the basis of this value, the monomer concentration at t = 0 was determined. A solution of the catalyst U3 (1.48 mg, 0.0019 mmol), dissolved in CDCl 3 (0.2 ml) (in case of catalyst U1 (1.83 mg, 0.0019 mmol)) dissolved in CDCl 3 (0.2 ml) was added via a syringe to yield the desired monomer to initiator ratio. After shaking, the tube was inserted into the NMR spectrometer, and the decrease in the monomer with respect to time was monitored by integrating the resonance peaks at 6.27 and 6.07 ppm. For determination of the monomer concentration at t = 0 and the monomer consumption, the corresponding signals at 6.27 and 6.07 ppm from monomer A 7 compared with the one at 8.20 ppm from the internal standard pyrene were integrated. The time between the addition of the initiator solution and the first measurement was added to the first measuring point.

chemsum

{"title": "Synthesis and crossover reaction of TEMPO containing block copolymer via ROMP", "journal": "Beilstein"}

oxidative_coupling_for_facile,_stable_carbon_modification_with_dna_and_proteins

## Abstract: Modification of electrodes with biomolecules is an essential first step for the development of biosensors. Conventionally, gold electrodes are used because of their ease of modification with thiolated biomolecules. However, carbon screen-printed electrodes (SPEs) are gaining popularity for the development of cost-effective platforms, as they do not require precious metals for the working electrode and are more consistent than most equivalent gold screen-printed electrodes. However, their universal modification with biomolecules remains a challenge; the majority of work to-date relies on non-specific amide bond formation to chemical handles on the electrode surface. By combining facile and consistent electrochemical modification to add an aniline handle to electrodes with a specific and biocompatible bioorthogonal oxidative coupling reaction, we can attach DNA and proteins to carbon electrodes. Importantly, for both proteins and DNA, the biomolecules maintain activity following coupling, ensuring that the DNA is in a biologically-relevant conformation and that proteins remain folded following coupling. We have further demonstrated the ability to generate self-assembled whole-cell films on these electrodes using DNA-directed immobilization to DNA-modified electrodes. This work provides an important platform for the modification of inexpensive carbon SPEs and is anticipated to be applicable to any carbon-based electrode material. ## Introduction Diagnostic testing has greatly improved the quality of health care by providing physicians with significant amounts of previously unavailable information. However, most diagnostics need to be performed in centralized laboratories with trained personnel and costly equipment, which can lead to lengthy time-to-results and limits their utility in many cases. Point-of-care diagnostics, in contrast, circumvent these issues; these technologies are, by definition, compatible with on-site measurements. For the development of effective diagnostic tools against novel targets, sensitive and specific detection of biomarkers is essential. 5 Electrochemical biosensing platforms have the unique ability to convert biological events, including substrate binding to proteins, enzymatic activity, and nucleic acid hybridization, directly into quantitative electronic signals. 6 However, commercialization of these technologies has remained limited. 7 We attribute the limited translation of electrochemical biosensors to limitations in the platforms available for sensing. The majority of electrochemical biosensors are developed on gold substrates due to their ease of modification using any molecule containing a free thiol. 8 Typically, biomolecule-modified electrodes are prepared through self-assembly of thiolated biomolecules on gold to form self-assembled monolayers. Despite the ease of generating these films, this strategy can pose challenges due to key issues with their implementation. First, for effective formation of goldthiol bonds, gold electrodes must be cleaned using harsh methods prior to assembly of molecules on them. Second, gold is a precious metal and can be costly. Third, gold-thiol monolayers are unstable and therefore not suitable for many commercial applications. 8 Finally, self-assembled layers of thiols on gold are frequently inhomogeneous, leading to inconsistencies in readout. 13, Further complications arise when we move from high-quality crystalline gold surfaces to the lower quality gold found in most disposable electrodes. Specifically, gold screen-printed electrodes (SPEs) are notoriously challenging to implement because the additives used in their inks can inhibit thiol-gold bond formation and yield highly inconsistent surfaces. 8 Because of these limitations, and the relatively high cost of gold, carbon SPEs are increasingly used for point-of-use devices. To-date, carbon electrode modification with biomolecules generally relied on either non-specific adsorption or appending a functional group to the carbon surface, followed by non-specific amide bond formation to couple a biomolecule to the electrode. Though these methods have generated carbon surfaces successfully modified with biomolecules, this strategy can be problematic due to the potential diversity of orientations of biomolecules on electrodes, which can decrease the sensitivity and reproducibility of the platforms. Here, we report a stable, efficient bioconjugation reaction to couple proteins and nucleic acids to inexpensive, disposable carbon SPEs with high chemoselectivity. The reaction series begins with electrochemical grafting of a nitrobenzyl moiety to the surface, 23 which is then reduced to generate an aniline as a chemical handle for subsequent bioconjugation. 24 An oxidative coupling 25 strategy using aminophenol-modified biomolecules in the presence of a mild oxidant (potassium ferricyanide) yields consistent surface modification on carbon electrodes. Both singlestranded nucleic acids and proteins were immobilized using this method. The advantages of this generalizable method are emphasized through its myriad applications. Here we focus on three: 1) complementary DNA strand biosensing; 2) controlled immobilization of DNA-modified cells using immobilized nucleic acids; and 3) controlled immobilization of HRP enzymes, which are important components of many bioassays. We anticipate this method to be a facile, low-cost, generalizable method to easily modify carbon surfaces with biomolecules. ## Figure 1. Electrochemical modification of carbon screenprinted electrodes with anilines. (a) A p-nitroaniline is chemically-activated to a diazonium in situ. This molecule is then electrochemically grafted to a carbon surface. The nitrobenzyl group covalently tethered to the carbon is subsequently electrochemically reduced to an aniline. (b) The cyclic voltammogram (CV) of the diazonium grafting to the carbon electrode (step 1). The CV was performed from 0.6 V to -0.2 V vs AgCl/Ag at 100 mV/s for two cycles; the loss of a reductive signal in the second scan confirms modification of the surface. (c) CV of nitrobenzyl reduction to aniline (step 2). The CV was performed -0.3 V to -1.3 V vs AgCl/Ag at 100 mV/s for two cycles. ## Results and Discussion Electrochemical surface modification and biomolecule conjugation. To modify carbon SPEs with biomolecules, we first required a consistent and stable electrode modification strategy to easily generate a chemical handle for bioconjugation on carbon. To accomplish this, we turned to the established electrocoupling between sp2 hybridized carbon on the electrode and diazonium salts to generate a carboncarbon bond (Figure 1a). 23 From p-nitroaniline, diazonium salts were generated in situ using sodium nitrite. Subsequent coupling to the carbon SPE was accomplished electrochemically, which was controllable through cyclic voltammetry (Figure 1b). This method provided consistent surface modification and appeared to minimize polymerization of the nitroaniline based on the observed currents. The combination of cyclic voltammetry rather than chronoamperometry and the application of a nitroaniline rather than p-phenylenediamine minimize random polymerization at the surface. 24 Subsequent reduction of the nitrobenzyl moieties on the surface to generate anilines was also performed electrochemically (Figure 1c). 24 From the cyclic voltammogram of the reduction, we estimate the surface coverage of anilines to be about 2 nanomoles/cm 2 . These aniline-modified electrodes were found to be stable following modification, enabling more flexibility in their preparation than selfassembled monolayers on gold. Biomolecule modification for specific coupling to anilinemodified electrodes. Following modification of the carbon SPEs with anilines, we undertook bioconjugation to these surfaces. We investigated both DNA and protein coupling to the modified electrodes. To enable specificity in the reaction while maintaining mild conditions compatible with biomolecules, we employed an oxidative coupling reaction that occurs between an aminophenol-modified biomolecule and an aniline. Though this reaction has been demonstrated in solution, 25,26 this is its first application to electroactive surfaces. To accomplish this, we first modified the biomolecule (in this case, single-stranded DNA (ssDNA) or horseradish peroxidase (HRP)) with a nitrophenol moiety using established protocols. The nitrophenol was reduced to an aminophenol prior to electrode coupling. Modification of these biomolecules was confirmed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS, Figures S2-S4). These biomolecules were then coupled to SPEs in the presence of a mild oxidant, as described below. Single-stranded DNA (ssDNA) surface coupling and quantification. ssDNA modified with a terminal o-aminophenol was coupled to aniline-modified electrodes by oxidative coupling in the presence of a mild oxidant: potassium ferricyanide. 25 Following reduction of the ssDNA-nitrophenol to ssDNA-aminophenol, size exclusion purification was performed, and a single peak was observed in MALDI-TOF, indicating near-quantitative conversion. Subsequent addition of ssDNA to aniline-modified SPEs was confirmed by hexaammineruthenium (III) chloride (RuHex) DNA quantification. RuHex interacts electrostatically with the DNA backbone, serving as a phosphate counter on the electrode surface. 27 From cyclic voltammograms of DNA-modified electrodes following RuHex treatment, the concentration of DNA can be determined by integrating the current response to determine a total charge on the surface (Figure 2a). 27 Cyclic voltammetry between -0.5 V and 0.1 V was performed at 100 mV/s on an electrode with the o-aminophenol DNA. As controls, we also evaluated an aniline-only electrode without any DNA and with amine-terminated DNA (which we would not expect to specifically couple). The anodic peak resulting from RuHex treatment of o-aminophenol-modified DNA oxidatively coupled to an electrode was significantly larger than the signals from the other surfaces (Figure 2b). We also observed a shift in the potential of the peak from -0.3V on the electrode without DNA (peak A) to -0.4V with o-aminophenol DNA (peak B), indicating surface attachment of DNA to the electrode (Figure 2b). The average charge for each electrode condition was determined from the cyclic voltammogram and subsequently converted to a DNA surface coverage using the following equation: 28 ΓssDNA=(nFAzNA)/(QM) Normalized to account for background RuHex signal, electrodes modified with o-aminophenol DNA had an average surface coverage of 10 pm/cm 2 . For gold electrodes modified with self-assembled DNA monolayers, coverages range from 1 pm/cm 2 to 100 pm/cm 2 depending upon the assembly conditions. Based on the RuHex quantification of DNA on our aniline-modified electrodes, the o-aminophenol DNA successfully coupled to the electrode surface with sufficient efficiency for medium-density DNA coverage. Hybridization of complementary DNA to ssDNA-modified electrodes. Though we confirmed that DNA was coupled to electrode surfaces using RuHex, its attachment to a surface does not indicate that the DNA is in a biologically-relevant, accessible conformation. To establish that the immobilized DNA is accessible for biological assays, we subsequently investigated the ability of complementary DNA to hybridize to the strands immobilized on electrode surfaces. DNA complementary to that on the electrode and non-complementary DNA were titrated onto DNA-modified electrodes. As an additional control, DNA was added to an aniline-only electrode that had not been modified with DNA. Only the electrodes with complementary ssDNA added to them showed a significant increase in DNA surface coverage measured by RuHex, with the higher concentration of ssDNA corresponding to a larger percent increase in the surface coverage (a 23% increase compared to 17%, respectively). Both the off-target DNA added to a DNA- modified electrode and the electrodes with aniline only on their surfaces demonstrated a slight decrease in the surface coverage of 2% and 1%, respectively, which is not statistically significant above the background (Figure 2c). The fact that higher concentrations of complementary ssDNA yielded the greatest increase in surface coverage, and that the controls experienced little change in signal, provides evidence for the accessibility of the DNA on the electrode surface and the potential of these DNA-modified electrodes for biosensing. DNA-directed immobilization for whole-cell capture on electrodes. Many microbes are of interest for bio-electrochemical systems to perform sustainable transformations including wastewater treatment and carbon capture as well as to act as bioengineered electrochemical sensors. 29 However, these species can require significant lengths of time to colonize electrodes and form biofilms. 30 One species in particular, Shewanella oneidensis, is considered a model organism for cells interfacing with electrodes. However, their biofilm growth can take several days on carbon cloth. We previously demonstrated that DNA hybridization between DNAmodified electrodes and DNA-modified cells enabled rapid current generation on gold electrodes. 31 Here, we sought to expand the application of that technology to carbon SPEs. Conventionally, S. oneidensis bioreactors require carbon cloth for current production, which has a high surface area to volume ratio and a significant three-dimensional structure, allowing for better colonization by these cells. 32 Less success has been found with carbon SPEs, but they are important for the development of deployable technologies with these cells. To observe DNA-directed immobilization of S. oneidensis on carbon SPEs, cells were surface-modified with ssDNA using established protocols. 33,34 Briefly, sugars on the surface of the cells were oxidized using a mild chemical oxidant (sodium periodate), followed by addition of hydrazine-DNA. This strategy yields cells that are surface-modified with ssDNA. These DNA-modified cells are then readily immobilized on SPEs modified with complementary DNA (Figure 3a). Cells were incubated with the DNAmodified electrodes for one hour, followed by washing with 1% SDS to remove any cells that were non-specifically bound. Electrodes were prepared with the following conditions: unmodified electrode with unmodified cells added; unmodified electrode with modified cells added; a ssDNAmodified electrode with unmodified cells added; and a ssDNA-modified electrode with DNA-modified cells added. After washing, only the modified electrodes with the modified cells showed cells adhered to the electrode surface (Figure 3b). This attachment provided evidence for the biologically-relevant conformation of the DNA coupled to the electrode surface and further demonstrated the fidelity of the DNA-directed immobilization of S. oneidensis on the carbon SPE. o-Aminophenol-modified horseradish peroxidase (HRP) coupling to electrodes. In addition to DNA, many proteins and enzymes are important for both biosensing and bioenergy applications. Specifically, enzymes capable of catalysis, such as HRP, can be important for both signal amplification of biosensing readout, transformations that enable the removal of reactive oxygen species from a solution, and performing reactions that may be essential to carbon capture and upcycling technologies. HRP in particular can oxidize a substrate and reduce hydrogen peroxide to generate the oxidized substrate and water. Thus, this enzyme has been used extensively for photochemical and electrochemical detection, as well as bioelectrochemical catalytic systems. We therefore sought to evaluate our coupling strategy with such a prevalent protein. We modified this enzyme with an aminophenol using a protocol similar to that used for DNA modification. We subsequently oxidatively coupled this protein to aniline-modified electrodes and characterized the electrodes using atomic force microscopy (AFM) imaging and an HRP activity assay (Figure 4). Atomic force microscopy (AFM) imaging of HRP-modified electrodes. AFM imaging was used to evaluate the nanoscale morphology of the electrode surface. This analytical technique can provide information both on whether the coupling occurred and the homogeneity of the resulting coupling. For AFM imaging, 30 µM o-aminophenol modified HRP was added to the aniline-modified surface. AFM images of an HRP-modified electrode are compared against an unmodified electrode for evidence that the protein is coupled to the electrode. From the images, differences in surface roughness between the modified and unmodified electrodes are apparent. Further, the HRP-modified electrode surface has higher variability in the height of the surface observed, which, when compared with the variations in height of the unmodified electrode, indicates electrode surface modification with HRP (Figure 4b-e). However, this evidence of coupling does not indicate that the enzyme is active after surface immobilization. Thus, we next evaluated the enzymatic activity of HRP following surface coupling. ## Colorimetric activity analysis of HRP. To evaluate the activity of the coupled protein on the electrode, we employed an assay conventionally used to quantify electrochemical enzyme-linked immunoassay (ELISA). Proteins are more sensitive to environmental factors such as temperature and pH fluctuations than DNA as they have a tertiary structure that must be maintained for them to remain active. Thus, we evaluated the activity of the HRP enzyme after coupling to electrodes to determine whether the enzyme immobilization procedure has any impact on its biocatalytic activity. To evaluate this, we performed a colorimetric assay using 3, 3', 5, 5'-Tetramethylbenzidine (TMB) as a substrate for HRP. HRP is a redox-active enzyme that, in the presence of hydrogen peroxide, can oxidize TMB to a colored charge transfer complex (Figure 4f). 40 Therefore, we added TMB and hydrogen peroxide to the HRP-modified electrodes and monitored the absorbance at 652 nm as a readout for the enzyme activity. A higher absorbance, indicating higher enzymatic activity, was observed on the HRP-modified electrodes than the unmodified electrodes. Additionally, the electrodes that were modified with a 30 μM HRP solution had on average an absorbance that was about 55% higher than that of the electrodes that were modified with a 10 μM HRP solution and 670% higher than that of electrodes that were not modified with HRP but were only treated with ferricyanide (Figure 4g). Overall, these results along with the AFM images of the modified electrodes verify that not only did the HRP enzyme successfully couple to the surface, but it maintained a biologically-active conformation. ## Conclusions Biomolecule-modified, carbon SPEs are a promising technology for the development of cost-effective diagnostic and sustainable energy platforms. They can have higher stability than gold-based platforms and are readily generated on flexible substrates. However, their universal and specific modification remains a challenge. We have developed an efficient biomolecule immobilization method that enables coupling of both proteins and single-stranded DNA onto carbon SPEs with high efficiency using mild reagents and reaction conditions. Importantly, we confirmed that both the biomolecules evaluated (ssDNA and HRP enzyme) remain adhered to the electrode surface in a biologically-relevant conformation indicating that our immobilization strategy did not affect their function and activity. We have further demonstrated that this platform can both immobilize biomolecules and capture whole cells on electrodes using DNA-directed immobilization. Taken together, our work highlights a simple and possibly generalizable strategy for the controlled biomolecule and whole-cell modification of carbon SPEs for potential use in point-of-care diagnostics, biosensing, and sustainable energy applications. ## ASSOCIATED CONTENT Supporting Information. Methods and materials, mass spec characterization of conjugated biomolecules.

chemsum

{"title": "Oxidative coupling for facile, stable carbon modification with DNA and proteins", "journal": "ChemRxiv"}

branched_bbb-shuttle_peptides:_chemoselective_modification_of_proteins_to_enhance_blood–brain_barrie

## Abstract: The blood-brain barrier (BBB) hampers the delivery of therapeutic proteins into the brain. BBB-shuttle peptides have been conjugated to therapeutic payloads to increase the permeability of these molecules.However, most BBB-shuttles have several limitations, such as a lack of resistance to proteases and low effectiveness in transporting large biomolecules. We have previously reported on the THRre peptide as a protease-resistant BBB-shuttle that is able to increase the transport of fluorophores and quantum dots in vivo. In this work, we have evaluated the capacity of linear and branched THRre to increase the permeability of proteins in cellular models of the BBB. With this purpose, we have covalently attached peptides with one or two copies of the BBB-shuttle to proteins in order to develop chemically welldefined peptide-protein conjugates. While THRre does not enhance the uptake and transport of a model protein in BBB cellular models, branched THRre peptides displaying two copies of the BBBshuttle result in a 2.6-fold increase. ## Introduction Biological barriers hamper the efficient delivery of therapeutic molecules to their intended target and are thus one of the main challenges in contemporary medicinal chemistry. The bloodbrain barrier (BBB) is a physiological and metabolic barrier resulting from a complex interaction between endothelial cells and several other cell types, such as pericytes and astrocytes (Fig. 1). The protective nature of the BBB is conferred mainly by restrictive cell-to-cell connections called tight junctions. Disorders of the central nervous system (CNS), which include Alzheimer's and Parkinson's diseases, epilepsy, and brain cancers, have a high prevalence in modern society. There is therefore a need to develop new therapeutics to treat the aforementioned disorders. For those drugs that cannot cross the BBB unaided, the alternative is to conjugate them to molecules that do have this capacity and can thus transport them into the brain. Cell-penetrating peptides (CPPs) were developed to increase the intracellular delivery of drugs. 8,9 A large number of CPPs have been described, including TAT, penetratin, SAP and octaarginine, for the delivery of a wide range of cargoes. 10,11 However, these molecules have several limitations, such as lack of selectivity or entrapment in endosomes. 12 These issues can be tackled in a number of ways. 11,13 On the one hand, the introduction of cyclic peptides opened the door to direct drug delivery to the cytosol. 14 On the other hand, CPP dendrimers display enhanced cell-penetrating properties. 15 Reymond's group reported a high multivalent effect for the divalent constructs of TAT, TP10 and Antp. 16 Moreover, the branched dimerisation of the TAT peptide improved permeation potency in a non-linear manner. 17 CPPs are not limited to intracellular delivery and have also been used to overcome other barriers, such as the gastrointestinal barrier and the BBB. 1,10,18 Although CPPs can enter brain endothelial cells, recent fndings show that the efflux rate of CPPs from the brain is higher than the influx in vivo. 19 Due to their lack of selectivity and transcytosis, CPPs do not easily reach the brain parenchyma, thereby limiting their use as brain delivery vectors. In recent years, a novel approach has been developed using molecular vectors, also known as Trojan horses or BBB-shuttles, to selectively deliver drugs into the CNS. One of the frst peptides reported to overcome the BBB without disrupting it was Angiopep-2, currently the most advanced BBB-shuttle in clinical trials. 23,24 Angiopep-2 belongs to a frst generation of BBB-shuttle peptides which also includes ApoE, 25 THR 26 and TGN. 27 The main flaw of this frst generation was the lability to hydrolysis catalysed by peptidases, thus reducing their half-life in serum and their efficiency in vivo. A second generation of BBB-shuttles was developed to overcome their stability limitations. 28 These improvements included the cyclisation of peptides, 29 the use of unnatural amino acids 30 and the retro-enantio approach. These strategies have proven useful in the feld of BBB-shuttle peptides, with examples such as THRre, 31 D Angiopep-2 32 and D CDX. 33 Here we present the design, synthesis and activity of a third generation of BBB-shuttles, namely branched BBB-shuttle peptides (Fig. 2). Although the second generation of BBBshuttles can fnd applications in the delivery of small molecules and inorganic nanoparticles (NPs), the real challenge is the delivery of proteins and antibodies to the brain. In the following sections, a branched version of the protease-stable peptide THRre is presented. In contrast to a single copy of the BBB-shuttle, the branched BBB-shuttle showed higher uptake and improved the permeability of a model protein in endothelial cells. ## Synthesis of branched THRre peptides The BBB-shuttle THR (THRPPMWSPVWP) is a 12-mer peptide discovered by phage display that interacts with the human transferrin receptor. 26 Although this peptide transports gold NPs across the BBB in vitro and in vivo, 34 it has low stability to serum proteases. The retro-enantio version of THR (THRre, pwvpswmpprht) is stable in serum and is able to transport a variety of cargoes, with higher efficiency than the parent peptide, in an in vitro BBB cellular model and in vivo in mice. 31 Branched peptides can be prepared in various ways. 35 Lys has been used as the branching unit for its easy incorporation using solid-phase peptide synthesis (SPPS). Lys presents two amino functions, at the aand 3-position, which provide two anchoring points. These Lys residues can either be intercalated in the main chain to create a decorated peptide 36 or be placed at the C-terminus to form tree-like constructs. 37,38 Multivalent peptides can be synthesised using stepwise SPPS or conjugation of two peptidic sequences forming the core and branches. Although the latter approach has the advantage of the pre-purifcation and characterisation of the separate peptides, the introduction of several chains into the core may be hampered by steric and hydrophobic interactions. 38 Therefore, here we designed a stepwise synthesis of the branched THRre BBB-shuttle peptide. We synthesised the peptide with several Fig. 2 Scheme representing three generations of BBB-shuttles. The 1 st generation is composed of peptides with L-amino acids that were first discovered to cross the BBB without disrupting it. The 2 nd generation improves their BBB-shuttle properties by conferring them with higher stability in front of serum proteases. In the 3 rd generation, the branched dimerisation of these BBB-shuttles improves the capacities of these peptides to transport more challenging cargoes. xxxx represents the amino acids corresponding to the core of the branched peptide. modifcations in order to facilitate the conjugation with proteins (Fig. 3 and Scheme 1). Fmoc-Lys(Fmoc)-OH was used as the branching unit, allowing the simultaneous deprotection of the two amino groups. A small spacer was introduced between the resin and the branching Lys in order to increase the solubility of the fnal peptides. ## Synthesis of THRre_2m and THRre_2f One of the Lys residues was orthogonally protected with an Alloc group to allow the introduction of other functionalities, such as a reactive moiety (6-maleimidohexanoic acid, Mal) or a fluorophore (carboxyfluorescein, CF) (Scheme 1a). Two Gly residues were placed in such a way as to flank the Lys(Alloc), thus serving as spacers. In this way, THRre_2m (with a Mal) and THRre_2c (with a CF) were obtained. ## Synthesis of THRre_2c An additional Cys residue was introduced at the C-terminus to provide a thiol as the reactive group (Scheme 1b). To improve the flexibility and solubility of the construct, the spacer 3,6dioxaoctanoic acid (O 2 Oc) was introduced between the branching Lys and the rest of the chain. Similar to the branched BBB-shuttles, linear versions were also synthesised without the introduction of the branching unit thus presenting only one sequence of the BBB-shuttle sequence THRre (Fig. 3, S1 and Table S1 †). ## Evaluation of the internalisation capacity in brain endothelial cells To evaluate the potential of the branched peptides as BBBshuttles, we examined their capacity to internalise a small cargo such as CF into cells. We assessed the amount of internalised peptides in a confluent monolayer of bEnd.3 cells, which is an immortalised mouse brain endothelial cell line that displays similar characteristics to the BBB. 39 In these experiments, the peptides labelled with the fluorophore were incubated with the endothelial cells and, after a period of time, the amount of peptides internalised was measured. The results obtained by flow cytometry show a 20-fold increase in uptake for the CF-labelled peptide containing two copies of THRre (THRre_2f) when compared with the one containing a single copy (THRre_1f) (Fig. 4a). We then performed internalisation studies by confocal microscopy. These experiments were performed on bEnd.3 and also on hCMEC/D3 to show that the peptide is internalised both by murine and by human cells. 40 Again, the amount of peptides internalised was higher for THRre_2f (Fig. 5, S3 and S4 †). In addition, THRre_2f was evaluated in a transcytosis BBB model that has a strong correlation with in vivo brain permeability. 41 This model consisted of human brain capillary endothelial cells, derived from pluripotent stem cells, forming a monolayer on the semi-permeable membrane of a transwell and co-cultured with bovine pericytes. The apparent permeability measured was (5.9 AE 0.8) 10 6 cm s 1 . These results show that the presence of two copies of the BBBshuttle enhances the uptake of these peptides by endothelial cells. They also point towards a multivalent effect, where uptake is increased in a non-linear fashion with the number of copies of the peptide. Thus this platform could greatly facilitate the transport of proteins that are not able to reach the brain unaided. ## Chemoselective conjugation to model proteins Site-specifc modifcation of proteins at a single residue is challenging and may not be enough to translocate the protein. Moreover, the introduction of several copies of a BBB-shuttle can provide an ill-defned mixture of chemical entities and may affect the physicochemical properties of the protein. Branched peptides provide a chance to incorporate several BBB-shuttles at a single point, thus facilitating the conjugation, enhancing the BBB permeability and minimising impact on the protein properties and activity. The presence of more than one copy of the peptide in a nearby space would also increase the avidity of the peptide, resulting in a multivalent effect. 17 Green fluorescent protein (GFP) was selected as a model for its ease of production and fluorescence properties. Moreover, GFP is not able to cross the BBB unaided, as is the case for most therapeutic proteins. 42 The GFP used in these experiments was expressed with an extra Cys residue to allow selective conjugation to the BBB-shuttles. In this way, the conjugates will have a chemically well-defned structure, be easy to characterise and maintain the original activity of the protein (Fig. S5 †). Two types of conjugates were envisaged, either with a thioether linkage or a disulphide bridge (Scheme 2). The latter have a linkage that can be reduced in the cytosol and tumour environment, due to the presence of a high concentration of glutathione (GSH). Therefore, disulfde linked conjugates present interesting properties for the delivery of proteins into cells and to solid tumours since the proteins are released in their quasi-native form. 43 To prove if GFP-THRre_1m and GFP-THRre_1c were released in the presence of intracellular levels of GSH and stable at blood levels of GSH the conjugates were incubated in a range of GSH concentrations (Fig. S7 †). At 10 mM, only 50% of GFP-THRre_1c was detected after 20 minutes. By contrast, when the concentration of GSH was lowered to blood levels (2 mM), 100% of GFP-THRre_1c was intact after 1 h. ## Conjugation via a thioether bridge Maleimide-containing peptides (THRre_1m and THRre_2m) were conjugated to GFP by the reaction of the maleimide moiety with the thiol of the Cys residue to form a stable thioether bridge. Following Scheme 2a, GFP was reduced by the addition of TCEP and conjugated without further purifcation with the maleimido-peptide to produce the conjugated product. The conjugation of THRre_2m with the protein did not proceed to completion due to the bulkiness of this peptide construct in comparison with THRre_1m (Table S2, Fig. S2 and S6 †). ## Conjugation via a disulphide bridge The disulphide bridge was achieved using the Cys-containing peptides (THRre_1c and THRre_2c) following Scheme 2b. The protein was reduced and reacted with Ellman's reagent to provide a more reactive intermediate thioester that reacted with the Cys-peptide to afford the fnal conjugated product. As in the previous conjugation, not all the protein reacted with the peptide in the constructs with THRre_2c (Table S2, Fig. S2 and S6 †). In order to assess the importance of having a chemically well-defned structure, we attempted preparing a GFP conjugate with an average of two THRre molecules randomly distributed throughout the surface. However, all conditions tested (Fig. S8 and Table S3 †) resulted in unstable conjugates. These results highlight the relevance of our strategy, which allows the incorporation of more than one BBB-shuttle molecule with a minimal effect on the protein properties. ## Evaluation of the BBB-shuttle properties of the conjugates After conjugation of the peptides to GFP, the BBB-shuttle properties of these constructs were evaluated through internalisation experiments in bEnd.3 cells and transport across an in vitro human BBB cellular model. For these assays, the conjugates were labelled with 125 I so as to facilitate their detection. The radiolabelling allowed using a concentration of the conjugate in the range of those attained by therapeutic proteins in vivo. As negative control we used GFP that had been previously S-alkylated with iodoacetamide in order to prevent the thiol group from reacting in the cellular environment. ## Internalisation experiments The internalisation experiments were performed by incubating bEnd.3 cells with the 125 I-radiolabelled proteins for 30 min. We assayed only the compounds with a thioether linkage because of its stability in a cellular environment. The uptake of GFP increased 2.6-fold when the protein was conjugated to THRre_2m while no increase was detected when conjugated to THRre_1m (Fig. 4b). These results are consistent with those obtained with CF as a cargo. The increase in uptake was exponential with the number of copies of the branched peptide, thereby suggesting that there is a synergistic effect between the peptides that enhances their BBB-shuttle properties. ## Permeability experiments Finally, an in vitro human BBB cellular model was used to further assess the BBB-shuttle properties of these conjugates. In these experiments, the radiolabelled conjugates were placed in the donor compartment and, after a 2 h incubation, the amount that translocated across the endothelial cells and reached the acceptor compartment was measured. For these experiments, we used the conjugates with a disulphide bridge. The conjugates with THRre_2c (Papp ¼ (3.9 AE 0.3) 10 6 cm s 1 ) had a higher Papp than the corresponding protein ((2.77 AE 0.03) 10 6 cm s 1 ) or conjugates with THRre_1c ((2.1 AE 0.3) 10 6 cm s 1 ) (Fig. 4c). Of note, the permeability of the protein did not increase when only one copy of the BBBshuttle was present. These results are along the same lines as those obtained in the uptake experiments, thereby reinforcing the idea that two copies of the BBB-shuttle enhance the permeability of the protein. The high proximity of two THRre molecules in the branched structure makes simultaneous binding to more than one receptor highly improbable. Therefore, the increase in transport originated by the bivalent BBBshuttles should be attributed to an avidity effect. Two vicinal ligands result an increased functional affinity due to a higher local concentration; 44 in the association-dissociation equilibrium established upon binding of one ligand, the close proximity of the other ligand further shifts the equilibrium toward binding. ## Conclusions Here we provide the frst description of the design, synthesis and evaluation of branched BBB-shuttles to be used in the transport of chemically diverse cargoes, such as proteins. In this regard, branched BBB-shuttles were also conjugated to the model protein GFP. Peptide-protein conjugates containing two copies of the BBB-shuttle showed a non-linear increase in uptake. The evaluation in an in vitro human BBB cellular model confrmed that branched BBB-shuttles increase the permeability of GFP while only one copy cannot. Importantly, with this methodology, the conjugates have only one point modifcation, thereby allowing a well-defned chemical structure and the maintenance of the protein properties. This approach could be further applied to increase the efficiency of other BBB-shuttles to allow the delivery of larger and therapeutically relevant cargoes into the brain. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Branched BBB-shuttle peptides: chemoselective modification of proteins to enhance blood\u2013brain barrier transport", "journal": "Royal Society of Chemistry (RSC)"}

techno-economic_analysis_of_atmospheric_water_generation_by_hybrid_nanofluids_to_mitigate_global_wat

## Abstract: Globally, multiple efforts are made to develop active atmospheric water generation (AWG) or atmospheric water extraction (AWE) systems, particularly using direct-air cooling technology to produce water from ambient air. However, this legacy technique is highly energy-intensive, it can only be operated when the local dew point is above the freezing point of water, allows bacteria to grow within the system, and does not scale to create enough water to offer solutions for most industries, services, or agriculture. Liquid desiccant-based AWG methods show promising performance advantages and offer a versatile approach to help address the thermodynamic, health risks, and geographic constraints currently encountered by conventional active AWG systems. In this study, we performed a techno-economic analysis of liquid desiccant-based AWG system with a continuous operating style. An energy balance was performed on a single design point of AWG system configuration while using LiCl liquid desiccant loaded with multi-walled carbon nanotubes (MWCNTs). We showed that the MWCNTs can be doped in LiCl for effective heat transfer during water desorption, resulting in lowering of the sensible heat load by ≈ 49% on the AWG system. We demonstrated that the specific energy consumption (SEC) can currently be obtained as low as 0.67 kWh/gal, while changing the inlet desiccant stream concentration of MWCNTs-doped LiCl at given conditions. While the production cost of water (COW) showed a significant dependency over region, economic analysis revealed that the cost of water can be produced at a minimum selling price of $0.085 per gallon based on the 2021 annual average wholesale electricity cost of $0.125 per kWh in U.S., thereby, providing a strong foundation for future research to meet the desirable and competitive water costs by 2026 but before 2031. ## INTRODUCTION Access to clean healthy drinking water has become a luxury in the current global environment. According to UNESCO/UN-Water 2020, half of the world's population will be living in water-stressed areas by 2050. It was reported that 1.2 billion people worldwide lack access to water, and a total of 2.7 billion experience water scarcity for at least one month of the year. - The increasing burden on the land to feed 7.95 billion people (around 7% of total sum of all humans that have ever lived) is causing increased strain on agricultural lands, leading to productivity loss. , As traditional methods of agriculture are replaced with modern and high output industrial machines, soil fertility diminishes. The unreplenished soil, therefore, requires 30% more water for irrigation. Thus, in the next 10 years, we will be producing less food and using more water. In addition, water shortages are causing friction and tensions where the human dependency on the water is greatest. In many cases, water needs are being exploited to influence government and even leading to conflict and war. Border closures and trade restrictions have limited free movement across continents and entirely isolated communities dependent on imports for their fragile existence. Ongoing issues with climate transformation and resource mismanagement will further exacerbate global water stress and hugely impact an increasing number of residential, commercial, industrial, and agricultural water consumers around the world. Unsurprisingly, children are frequently most affected by the systematic destruction or contamination of potable water; this directly affects their "rights to life, to the enjoyment of the highest attainable standard of health, to an adequate standard of living, education, and dignity." - Atmospheric water generation (AWG) is a nature mimicking (biomimicking) technology inspired by the hydrological cycle which harvests fresh water from ambient air. - Water within the Earth's dynamic atmosphere is continuously moving and changing its physical state of form. It is estimated that Earth's ubiquitous atmosphere contains around 13 sextillion (10 21 ) liters of fresh water at any given time, 98% of which are a vapor, enabling water production from the ambient air a strong pathway to solve global water scarcity. Such water is generally equivalent to 75 times of the fresh water contained in all of earth's lakes, rivers, and streams. Therefore, regions affected by severe drought, trade restrictions, border conflicts, and adverse climate related challenges may consider sustainable AWG technologies which tap into this overhead reservoir of water in Earth's atmosphere. Previous studies reported various AWG systems using both active and passive techniques including fog harvesting, - dewing, direct-air cooling, and desiccant-based methods. - Fog harvesting and dewing, despite being widely studied and practically implemented as conventional technologies over the decades, suffer from several limitations thermodynamically, climatically, and geographically. , On the other hand, several efforts are made to develop active AWG systems, particularly using direct-air cooling technology to effectively produce water from ambient air. This method relies on the forced cooling of bulk ambient air through a refrigerated surface to below dew point to provoke the water condensation to form liquid water. The associated technology is well-studied and commercial systems are currently available from several vendors, but it is also limited by multiple factors. To name a few, direct-air cooling can only operate when the local dew point is above the freezing point of water, which is a caveat in relatively low humidity conditions. Moreover, direct-air cooling fails in arid regions such as the Southwestern United States due to the high energy consumption to produce a unit of water, which was estimated to be between 270 kWh/m 3 and 460 kWh/m 3 . Although several attempts have been made to modify direct-air cooling AWG systems by introducing vapor separation techniques to bring down the energy requirements, such are still limited by their feasibility on a global scale due to their constrained scalability. In addition, according to the U.S. Environmental Protection Agency, some air conditioning-based systems including WaterGen™ have issues with bacteria growth. , Such bacterial growth can lead to water borne illnesses when left unchecked. Desiccant-based AWG systems, , particularly using liquid desiccants , began to gain significant interest owing to its potential for sustainability, system continuity, scalability, and freedom associated with system integration in a wide range of climatic conditions. Although there have been multiple desiccant types including solid desiccants, , and composite desiccants - that are commonly studied for AWG systems, such are limited by hydrophobic nature, poor wettability, structural defects, and bacterial growth which reduces dehumidification and water regeneration performance due to ineffective mass and heat transfer. To date, several hygroscopic liquid desiccants including ionic liquids (ILs), organic, and inorganic desiccants have been utilized in various industrial and laboratory applications. , Organic desiccants such as diethylene glycol and triethylene glycol , were mostly used in air conditioning applications but limited by volatility due to low surface vapor pressure which lead to desiccant losses and equipment contamination. ILs, which are essentially organic salts, often referred to as "target specific" solvents, could serve a wide range of applications due to their structural diversity and highly tunable properties. Inorganic liquid desiccants including halide salts such as LiBr, LiCl, and CaCl2 are prominent for AWG system configurations due to the large water uptake capacity as governed by their low vapor pressure, low cost, low toxicity, low viscosity, eco-friendliness, and ready availability; but these gains frequently come with the penalty of corrosiveness. , While some commercial units of active AWG systems using liquid desiccants are already being developed, the performance of these systems is primarily dependent on the local atmospheric conditions and desiccant media. Particularly, liquid desiccants can be used to first capture water from the ambient air, accounting for 20-65% in immediate energy savings due to vapor separation process. However, for AWG systems that are being operated at large scale, forced convection by blowers is typically used to move ambient air to the liquid desiccant, which does contribute to a small degree to the overall energy consumption. The captured water can be actively regenerated with external electric and/or thermal energy input by heating desiccant solution. Although 80-90% of the total energy is consumed for liquid desiccant regeneration, liquid desiccant technology allows for the recovery of both latent and sensible heats from the liquid desiccant between the scrubber and flash vessel, which generally occurs at different temperatures. The minimum specific energy consumption (SEC) for AWG using desiccant technology if energy recovery is not implemented, is equal to the enthalpy of vaporization of water 694 kWh/m 3 , which is significantly higher when compared to other water production techniques including seawater reverse osmosis (RO) which requires only 3-4 kWh/m 3 . However, little has been reported to date on AWG system level design with approaches to reduce this energy consumption, and likely the cost of water produced with AWG by liquid desiccant. In this work, we provide a novel theoretical framework on AWG system configuration using hybrid nanofluids. We perform a technoeconomic analysis with the intention to reduce the overall energy demand of AWG systems and enable the lower-cost water production from ambient air. The ability to produce low-cost water can translate into supporting all United Nations Sustainability Development Goals. In turn, advancements in this area of technology can help uplift humanity by providing access to affordable water even in water scarce regions or locations that are land-locked or where water transportation infrastructure (i.e., aqueducts, pipelines, water trucks etc.) may not be practical or cost efficient. ## METHODS Description of the system. The proposed liquid-desiccant based AWG system configuration was designed to operate continuously in a closed-loop; the desiccant regeneration requires only lowgrade thermal energy. The system configuration in Figure 1 consists of four major unit operations: (1) a scrubber column that absorbs water vapor from the ambient air, (2) a liquid-liquid heat exchanger (HE1), (3) a flash vessel that facilitates liquid desiccant regeneration by desorbing water, and (4) a two-phase heat exchanger that acts as a condenser (HE2). The scrubber is considered as an adiabatic packed-bed tower filled with a high surface area packing material on which LiCl trickles to maximize the water absorption capacity. In operation, air, at ambient temperature and pressure, and LiCl enter the scrubber column in a counter-flow configuration. In general, liquid desiccants such as LiCl possesses surface vapor pressure as low as 0.36 kPa. As air contacts LiCl in the scrubber column, cooling-free water condensation with almost no energy input happens due to a vapor pressure gradient, unlike direct-air cooling techniques that require external energy input to provoke condensation. In addition, liquid desiccants allow the molecules into the bulk volume, where an initial surface capture happens followed by a subsequent internal permeation process. This mass transfer caused by the vapor pressure gradient will occur until an equilibrium is reached between water in the air and LiCl. Ideally, 100% efficiency in the scrubber column can be achieved when the vapor pressure of water in air at the column outlet is equal to the vapor pressure of the LiCl at the column inlet. The concentration of LiCl drops and becomes diluted by the water absorbed until equilibrium is achieved. Dilute LiCl from the scrubber column that is operated adiabatically must be regenerated with external energy input where the internal water molecules continue to diffuse toward the surface by the concentration gradient. A fraction of dilute LiCl is regenerated, allowing it to initially go through HE1 where it is preheated with returning hot regenerated (rich) LiCl to recover the sensible heat. In the second stage, LiCl from HE1 is passed through a two-phase heat exchanger HE2, that functions as a condenser to recover latent heat of condensation from the returning superheated vapor at saturated pressure exiting the isobaric compressor. The hot LiCl stream from HE2 enters the heater to further elevate the temperature, followed by a flash vessel where the LiCl regeneration happens at sub-atmospheric pressure. Since LiCl has a very low vapor pressure at relatively high temperatures, it does not evaporate with the water, thereby mitigating the liquid desiccant carryover. A final concentration of LiCl is achieved at saturation temperature of water in the flash vessel when all the water content is being desorbed. The rich LiCl desiccant is then recirculated back to the scrubber column through HE1 where majority of the sensible heat is recovered. The precooled LiCl from HE1 is mixed with a split dilute LiCl to help further lower the temperature of inlet LiCl desiccant stream to the scrubber column, allowing it to regain its moisture absorption capacity. Finally, the condensed water from HE2 is collected in a storage tank. Thermodynamic analysis. The AWG system configuration shown in Figure 1 was mathematically defined using an energy balance. To conduct the performance analysis, the following assumptions were made: • All system components are operated at steady-state conditions • Kinetic and potential energy changes are negligible • Pumps and blower energy requirements are negligible • Isothermal compressor is operated at an efficiency of 85% • The coefficient of performance is 2.5 • LiCl leaves the scrubber and flash vessel at saturation state Mathematical modeling. To evaluate the comprehensive system performance, an energy balance was performed on AWG with continuous-style operation using LiCl liquid desiccant, which has been studied previously and the material properties have been well-documented. Besides atmospheric conditions, SEC analysis is also a function of materials properties of the desiccant media, allowing the analysis of AWG system performance by introducing the new dimension of operational conditions. At a given dry bulb temperature and water vapor pressure, it was assumed that the LiCl reaches the saturation state at equilibrium concentration due to vapor pressure gradient with no energy input during the absorption phase. The energy requirement of the blower for the air convection to LiCl is negligible, though it is not always the case in real scenarios, particularly when the system is operated at large scale. During regeneration, the equilibrium LiCl concentration reaches its final desiccant concentration where all the absorbed water has been desorbed to pure water vapor at a saturation state. Typically, regeneration temperatures range between 50 and 260 ºC, which is an energy demanding operation. However, to avoid the higher regeneration temperatures, in this design, the flash vessel is operated at sub-atmospheric pressure. While assuming that the energy contribution of blower is negligible, the energy requirement of the system is primarily attributed by the desiccant regeneration. Thus, the total thermal energy requirement, 𝑄 𝑡𝑜𝑡𝑎𝑙 , to operate the AWG system with no energy recovery is thermodynamically given by the sum of sensible heat, and latent heat of vaporization: Where, 𝑞 𝑠 refers to the sensible heat, which is dependent on the regeneration temperature 𝑇 𝑅𝑒𝑔 , ambient temperature 𝑇 𝑎𝑚𝑏 , specific heat capacity 𝑐 𝑝,𝑑𝑒𝑠 of LiCl with inlet and outlet concentrations of flash vessel to be 𝐶 1 , and 𝐶 2 , respectively. 𝑞 𝑙 corresponds to the latent heat of vaporization. The difference between the mass of LiCl solution entering and exiting the flash vessel is equal to the total mass of water 𝑀 𝐻2𝑂 produced, based on the assumption that mass of LiCl is unchanged and is given by Equations 3 & 4. Since the AWG system is operated continuously and, using liquid desiccants in general, both the latent and sensible heat can be recovered in HE2 and HE1, respectively. Then, the amount of total energy is expressed as: The saturated water vapor exiting the flash vessel is compressed from the regeneration pressure 𝑃 𝑟𝑒𝑔 to the compressor pressure 𝑃 𝑐𝑜𝑚𝑝 , resulting in a pressure ratio 𝑃𝑅 across the compressor. 𝛾 refers to the specific heat ratio. Thus, the work done by the compressor 𝑊 𝐶 with an efficiency 𝜂 𝐶 is given in Equation 6, under the assumption that water may be approximated as an ideal gas at relatively low pressures. From equations 5, 6, & 7, the minimum specific energy consumption 𝑆𝐸𝐶 𝐴𝑊𝐺 to operate the comprehensive active AWG system is expressed as: Where, 𝛽 and 𝛼 corresponds to the recovery factor of latent heat of condensation of superheated vapor and sensible heat during dilute LiCl regeneration, respectively. ∆ℎ 𝑠ℎ𝑣 and ∆ℎ 𝑠𝑐𝑙 are the heat recovered from the sensible heat of both superheated vapor and subcooled liquid leaving HE2. 𝐶𝑂𝑃 is the coefficient of performance. 𝑊 𝑝𝑑𝑝 is the work done by the positive displacement pump for compressing condensate back to the atmospheric pressure since the system is operated continuously at sub-atmospheric pressure, and therefore it must be pressurized prior to the water storage. However, in this case, the energy requirement of pump is assumed to be negligible. ## RESULTS AND DISCUSSION Using the previously described thermodynamic analysis, the minimum specific energy consumption 𝑆𝐸𝐶 𝐴𝑊𝐺 was analyzed for an AWG continuous-style operating system (Figure 1) using LiCl liquid desiccant solution. The parameters stated in Table 1 were used to estimate the 𝑆𝐸𝐶 𝐴𝑊𝐺 values for a system operated with both sensible and latent energy recovery. To minimize the high temperature swing in the flash vessel, LiCl regeneration was carried out at sub-atmospheric pressure while maintaining the saturation temperature in the vessel at 76 ºC. Initially, dilute LiCl at a concentration of 35.1 wt.% was sent to HE1 at 27 ºC, where 70% of the sensible heat was assumed to be recovered in exchange with incoming hot LiCl at 36.7 wt.% from the flash vessel. The resulting water vapor after water desorption with a specific heat capacity of 1.99 kJ kg -1 K -1 at a saturation state was sent to a compressor that is operating with a pressure ratio of 4.0 and 85% compressor efficiency. The enthalpies of superheated vapor ∆ℎ 𝑠ℎ𝑣 and subcooled liquid ∆ℎ 𝑠𝑐𝑙 were considered as 264.35 kJ kg -1 and 37.65 kJ kg -1 , respectively. The exiting LiCl stream from HE1 was sent to HE2, where the latent heat of condensation was assumed to be 70% recovered in exchange with the condensate while assuming that the unit is operated nonadiabatically. Although most of the energy consumption is attributed during water release, liquid desiccants enable the opportunity to recover significant latent and sensible heat through an internal loop, thereby, minimizing the 𝑆𝐸𝐶 𝐴𝑊𝐺 . , Regardless, the current energy trends for liquid desiccant-based AWG systems have been relatively higher when compared with other technologies including freshwater production from liquid water resources (i.e. desalination). However, this energy penalty can be mitigated with controlled choice of liquid desiccants, , , while simultaneously introducing nanomaterials-doped hybrid nanofluids with improved thermal, mechanical, and physical properties. , Based on the classical and statistical mechanics, a detailed model was previously developed , to correlate the specific heat capacity of a nanofluid 𝐶 𝑝,ℎ𝑓 as a function of nanomaterial volume fraction 𝜑 in the liquid desiccant. If we assume that the thermal equilibrium is established between the particles and the surrounding fluid, then the specific heat capacity of a hybrid nanofluid 𝐶 𝑝,ℎ𝑓 is given by: Previous studies investigated the dependency of various nanoparticles including carbon nanotubes, graphene nanoplatelets and alumina nanospheres and their loading concentrations on the specific heat capacities of nanofluids. , Results showed that the specific heat capacities decrease with an increase in volume fraction of nanoparticles due to reduced surface atomic contributions. , In this study, the influence of nanoparticles such as multi-walled carbon nanotubes (MWCNTs) and their loading concentrations in LiCl on the 𝑆𝐸𝐶 𝐴𝑊𝐺 of the system was modeled. The specific heat values of MWCNTs and LiCl are taken as 0.833 kJ kg -1 K -1 and 2.69 kJ kg -1 K -1 , respectively, at the regeneration conditions. , Besides, the density values of MWCNTs and LiCl are recorded to be 1.35 g cm -3 and 1.23 g cm -3 , respectively. By replacing the specific heat values of liquid desiccants with hybrid nanofluids, the 𝑆𝐸𝐶 𝐴𝑊𝐺 of a system operated by hybrid-nanofluids can be stated as: Using parameters in Table 1 and Equation 10, we estimated the 𝑆𝐸𝐶 𝐴𝑊𝐺 values of a system configuration (Figure 1). Table 2 shows the breakdown of AWG system performance at two different liquid desiccant conditions: (1) LiCl and (2) LiCl loaded with MWCNTs. In the case of LiCl desiccant solution, the sensible heat load was estimated to be 342 kJ kgH2O -1 , which has been significantly dropped by ≈ 49%, with the loading of 0.5 vol.% MWCNTs in LiCl. Thereby, the sensible heat load required to heat the same amount of LiCl doped with MWCNTs is reduced during hybrid nanofluid regeneration. While the heat capacity of hybrid nanofluids showing a strong influence on SEC, it is worth noticing that the SEC also depends on viscosity of nanofluids, especially in the case of turbulent flow region which is typical in practical applications. Several studies previously investigated the rheological behavior of water-based nanofluids containing carbon nanotubes. , Results showed that the flow properties of nanofluids doped with MWCNTs is dependent on the temperature and the volume fraction of nanotubes. , At higher concentrations of nanotubes, the shear viscosity of nanofluids showed a pronounced shear thinning behavior as a function of shear rate at 20 ºC. This could be attributed due to the alignment of structural network of MWCNTs under enough shear and, along with their lubrication properties, - resulting in lowering of viscous force, which would greatly improve the performance of thermal management equipment such as heat exchangers in practical applications. Besides loading concentration of nanoparticles in base fluids, SEC also depends on size, shape, and morphologies of nanoparticles. - Regardless, the typical operating conditions of AWG at large scale produce highly turbulent regime where the role of thermophysical properties of hybrid nanofluids is critical. Therefore, the selection of operating temperature and loading concentrations of MWCNTs with controlled size and shape in LiCl liquid desiccant play a vital role and can be rationalized by future research work for comprehensive AWG system improvement. To determine the optimal 𝑆𝐸𝐶 𝐴𝑊𝐺 values, the inlet stream concentration of LiCl to the flash vessel was varied from 30 wt.% to 35.1 wt.%. We compared the results for two cases: (1) LiCl without MWCNTs and (2) LiCl with MWCNTs at a concentration of 0.5 vol.%. For the case of LiCl, results showed that the 𝑆𝐸𝐶 𝐴𝑊𝐺 values gradually increased up to LiCl concentration of 33 wt.% and then, a significant increase in 𝑆𝐸𝐶 𝐴𝑊𝐺 values were observed as it approaches closer to the LiCl saturation concentration in the flash vessel. Whereas in the case of LiCl loaded with 0.5 vol.% MWCNTs, a significant drop in 𝑆𝐸𝐶 𝐴𝑊𝐺 values were observed, especially at higher inlet stream concentrations (≥ 33.3 wt.%) of LiCl, indicating that the presence of MWCNTs in LiCl played a key role in lowering the sensible heat load. An optimum 𝑆𝐸𝐶 𝐴𝑊𝐺 value of 0.67 kWh/gal was recorded at an inlet LiCl concentration of 30 wt.%, indicating that relatively low energy is required if the hybrid nanofluids with a high-water carrying capacity and enhanced heat transfer properties are used. At optimal conditions, we also modeled the AWG system carrying LiCl with MWCNTs at varied concentrations between 0 and 0.5 vol.%. A gradual decrease in 𝑆𝐸𝐶 𝐴𝑊𝐺 was observed with respect to MWCNTs concentration, leading to a maximum drop of ≈ 7.5% at 0.5 vol.% of MWCNTs in 30 wt.% LiCl solution. This behavior could be attributed due to the constrained liquid desiccant layering at the surface of nanoparticle free boundary caused by the changes in the phonon vibration mode at solid-liquid interface. , 11). While wholesale electricity cost is determined by various factors specific to region, the annual average wholesale electricity costs in 2021 (U.S. Energy Information Administration) are recorded over region to determine the 𝐶𝑂𝑊 𝐴𝑊𝐺 values at a given 𝑆𝐸𝐶 𝐴𝑊𝐺 value of 0.67 kWh/gal. In Figure 3a, U.S. recorded the lowest 𝐶𝑂𝑊 𝐴𝑊𝐺 values, while Germany reported the highest. It is important to note that these costs are highly seasonal due to heating and cooling needs, thereby significant changes can be observed in the 𝐶𝑂𝑊 𝐴𝑊𝐺 values at different regions. ## 𝐶𝑂𝑊 𝐴𝑊𝐺 [ Performance of the hybrid nanofluids-based AWG system and off-the-shelf direct-air cooling AWG systems are compared in Figure 3b. For direct-air cooling technology, the energy requirement is used nearly entirely for the processing of bulk ambient air to provoke the condensation, which is a major drawback to effectively scale up these systems. As a result, directair cooling AWG systems will not fit global climatic conditions while meeting the expectations of low energy and cost requirements. Theoretically, liquid desiccant-based AWG has the best SEC value relative to the other commercial AWG systems, due to the special advantages associated with the tunability of liquid desiccants while also recovering sensible and latent heat. A secondlaw analysis , was performed on air-to-water separation system to determine the theoretical minimum work of separation (process agnostic). At the baseline conditions of 80 ºF and 60% RH, the theoretical minimum work of separation required to remove a gallon of water from ambient air was determined under the assumption that the moisture in the air is being captured completely. Based on the analysis performed by CHEMCAD TM modeling, the minimum work of separation was found to be ≈ 0.07 kWh/gal. Figure 3b shows the projection of 𝑆𝐸𝐶 𝐴𝑊𝐺 values towards the theoretical minimum value by 2031 if present research advancements continue based of present trend-lines observed between 2018-2022 (in liquid desiccant air to water capture simulations, proof of concepts, digital twins, and pilot plants). This efficiency trend suggests more optimal system configuration and fluids can be designed and achieved by carefully considering both thermodynamic and economic modeling together. Qualitatively, it is a reasonable hypothesis that the time period required to advance more rapidly than the current trend-line (toward the theoretical minimum SEC values as low as ≈ 0.07 kWh/gal) could be achieved with focused research and development informed by both thermodynamic and economic factors (Figure 3b) illustrated by the dashed-line. ## CONCLUSION In summary, we have performed the techno-economic analysis of liquid desiccant-based atmospheric water generation system with a continuous closed-loop operating style. An energy balance was performed on a single design point parameters of AWG system configuration while using LiCl loaded with MWCNTs. Results show that MWCNTs doped LiCl is an effective liquid desiccant by reducing the sensible heat load by ≈ 49% thus enabling new avenues to test with wide range of nanomaterials for efficient heat transfer applications during water desorption. This method is vastly superior to air conditioning-based AWG systems in terms of efficiency, cost, and ability to scale up. We showed that the specific energy requirements can be obtained as low as 0.67 kWh/gal, while changing the inlet desiccant stream concentration of MWCNTs-doped LiCl at saturation conditions. While the production cost of water shows a significant dependency per region, economic analysis revealed that the cost of water can be produced at a minimum selling price of $0.085 per gallon based on 2021 average annual wholesale electricity cost of $0.125 per kWh in U.S., thereby, providing a strong foundation for future work to meet the desired water production costs not dependent on legacy water sources during or before the next decade.

chemsum

{"title": "Techno-economic analysis of atmospheric water generation by hybrid nanofluids to mitigate global water scarcity", "journal": "ChemRxiv"}

sequential,_all-bioorthogonal_reaction_cascade_catalyzed_by_a_dual_functional_artificial_metalloenzy

## Abstract: Spatial control over chemical reactions is a ubiquitous feature shared by all living organisms. In the present article, we describe the design of a proteinaceous, synthetic capsid based on the well-described bacterial nanocompartment encapsulin.Our engineered virus-like particle carries a dual functional artificial metalloenzyme (ArM) as non-native guest protein. This ArM, created from a fusion protein of HaloTag and monomeric rhizavidin, serves as catalyst for a fully bioorthogonal, linear, two-step reaction cascade. A ruthenium-catalyzed alloc deprotection is followed by a gold-catalyzed, ring-closing hydroamination reaction leading to indoles and phenanthridines with up to 67 % overall yield in aqueous solutions. ## Introduction Compartmentalization enables simultaneous but orthogonal chemical processes, while providing unique selectivities and reactivities e.g. by establishing local concentration gradients. Therefore, researchers from various fields are developing artificial reaction vessels. 1-3 A platform with increasing recognition amongst synthetic biologists and biological chemists are encapsulins (Encs). These protein-based capsids host a variety of natural guest enzymes and are widespread throughout bacterial and archaeal phyla. 4 Previously, we described the application of a nanoreactor for bioorthogonal pro-fluorophore activation inside living cells using encapsulin from Mycobacterium smegmatis (M. smegmatis). 5 The smallest representatives of Encs known to date consist of 60 protein monomers, which self-assemble in a pentagon dodecahedron-like structure with a sphere diameter of 24 nm. Small molecule reactants can pass the proteinaceous shell via pores lying in the symmetry axes of the dodecahedral structure. 6 During assembly, these nanometer sized virus-like particles recognize their guest proteins via a short conserved peptide sequence found at their C-terminus (ELS: encapsulin localization sequence). 7 During the last years, activity of various (non-) natural guest enzymes inside these capsids was investigated. Figure 1: The present research work is highlighted as a merging concept of several previously reported approaches. Moreover, future perspectives in ArM and nanoreactor research are shown. 5, Here, we describe the first proteinaceous nanoreactor catalyzing a two-step, bimetallic, sequential, allbioorthogonal reaction cascade. In this context, Figure 1 shows previous research, the present work and potential developing goals being addressed. For stable immobilization of two orthogonal transition metal (TM) complexes inside Enc, we designed a dual functional fusion protein based on HaloTag (Halo) 5,16 and monomeric Biotin-(strept)avidin. 17 Several monomeric analogues of avidin-type proteins have been described, but only one report of a monomeric avidin based ArM exists. In the present work, two different guest proteins with either monomeric streptavidin (mSA) 22 or circular permuted monomeric avidin (cpMA) 21 fused to HaloTag were investigated. The first step of the presented reaction cascade involves the alloc-deprotection of a secondary amine catalyzed by a ruthenium catalyst (RuH, see Figure 2) covalently bound to HaloTag. 5 The deprotected amine serves as nucleophile in a subsequent gold-catalyzed hydroamination reaction at the monomeric avidin analogue. 23 This reaction is catalyzed by a Buchwald-type Au(I)-phosphine complex (AuB, see Figure 2), that we introduce here as a novel bioorthogonal catalyst. Vidal et al. 24 already studied other Au(I)-phosphine complexes in bioorthogonal catalysis, whereas Wang et al. 25 described (1,1'-biphenyl)-2-ylphosphines as privileged ligands in gold catalyzed nucleophilic attacks of alkynes. We identified hydroamination reactions as suitable second reaction step based on previously published fluorescent gold-anion sensors. 26,27 Recently, two other ArMs catalyzing hydroamination reactions have been reported, both utilizing gold-N-heterocyclic carbene (Gold-NHC) complexes. Chang et al. describe the synthesis of phenantridinium core structures by an albumin ArM. 14 The streptavidin-ArM published by Vornholt et al. catalyzes indole formation from 2-ethynylaniline. 23 Our catalyzed reaction cascade combines both major mechanisms of (bio)-precursor prodrug activation (see Figure 2). A phase I bond cleavage step is followed by a phase II bond forming step. 28 Products of the nanoreactor described here are functionalized, heterocyclic amines with various reported biological activities. As Sravanthi et al. highlight, "among the indole class of compounds, 2-arylindoles appear to be a most promising lead for drug development." 29,30 Therefore, in our study, we focus on the synthesis of functionalized indoles with reported antibacterial 31 or NorA efflux pump inhibitory 32 properties. Furthermore, potential products of our reaction sequence are phenanthridines and phenantridinium cations. These are core structures of multiple DNA intercalating substances with established applications as molecular probes and anti-tumor reagents. 14,33 Overall, the successful implementation of bimetallic catalysis in a confined and programmable nanoreactor opens significant opportunities for reactivity design. ## Results and Discussion This section is split into four parts each discussing one major step in the design of the nanoreactor. First, we evaluate the catalytic activity of the TM-complexes in protein-free, aqueous solutions. The biotechnological work of designing and expressing the ArM-scaffolds as well as the encapsulin nanocompartments is described next. Subsequently, we combine both TM-complexes and protein scaffolds to fully assemble ArMs. Finally, the catalytic activity of the nanoreactors is studied regarding efficiency and scope. ## TM-Catalysis in Aqueous Solutions Figure 2 provides an overview of the reaction cascade towards both hetero-aromatic core motives (indole and phenantridinium). Furthermore, the applied catalysts, as well as their functionalized ligands are shown (for synthetic procedures please see supporting information). By studying the two-step conversion of the starting material SMe5, catalytic properties of the TM-catalysts in aqueous and proteinaceous solutions were analyzed (see Table 1). First experiments prove the catalytic activity and selectivity of the catalysts for the selected reactions (see Table 1, entry 1 to 3). SMe5 was converted to IMe5 (see Figure 2) in 60 % yield by RuH and to PMe5 in 80% yield by addition of both RuH and AuB. During Ru catalyzed deprotection, the literature known side-reaction of amineallylation was observed leading to a significant deviation between conversion of starting material and product yield. 34 Importantly, the amount of allylation product is reduced during cascade reactions, as the reactive amine is consumed during the subsequent hydroamination step (see Table 1, entry 3 and Supporting Figure S1). ## Design and Expression of ArM-Scaffolds With two compatible TM-complexes in hand, a protein scaffold for the dual functional ArM was developed. Therefore, Halo was fused to either mSA or cpMA via an enterokinase site as linker (see Figure 3 A and Figure S2). At the C-terminus a 14 amino acids ELS was fused to both proteins. Halo-mSA and Halo-cpMA were coproduced with His-tagged Enc in a two-vector system in E.coli. After immobilized metal affinity chromatography (IMAC) and size exclusion chromatography (SEC), the nanocompartments Enc::Halo-mSA and Enc::Halo-cpMA were characterized by polyacrylamide gel electrophoresis (PAGE, see Figure 3, S3 and S4). Production and purification of the guest proteins alone without encapsulin was not possible, most probably because the majority of guest proteins were not produced in soluble form. For Enc::Halo-mSA, only a single protein band corresponding to 30 kDa, the expected size of one Enc monomer, was observed under denaturing-PAGE conditions (see Figure 3 B). Potentially, effective encapsulation of Halo-mSA into Enc during assembly was not possible due to the reportedly low solubility of mSA and its fusion proteins leading to a low cytosolic concentration of Halo-mSA. Alternatively, the encapsulation peptide was not accessible for Halo-mSA. 22 Here, the pore profile for EncbpTC was calculated. However, for the Enc::Halo-cpMA construct, encapsulation did take place as an additional, second protein band of the guest protein (53 kDa, Halo-cpMA) was visible in PAGE (see Figure 3 B). Thus, only Halo-cpMA was further investigated for nanoreactor development. Subsequent experiments with Enc::Halo-cpMA showed that the high abundance of His-Tags outside the encapsulin shell interfered with catalytic activity. Therefore, a Tevcleavage site for removing the C-terminal His-Tag was introduced (EncTC, encapsulin TEV cleaved). Furthermore, a previously described mutation of the 5-fold encapsulin pore was reproduced in order to study the influence of the pore environment on substrate transport across the proteinaceous shell (EncbpTC, encapsulin big pore, TEV cleaved). 36 The mutant shows increased pore diameter as well as an altered polarity and charge distribution (wild type: 7 , mutant: 14 , see Figure 3 D and F, measured with moleonline 35 ). After expression and purification EncTC::Halo-cpMA and EncbpTC::Halo-cpMA were fully characterized including scanning transmission electron microscopy (STEM) and dynamic light scattering (DLS). Both analyses supported the prevalence of ~25 nm-sized, spherical particles (for detailed discussion on DLS and STEM results see supporting information and Figure S5 and S6). ## Bioconjugation A crucial step in ArM synthesis is targeted and exhaustive metal to protein binding as well as subsequent removal of unbound components. 5 For establishing a labeling strategy of the dual functional guest protein inside encapsulin, we developed a pulse chase experiment. It includes the TM-complexes RuH and AuB besides the fluorescent Halo ligand AlexaFluor660 (AF660L, λexc=668 nm, λem=698 nm) and the avidin ligand, fluorescein-4-biotin (f4b, λexc=498 nm, λem=517 nm). In the experiment, EncTC::Halo-cpMA was incubated with the TMcomplexes. Subsequently, the fluorescent dyes were added and after a second incubation period, the samples were analyzed by native PAGE. During native PAGE the labeled construct remains intact. As all binding sites at Halo-cpMA were occupied, labeling with the fluorescent dyes was not possible. No fluorescence could be observed for the respective bands in native PAGE (see Figure 4 B, lane 2; Figure 4 C, lane 2; Figure S7). In control experiments where no TM complexes were added, the guest protein could be labeled with the fluorescent dyes. Native PAGE showed the expected high molecular weight bands for the fluorescently labeled nanocompartment (see Figure 4 B, lane 1 and Figure 4 C, lane 1). This experiment proves diffusion of the fluorescent dyes as well as TM-complexes across the proteinaceous shell, activity of Halo-cpMA fusion protein inside encapsulin and noncompetitive binding of cpMA to the biotinylated TM-complex (for detailed procedures see methods text and Figure S7). For exhaustive labeling, an excess of TM-complexes (20 mol eq. RuH) was necessary. Previously described procedures for removing non-specifically bound TM-complexes applying dialysis and spin-concentration are tedious with significant hands-on time required, especially if performed under oxygenfree conditions. 5 Therefore, we developed a novel chromatography-based method (see Figure 5). After labeling of the nanocompartment with the necessary excess of TM-complexes, a saturated solution of BSA was added to the sample, providing non-specific binding sites for the unreacted TM-complexes. 39 Subsequently, BSA and BSAbound TM-complexes were removed by SEC. We verified our purification protocol in a control experiment, where Enc::GFP displaying no binding sites for RuH and AuB was incubated with the TM-complexes. Then, after purification by addition of BSA and subsequent SEC, we added the catalyst substrate SMe5 and the remaining catalytic activity was determined. In this control experiment, no conversion of SMe5 was observed for the optimized purification protocol, underlining its efficiency (see Table 2, entry 4). In consequence, the assembled and purified nanoreactors were applied for catalytic activity screens studying the two-step conversion of SMe5 to PMe5 via IMe5. ## Biocatalysis-Nanoreactor screening The assembly of the catalytically active nanoreactor takes 26 h enforcing high demands on TM-complex stability. To simulate such conditions RuH and AuB were pre-incubated in aqueous, protein-free buffer for 26 h prior to substrate (SMe5) addition. This led to a tremendous decrease in catalytic activity from 99% conversion to 11% conversion only (Table 2, entry 1). By applying inert-gas conditions (N2 atmosphere to avoid catalyst oxidation), this loss in activity was reduced, albeit not fully avoided (37% conversion, see Table 2, entry 2 and 3). Thus, unless stated otherwise, the labeling and purification protocol was performed under exclusion of air. EncTC::Halo-cpMA*RuH*AuB with both TM-complexes distinctively bound to Halo-cpMA inside EncTC catalyzed the outlined reaction cascade in 19% yield over two steps (see Table 2, entry 6). Thus, a functional nanoreactor was designed allowing substrates and products of the reaction cascade to pass its proteinaceous shell. Additional information about the influence of this third coordination sphere on diffusion and catalytic activity can be obtained from considering the result obtained with the pore-mutated nanoreactor EncbpTC::Halo-cpMA*RuH*AuB. The yields obtained during catalysis were generally lower for EncbpTC::Halo-cpMA*RuH*AuB as compared to EncTC::Halo-cpMA*RuH*AuB with wild-type pore. For SMe5 the conversion was reduced from 36% in the system with wild-type pore to 22% conversion in the system with mutated pore (Table 2, entry 6 and 7). The mutated pore not only shows increased diameter (from 7 to 14 ) but also inverted surface polarity from positive in the wild type to negative in the mutated form. The mutation also significantly reduced the lipophilicity of the pore from logD = -0.08 for the wild type to logD = -1.63 for the mutant (see Figure 3 C-F). LogD is defined as a distribution constant between octanol and water, and serves as a measure of the relationship between lipophilicity and hydrophilicity of a substance. 40 The observed loss of activity in EncbpTC::Halo-cpMA*RuH*AuB could further be explained by increased diffusion of residual oxygen through the pores and thus increased decomposition of the catalyst. Furthermore, the reduced catalytic activity could be the result of reduced diffusion of substrates through the protein shell. If a significant steric hindrance on diffusion of the substrates into the compartments exists, an increased pore size should have led to increased catalytic activity. Thus, decreased substrate diffusion would most probably be due to the altered pore charge and lipophilicity. Considering the lipophilic character of the aromatic reaction partners this assumption seems reasonable. Another explanation for the reduced catalytic activity is the increased diffusion of substrates out of the compartment, suggesting an accumulation of SMe5 inside EncTC::Halo-cpMA*RuH*AuB. Here both pore size or pore charge might be causative. As not only the electrostatic potential of the pore itself but also its surrounding area was altered by the mutation, decreased catalytic activity could also be due to interactions between the substrates and the inner surface of EncbpTC leading to reduced availability of the substrates for catalysis. Under aqueous, protein-free conditions, AuB catalyzes quantitative conversion of IMe5 to PMe5 (see Table 1, entry 3). In contrast, for the nanoreactors, approximately only half of the intermediate IMe5 being formed is converted to the final product PMe5 (see Table 2, entry 5-9). A direct explanation for the reduced catalytic activity of cpMA*AuB is inhibition by the protein scaffold. Similar findings are frequently observed for non-optimized (avidin-type) ArMs usually indicating an influence of the protein on the transition state of the reaction. 23,41 As cpMA could not be purified in a non-encapsulated fashion, verifying this hypothesis was not possible. Other possible explanations are non-specific binding of IMe5 to the inner surface of encapsulin, thereby preventing diffusion to the catalytic center of cpMA or substrate inhibition due to an increased local concentration of IMe5 inside the nanoreactors. However, experiments with up to 25-fold increased amounts of IMe5 still showed linear relationships between substrate and product concentrations (see supporting information, Figure S8). Similar inhibition was not observed for Halo*RuH. This is in accordance with known properties of HaloTag based ArMs providing little secondary sphere interactions, minor influences on the transition state and small effects of the protein on the catalytic activity of the complex. 42 In accordance with previous observations, we noticed reduced catalytic activity of the nanoreactor in the presence of oxygen (from 19 % to 6% product yield, see Table 2, entry 5 and 6). However, we could successfully enhance the reaction yield upon increasing the sample volume by four and additionally preparing the nanoreactor in degassed buffer. The subsequent catalytic reaction was taking place in the presence of air, yielding 88% conversion of SMe5 (see Table 2, entry 8). We explained this finding by a relative decrease in oxygen dependent catalyst decomposition. As inert handling is more reliable at increased sample sizes, we were able to further enhance catalytic activity when setting up the reaction at fourfold increased batch size and air-exclusion (see Table 2, entry 9). Indeed, under these optimized conditions we obtained 90 % substrate conversion, yielding 23 % IMe5 and 67 % PMe5 after two steps. Catalytic parameters were determined for both active sites. Obtained turn-over frequencies (TOF) of 1.2 (for Halo*RuH) and 0.6 (for cpMA*AuB) resemble those of similar, non-optimized ArMs. 43,44 With a value of 400 the turn-over-number (TON) for cpMA*AuB is especially remarkable (TON (Halo*RuB) > 60). For details on TON and TOF determination see methods text and supporting information Figure S8 and S9. ## Phenantridine/phenantridinium precursors are shown in blue. All conditions contained catalyst (1 µM) and substrate (20 µM). Reactions were carried out in aq. HEPES-buffer (50 mM, pH=7.5) Yields are displayed using a color code from red to green, as indicated. Experiments marked with a * were performed on large scale (400 µL), others were screened with 100 µL sample size. ## Substrate scope of the nanoreactor In a next set of experiments, we screened various indole and phenantridinium precursors as substrates for the nanoreactors EncTC::Halo-cpMA*RuH*AuB and EncbpTC::Halo-cpMA*RuH*AuB as well as RuH, AuB as control (see Figure 6 and supporting information for synthetic procedures of substrates, intermediates and products). Only selected substrates were investigated with the increased sample size (i.e. 100 µL vs. 400 µL) due to the huge consumption of nanoreactors. The first reaction step catalyzed by RuH yielded high conversion rates for all substrates except S5 (<10% conversion) and S6 (<5% conversion), where very low yields were observed. This observation is in accordance to literature, describing Ru-complexes as promiscuous with regards to substrate scope. 15,23,45,46 Notably, we report the first examples of RuH-type catalyzed alloc-deprotection of secondary amines. In contrast to RuH, AuB catalyzed reactions showed high substrate dependency. In general, the intramolecular annulation reaction is favored for electron-rich (I2, I3) over electron-poor substrates (I4). Furthermore, the annulation reaction is enhanced when the electron donating substituent is located para to the amine rather than para to the alkyne at the distal C13 atom. N-Methylation enhances hydroamination for all substrates tested. Thus, nucleophilicity of the reactive amine is one key factor for the desired reactivity. 25,47 In general, the phenantridinium precursor SMe5 was the most efficient substrate. ## Conclusion In summary, this study provides various innovations in the field of ArM design. The herein presented nanoreactor is the first report of a two-step, dual-functional ArM driven, fully bioorthogonal reaction cascade inside a proteinaceous capsid opening up exciting possibilities in the field of designing artificial organelles with compartmentalized reaction pathways. 8,9, The novel gold-phosphine complex AuB is established herein for both bioorthogonal catalysis in aqueous solutions and as an avidin cofactor. With a TON above 400 its catalytic activity is comparable to the previously described gold-NHC complexes. 14 The monomeric rhizavidin analogue cpMA 21 shows noncompetitive binding to biotinylated TM-complexes and is a suitable alternative to previously described avidin derivatives. 51 As a fusion protein with HaloTag, cpMA is suitable for designing dual functional ArMs, even though it does not express in soluble form. However, encapsulin can stabilize the fusion protein facilitating convenient isolation and application, highlighting the tremendous potential of this approach. The pore engineering detailed herein, indicates, that influencing substrate selectivities is possible. Recent investigations in our group focus on studying the tertiary sphere effects provided by the encapsulin shell. The possibilities offered by encapsulins go far beyond pore mutations. They include further surface functionalizations, e.g. for targeted delivery of the nanoreactors. Along these lines, we believe that the prospects of encapsulin engineering are ideally suited for achieving future applications shown in Figure 1, such as targeted prodrug activation. ## STEM imaging: For sample preparation, a carbon coated Cu-grid was floated on top of a droplet of protein solution (30 µL, 0.5 mg/mL) for 10 sec. Afterwards, excess of solvent was removed and the grid was floated on top of a droplet of aqueous uranyl acetate (UA) solution (30 µL, 1% UA) for another 10 sec. After removing the excess of solvent, the samples were stored until image acquisition. STEM images were recorded with a Hitachi SU8200 cold field emission gun high resolution scanning electron microscope (FEG-HRSEM) operated at 30.0 kV and equipped with a transmission electron detector (see supporting information Figure S5). ## Dynamic light scattering (DLS): DLS measurements were performed on a Zetasizer NanoZS, Malvern Instruments, UK. Measurements were performed with a protein concentration of 1 mg/mL in 50 mM HEPES, pH=7.5 at 25 °C. Measurements were performed with the following set up: attenuator: 8; mean count rate (kcps): 260-420. The data were analyzed and presented with the Zetasizer Nanoseries software (Malvern Instruments, Malvern, UK) using the general purpose analysis model. The intensity-weighted mean hydrodynamic diameter for each measurement was calculated and reported (see supporting information Figure S6). ## Bioconjugation-Pulse chase experiment EncTC::Halo-cpMA or EncbpTC::Halo-cpMA (15 µL, 10 µM, 18 mg/mL) was incubated (4 °C, overnight) with the TM-complexes (0-28 eq., 10 mM stock in DMSO). Afterwards, the fluorescent ligands AlexaFluor660Ligand (0.2 µL, 80 µM stock in DMSO, Promega) and fluorescein-4-biotin (0.2 µL, 80 µM stock in DMSO, Sigma) were added. After a second incubation step (37 °C, 2 h), the samples were analyzed by native page. When catalyst labeling was successful and quantitative, no binding of the fluorescent substrates to the proteins should be observable (see supporting information Figure S7). ## Bioconjugation-Purification protocol Unless stated otherwise all buffers and sample solutions applied during this protocol were degassed by several cycles of applying vacuum and flushing with argon. The metalloenzymes were prepared by incubating the purified protein (100 µL, 11 mg/mL) with RuH (0.5 µL, 50 mM stock solution in DMSO, 20 mol eq.) and AuB (0.3 µL, 50 mM stock solution in DMSO, 12 mol eq.) for 8 h at 4 °C under mild shaking (200 rpm). Afterwards, a concentrated solution of BSA in aq buffer (0.6 mL, 50 mg/mL BSA, 50 mM HEPES, pH=7.5) was added and the solution again incubated under mild shaking at 4 °C overnight. The assembled nanoreactor was purified by SEC. ## Biocatalysis The obtained target nanoreactor was concentrated to 1 mg/mL (10 kDa cut-off spin-concentrators, Sartorius). Reactions were performed in HPLC-vials with a total volume of 100 µL (400 µL for selected conditions) under inert conditions (unless stated otherwise) by addition of substrate (0.1 µL, 20 mM stock in DMSO). The samples were incubated at RT under mild shaking (200 rpm). The sample was transferred to a centrifuge vial (1.5 mL, Greiner Bio-One) and the protein was removed by addition of cold acetonitrile (3 eq. V:V, incubation at -20 °C, 10 min) and subsequent removal of precipitate by centrifugation. The supernatant was analyzed by HPLC. ## Determination of TON For determining, the maximum number of catalytic cycles per reactive center either SMe5 or IMe5 (20 mM stock in DMSO) were added in different concentrations of up to 500 µM to the assembled nanoreactor (100 µL, 1 µM, 1 mg/mL in 50 mM HEPES pH=7.5). For higher concentrations the limit of solubility was reached. After 48 h no further conversion was observed by HPLC. By dividing the total amount of product by the amount of nanoreactors, the TON was calculated. Hamburg, Germany). Fluorescent-gel images were taken on a UVP chemstudio imager (Analytic Jena, Jena, Germany). For PAGE, Mini-Protean Precast Gels, Powerpacs and Tetracells (Bio-RAD, Germany) were applied. Details on procedures, plasmids, primers e.g. are provided in the supporting information. Enc::GFP was expressed and purified as described in previously reported procedures. 5 All protein concentrations were determined photometrically (Implen, NP80, extinction coefficients of proteins were calculated using ExPASy 53 ). ## Protein expression Chemically competent E. coli BL21 Star (DE3) cells were co-transformed with pET28b(+):: 6xHis-TEV Enc and pET51b(+):: Flag Halo-cpMA-ELS. Transformants were selected on LB-agar plates containing kanamycin and ampicillin. Single colonies were isolated and validated by PCR analysis using the primer pairs as indicated in Table S1. One colony was transferred into LB-medium (containing kanamycin and ampicillin) and grown overnight at 37 °C, 180 rpm. The pre-culture (5 mL) was transferred into auto induction medium (1 L, CYP-5052, containing kanamycin and ampicillin) and grown at 18 °C for 60 hours. Cells were harvested by centrifugation (5000 x g, 4°C, 30 min) and incubated in cell lysis buffer (5 mL per gram wet weight, 50 mM HEPES, 300 mM NaCl, 0.1 % Tween 20, 1 mM DTT and 1x protease inhibitor cocktail, pH=8.0) containing lysozyme (1 mg/mL, incubation on ice for 30-45 min). Cell lysis was achieved by sonication (6 x 10 s bursts, 40 % power, 10 s cooling period between each burst, Sonoplus from Bandelin). The obtained crude cell lysate was centrifuged (10000 x g, 4 °C, 30 min) and DNase (1 µg/mL) was added to the supernatant. ## IMAC purification For subsequent HisTag purification automated affinity column chromatography was applied (HisTrap HP 5 mL, 3 columns installed in sequence, Äkta Protein Purification System, GE Healthcare Germany; Washing buffer: 50 mM HEPES, 300 mM NaCl, 25 mM imidazole, 0.05% Tween 20; Elution buffer: 50 mM HEPES, 300 mM NaCl, 500 mM imidazole, 0.05% Tween 20). The target proteins eluted at 100% elution buffer (see supporting information, Figure S3 A). After immobilized metal affinity chromatography (IMAC) purification, EDTA (0.1 mM) was added and the sample was dialyzed against aq. HEPES buffer (50 mM, pH=7.5, 10 kDa MWCO dialysis tube, cellulose membrane, overnight). TEV cleavage EDTA (0.5 mM), glutathione (3 mM) and TEV-protease (25 µL, >= 10 units/µL, Sigma) were added to the protein sample (10 mL, 2 mg/mL). After incubation for 2 days at 4 °C the solution was again purified by HisTrap. Cleaved protein was eluted from the column with washing buffer (300 mM NaCl, 25 mM imidazole, 0.05% Tween 20, see Supporting Information, Figure S3, B). ## Size exclusion chromatography During size exclusion chromatography (aq. HEPES buffer, 50 mM, pH=7.5, HiLoad Superdex 200 PG column, Äkta Protein Purification System, GE Healthcare Germany) target protein elution starts after 40 mL (see supporting information, Figure S3 C). Fractions were analyzed by SDS-PAGE, pooled and concentrated (Pierce 50 kDa MWCO centrifugal filter unit, Thermo Fisher Scientific) and stored at -20 °C until further use. The described procedure yielded ~30 mg of purified protein per liter of culture. ## Synthesis of chemical compounds RuH was synthesized according to Lohner et al. 5 The ligand of AuH was synthesized according to Nishimura et al. 54 Details on synthetic procedures of substrates, intermediates and products are provided in the supporting information.

chemsum

{"title": "Sequential, all-bioorthogonal reaction cascade catalyzed by a dual functional artificial metalloenzyme inside encapsulin", "journal": "ChemRxiv"}

pushing_property_limits_in_materials_discovery_<i>via</i>_boundless_objective-free_exploration

## Abstract: Materials chemists develop chemical compounds to meet often conflicting demands of industrial applications. This process may not be properly modeled by black-box optimization because the target property is not well defined in some cases. Herein, we propose a new algorithm for automated materials discovery called BoundLess Objective-free eXploration (BLOX) that uses a novel criterion based on kernel-based Stein discrepancy in the property space. Unlike other objective-free exploration methods, a boundary for the materials properties is not needed; hence, BLOX is suitable for open-ended scientific endeavors. We demonstrate the effectiveness of BLOX by finding light-absorbing molecules from a drug database. Our goal is to minimize the number of density functional theory calculations required to discover out-of-trend compounds in the intensity-wavelength property space. Using absorption spectroscopy, we experimentally verified that eight compounds identified as outstanding exhibit the expected optical properties. Our results show that BLOX is useful for chemical repurposing, and we expect this search method to have numerous applications in various scientific disciplines. ## Introduction Important properties for the discovery or design of novel functional materials are often either correlated or conflicting. If some materials are plotted in the space that is spanned by their various properties (property space), a distribution trend can be observed. For instance, the organic molecules as a function of excited states and their oscillator strengths are represented by a Gaussian distribution with a peak near 250 nm. 1 However, materials chemists make efforts to develop out-of-trend materials. As an example from recent research on functional organic molecules, molecules that show thermally activated delayed fluorescence (TADF) have received much attention as promising materials with drastically improved emission yields. 2 Commonly, for TADF molecules, it is necessary that the singlet excited state is close in energy to a triplet state. To achieve this, many chemists try to design molecules with minimal overlap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), but this tends to result in low emission efficiencies. Similarly, photosensitizing molecules that efficiently absorb long-wavelength light are necessary for solar cells. 3 However, the absorption of longwavelength light results in a low molar absorption coefficient. Molecules that act as UV flters 4 require the absorption of light with short wavelengths, which also results in low molar absorption coefficient. In such cases, chemists typically attempt to develop optimum materials that satisfy these conflicting demands without any information about the distribution profles of the molecules in the property space. Recently, machine-learning-based (ML-based) exploration algorithms have been investigated to optimize the materials properties. As a notable example, Gómez-Bombarelli et al. 5 have succeeded in identifying promising novel organic light-emitting diodes (OLEDs) from 1.6 million molecules by combining density functional theory (DFT) 6 simulation and ML with chemical knowledge. Among ML-based exploration approaches, efficient material searches based on black-box optimization, 7 which is a problem that fnds the maximum of an unknown (black-box) function with a limited number of evaluations, such as Bayesian optimization have been applied in various felds, and many successful examples have been reported. Drug-like molecule generation methods combining deep learning and Bayesian optimization have also been proposed. However, black-box optimization generally requires an appropriate optimization target (objective) in advance. Unfortunately, the optimal objective is not always obvious, especially when optimizing multiple properties simultaneously (so-called multiobjective problem). 11, In contrast, objective-free methods such as random goal exploration (RGE) have been proposed to search for out-of-trend materials or conditions without any explicit optimization targets. 20,21 For example, in RGE, the target properties are randomly selected in the predetermined region of the property space. Then, RGE recommends the candidate material whose properties, as predicted by ML models, are closest to the target point and then repeats this procedure to fnd out-of-trend materials. Recently, the discovery of a new protocell droplet phenomenon has been reported using a combination of RGE and robotics. 22 However, such objectivefree methods require a boundary in the property space, and search beyond the boundary is basically not assumed. Thus, if out-of-trend materials exist outside the expected boundary, we will miss an opportunity to fnd innovative materials. Here, to address the above issues, we propose a BoundLess Objective-free eXploration method, called BLOX. BLOX repeatedly recommends out-of-trend materials that lie around the edge of a distribution boundlessly, as follows. First, an MLbased model is built to predict the property values based on various materials for which current data on calculated or measured properties is available. For the predicted locations of candidate materials without true properties in the property space, BLOX selects the most deviated material with the criterion of "similarity" to the uniform distribution. That is, if the predicted properties of a candidate material deviate from the distribution of the current data, the entire distribution is scattered and consequently approaches the uniform distribution. For these calculations, BLOX employs Stein discrepancy, 23,24 which can boundlessly evaluate a kind of distance (similarity) between any two distributions in any dimensional space. For the recommended most deviated material, its properties are measured through experiments or simulations. By repeating these recommendations and measurements, BLOX realizes an efficient exploration that expands the limit of the distribution in the property space boundlessly. To demonstrate the performance of BLOX, we searched for effective light-absorbing molecules (that is, chemical compounds that absorb light with high intensity) from the drug candidate database ZINC, 25 which has not previously been investigated as a molecular database for determining photochemical properties through calculations and experiments. To evaluate the performance of BLOX, we have also carried out a search based on random sampling, which randomly selects molecules with the fxed number among the prepared dataset of candidate molecules. We succeeded in fnding out-of-trends molecules using a small number of trials based on BLOX and DFT calculations more effectively than random sampling. Furthermore, we selected eight of the out-of-trend molecules obtained by BLOX for experimental verifcation and confrmed that their absorption wavelengths and intensities were almost consistent with the computational results. This demonstration suggests that BLOX has potential as a tool for discovering outstanding materials. ## BLOX We show an overview of BLOX in Fig. 1. Our implementation of BLOX is available at http://github.com/tsudalab/BLOX. In BLOX, after the initial preparation step (Step 1), the search is performed by repeating the following three steps: the construction of a property prediction model (Step 2), the selection of a candidate using the Stein novelty score based on Stein discrepancy (Step 3), and the evaluation of the selected candidate by experiment or simulation (Step 4). The details of each step are as follows. In Step 1, a dataset of samples (materials/molecules) are chosen for searching and objective properties are determined. Two or more objective properties can be set. Here, there is no need to design an appropriate evaluation function and set a search region (boundary). A small amount of property data obtained from experiments or simulations is needed because BLOX uses ML to predict properties. Previously measured property data can be used, if available. If no property data is available, experiments or simulations must be conducted for a small number of randomly selected candidates. As a demonstration, in this study, we employed the ZINC database and selected 100 000 commercially available molecules with small ZINC indexes from ZINC000000000007 to ZINC000002386999 as a candidate molecules database. We used the absorption wavelength for the frst singlet excited (S 1 ) state and its oscillator strength (as an alternative to the experimental intensity) as objective properties to fnd molecules. For the initial sampling, we selected 10 molecules randomly and calculated the values of their objective properties using DFT. The computational details are given in Step 4. In Step 2, an ML-based prediction model is built for the objective properties based on the already evaluated materials and their property data. Any method that can predict the desired properties of materials can be used. In our demonstration, as a simple example, we built two models for predicting the absorption wavelength and intensity using the Morgan fngerprint, 26 which is widely used in chemoinformatics, and classical ML methods. For each molecule, we calculated its fngerprint, which is a 2048-dimensional vector consisting of values of 0 and 1, using RDKit. 27 We normalized the calculated fngerprints and property values. For the training dataset (pairs of fngerprints and property values for already evaluated molecules), we train two prediction models using standard ML techniques, namely, Lasso regression, 28 Ridge regression, 29 support vector regression (SVR), 30 random forest (RF), 31 and neural network (NN). A frst-degree polynomial function is employed as the basis function of Lasso and Ridge regression. Although it has been reported that NNbased methods such as Graph Convolutional Networks (GCNs) 32 are superior in predicting chemical/physical properties of molecules, such NN-based methods, particularly deep NN-based models, generally require large dataset to be effective. In BLOX, it is required to train the prediction model with a very small dataset, especially at the beginning of the search. Therefore, in this study, we mainly used conventional ML methods. In this study, as the NN model, we utilized a multilayer perceptron used in the previous studies. 33,34 The network has three hidden layers and the number of neurons in each layer is 500, 500, and 100. We used the scikit-learn library 41 to perform the above calculations. In Step 3, a candidate is selected for evaluation in Step 4 based on Stein discrepancy. 23,24 First, for unchecked materials in the database, we predict their properties (open triangles in Step 3, Fig. 1) using the prediction models developed in Step 2. Most of the predicted points are expected to be distributed around some trends. However, the trends are generally unde-fned. Next, we select the most deviated candidate (triangle surrounded by the red circle in Step 3, Fig. 1) using Stein discrepancy (see the ESI † for computational details of Stein discrepancy). Fig. 2 illustrates the concept of Stein novelty (SN), which is our introduced index to select a next candidate, based on Stein discrepancy and the observed distribution. We can quantify the discrepancy SD(V) between the observed distribution V (Fig. 2a) and the uniform distribution U (Fig. 2b) using Stein discrepancy. Here, we evaluate the Stein discrepancies when a new point (predicted unchecked candidate) denoted by p is added to the observed distribution, as in Fig. 2c and d. If the new point deviates more from the observed distribution, as in Fig. 2c, its discrepancy is smaller. Then, we introduce SN to measure the degree of deviation, as given in the following equation: where p is a predicted unchecked point by ML (see the ESI † for computational details of the SN). As the SN increases, the deviation grows. Fig. 2e shows the visualized SN distribution for the observed distribution V. In this step, we select the candidate with largest SN. In Step 4, the candidate selected in Step 3 is evaluated by experiment or simulation. In the demonstration, for the selected molecule, we calculate the absorption wavelength for the S 1 state and its oscillator strength using DFT, as follows. A three-dimensional structure is converted from the molecule in SMILES string with RDKit. After optimizing its conformational structure at the universal force feld (UFF) level, we optimize it using DFT at the B3LYP/6-31G* level. Then, we compute the absorption wavelength and the oscillator strength of the molecule using time-dependent DFT (TD-DFT) at the same level. For the TD-DFT computation, the lowest ten excited states are computed. All DFT calculations were performed with the Gaussian 16 package. 42 In this study, when a calculation failed in the middle, we stopped the computation and instead performed the calculation for the molecule with the second-highest SN score. ## Absorption spectra We experimentally measured the absorption wavelengths and intensities of the selected test molecules (Table S1 †). Test molecules were used as received expect for molecule (ii). Molecule (ii) was purifed with column chromatography on silica gel since it involved some impurities in 1 H-NMR analysis. Absorption spectra in solution were recorded using a Shimadzu UV-3600 UV-vis-NIR spectrophotometer. A quartz cell with 1 cm path length was used. Spectroscopic grade solvents, purchased from Tokyo Chemical Industry (TCI) and Wako Pure Chemical Industries, were used as received. Solvatochromic effect and concentration dependencies are detailed in the ESI. † To exclude the influence of trace impurities, test molecules were analysed by high performance liquid chromatography (HPLC) (see Table S3 and Fig. S23-S29 †). It is confrmed that absorption spectra of main fractions in HPLC were consistent with the absorption spectra shown in Fig. 6. In addition, purity of test molecules were analyzed by HPLC. Note that molecule (i) was omitted in HPLC analysis due to the lack of measurable absorption spectrum. See ESI † for experimental details of HPLC. ## Characterization of test molecules Test molecules were characterized by 1 H-NMR spectrometry (Fig. S5-S12 †) and high resolution mass spectrometry (HRMS, Table S2, Fig. S16-S22 †). 1 H-NMR spectra were obtained using an AL300 BX spectrometer (JEOL, 300 MHz). HRMS was recorded on a Bruker TIMS-TOF spectrometer with samples dissolved in 1 : 1 acetonitrile : methanol (0.1 mg mL 1 ). ## Results and discussion We performed a BLOX trial using RF-based prediction models to fnd out-of-trend molecules from the ZINC database using the absorption wavelength and intensity as objective properties. The orange points in Fig. 3 indicate the molecules (samples) found by BLOX sampling in the property space consisting of the absorption wavelength for the S 1 state and its oscillator strength (intensity) after 200 (A), 500 (B), and 2000 (C) samplings. To compare the performance of BLOX with RF, we also sampled molecules randomly from the database and evaluated their properties (blue triangles, Fig. 3). The distribution obtained by random sampling suggests the presence of the trend, that is, molecules whose absorption wavelength distributed in the range of 250-400 nm with high intensities. In comparison, BLOX with RF (orange points, Fig. 3) successfully found out-oftrend molecules that have high intensities with shorter (<250 nm) and longer (>400 nm) absorption wavelengths. We picked molecules (i)-(viii) in Fig. 3C as examples of out-of-trend molecules for further experimental verifcation by UV-vis absorption spectrum measurements, as discussed later. To investigate the effect of different prediction models on the search results, we also performed BLOX trials using the Lasso, Ridge, SVR, and NN models. The initial 10 molecules were the same in all the searches, including the random sampling. Fig. S1 † shows the search results using Lasso ((a)-(c)), Ridge ((d)-(f)), SVR ((g)-(i)), and NN ((j)-(l)). Fig. 3 and S1 † clearly show that the molecules found using the RF and SVR models are distributed over a wide range, i.e., many out-of-trend molecules are found. To evaluate the difference between the prediction models quantitatively, we show the Stein discrepancy values as functions of the number of samplings in Fig. 4. A Stein discrepancy value closer to 0 indicates a greater similarity between the observed distribution and the uniform distribution, i.e., the obtained molecules are distributed more widely in the property space. The results in Fig. 4 show that the Stein discrepancy values decrease immediately after the start of sampling in all samplings using BLOX except for NN, and that the Stein discrepancy values of the BLOX trials are signifcantly lower than those of the random sampling. In addition, the nonlinear prediction models (SVR and RF) have lower Stein discrepancy values than the Lasso and Ridge models. From Fig. 4, NN fnally showed high performance after 2000 sampling, but it was comparable to random search when the number of sampling is small. Thus, it was quantitatively confrmed that the molecules searched using RF and SVR were distributed over a wide range in the property space. We adopted some ML methods to build the prediction models. As references, the performance of the predictions with RF and NN for the absorption wavelength and intensity are evaluated in Fig. S2 and S3 in ESI. † In RF, although the prediction accuracy in the demonstration was low when the number of evaluated data was small, this did not seem to cause fatal problems because the BLOX trials successfully found outof-trend molecules more effectively than the random sampling, even with a small number of samplings, as shown in Fig. 4. Furthermore, the prediction accuracy of BLOX can be enhanced by increasing the amount of sample data. When the number of sampled molecules increases, the prediction accuracy is improved, as shown in Fig. S2C and F. † Recently, property prediction methods using various ML methods, including deeplearning techniques, have been proposed. 32,37,38, As stated in the method section, although NN-based, particularly deeplearning-based, prediction models are known to have high accuracy, they are not always practical because the amount of data is limited and training time is required for each sampling in BLOX. In fact, from the prediction performances of the NN model as shown in Fig. S3, † we can see that the prediction accuracy is low when the number of training data is small, such as Fig. S3A, B, D and E, † whereas the accuracy improves with the increase of the number of training data, such as Fig. S3C and F. † Due to this low accuracy, out-of-trend molecules are not found at the beginning of search. Thus, it is important to use an appropriate ML technique so that the prediction accuracy is not too bad. As another approach to increase the accuracy of the prediction model, Proppe et al. have proposed a strategy to select dissimilar molecules to use as a training dataset by combining Gaussian process and active learning to build an accurate prediction model of dispersion correction parameter in DFT calculations. 51 Although their method has a different objective from BLOX, in the future, incorporating their method may improve the exploration performance of BLOX by enhancing the accuracy of the prediction model. In addition, the framework of BLOX is applicable for other materials, if properties can be predicted to some extent using an ML-based model from materials descriptors. For example, in solid materials, the magpie descriptor has been reported for predicting physical properties such as superconducting temperature and bandgap. 52,53 Also, in solid-state materials community, various types of descriptors for composition and structure of atoms have been prepared, and using some tools, we can easily obtain these descriptors using libraries like RDKit. 27 For actual application of BLOX to other materials, it is required to select both an appropriate prediction model and a descriptor that match the dataset and target properties with taking a balance between the prediction accuracy and training time. The time required for the BLOX search consists of three components: the time to train the ML-based prediction model (training time), the time to select the next candidates based on the SN (selection time), and the time to evaluate the selected candidate through experiments or simulations (evaluation time). The appropriate allocation of these computational times depends on the size of the database, the prediction model used, and the cost of the experiments or simulations. The training and selection times required in this study on a 12 core (Intel Xeon Gold 6148 CPU) server are shown in Fig. S4. † Although the calculation time tended to increase with an increase in the amount of observed data, the calculation was completed in a few tens of seconds to $2 min. The average simulation time on the same server was 29.74 min per molecule. Therefore, in this study, the training time for the ML-based prediction models and the selection time were sufficiently short in comparison with the evaluation time. In this study, we used 100 000 molecules in the ZINC dataset as an exploration demonstration. However, BLOX is applicable to a larger dataset because only the selection time increases when searching in a larger dataset. As shown in Fig. S4, † the selection time is much shorter than for experiments and detailed simulations, and these predictions and Stein novelty calculations for each material candidate in a dataset can be performed independently. This indicates that the selection time will only increase linearly with respect to the increase of the dataset size. In addition, because of the independence, the selection can be further accelerated by using more CPUs, while 12 CPUs were used in this study. Therefore, there is nothing to hinder our method from applying to a larger dataset. Furthermore, to explore an open chemical space beyond the fnite dataset, the combination of BLOX and de novo molecule generation methods 15, can be a promising approach. Most of these generation methods to generate molecules with target properties by sequentially evaluating the properties (scores) of the generated molecules. Here, using the BLOX framework (especially Step 2 and Step 3 in Fig. 1), we can evaluate the degree (scores) of deviation (out-of-trend) for generated de novo molecules. The BLOX framework would be extended to be truly boundless by combining such an evaluation strategy and de novo molecule generation. We successfully demonstrated the effectiveness of BLOX using the example of light absorption for molecules described at the DFT level. Hereafter, we discuss the validity from the experimental viewpoint considering the error of DFT. Each of the selected molecules ((i)-(viii) in Fig. 3C) is shown in Fig. along with its dominant electronic transition to the S 1 state calculated at the B3LYP/6-31G* level. The S 1 state of each molecule is attributed to a HOMO-LUMO single electron transition. The excitation of (i) is attributed to s-s* excitation, as reflected by the high energy of the excitation (133.5 nm). For (ii)-(viii), the S 1 excited states are attributed to p-p* excitation. In particular, the S 1 states of (ii)-(iv) induce bond alternation (double bonds in the S 0 state becomes single bonds in the S 1 state). Hence, the oscillator strength is strong because the overlap between the HOMO and the LUMO is large. However, (v)-(viii) absorb light at wavelengths longer than 500 nm, indicating that the S 1 states have charge-transfer properties. Thus, the overlap between the HOMO and the LUMO of each of these molecules is small, as reflected by their low oscillator strengths. We validated the calculated absorption properties of (i)-(viii) using UV-vis absorption spectra measurements, as shown in Fig. 6 along with images of the solutions (see Fig. S13 and S14 † for UV-vis absorption spectra at other concentrations and in other solvents, respectively). The solutions of (i)-(iii), which absorb light at wavelengths shorter than 300 nm, are transparent and colorless, whereas those of (iv)-(viii), which absorb light at wavelengths longer than 400 nm, are colored. We could not record the absorbance of molecule (i) in any available solvent. As mentioned previously, the S 1 states of (v)-(viii) have charge-transfer properties, which results in a computational underestimation of the S 1 energy, 62 as reflected in Fig. 6(v)-(viii). However, the absorption spectra of (ii)-(viii) indicate that the experimental absorption peaks nearly correspond to the calculated S 1 energies (broken red lines in Fig. 6). Concerning the intensity, the lowest energy absorption bands of (iii) and (iv) show high molar absorption coefficients on the order of 0.4 10 5 mol 1 L cm 1 , whereas those of the other molecules are at least 0.4 10 4 mol 1 L cm 1 . The molar absorption coefficients observed for (ii)-(viii) correlate well with the calculated oscillator strengths, as shown in Fig. 7. Therefore, the BLOX framework can fnd plausible molecules despite the evaluation being performed at the DFT level. Herein, we employed the ZINC database, which consists of drug candidates. For example, (i) (paraldehyde) is widely used as a sedative, hypnotic, and anticonvulsant. 63,64 (vii) has been reported as one of the anticancer drug candidates. 65 However, as (iv)-(viii) are colored molecules, they may also be useful as benign dyes, e.g., for organic solar cells. Furthermore, (i) and (ii) may be useful as harmless UV flters that block strong light (at short wavelength). In a similar attempt, the repurposing of deoxyribonucleic acid topoisomerase inhibitors as organic semiconductors has also been reported. 66 The results of our demonstration suggest that BLOX has the potential to accelerate the discovery of new materials by using databases collected for one purpose in other unintended felds. ## Conclusion In conclusion, we proposed a novel search method (BoundLess Objective-free eXploration; BLOX) for the effective discovery of out-of-trend materials from a dataset based on Stein discrepancy. Notably, BLOX does not require any information about the distribution of the materials in the property space as input. To demonstrate the utility of this method, we applied BLOX combined with DFT-based simulations to fnd lightabsorbing molecules with high molar absorption coefficients in a database of drug candidates. BLOX showed better performances to fnd out-of-trend molecules, compared to random sampling. Furthermore, it was experimentally confrmed that the discovered compounds absorbed at the expected wavelengths. We believe that this method will be useful for fnding unexpected and out-of-trend materials that have the potential to push their property limits by developing derivatives. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Pushing property limits in materials discovery <i>via</i> boundless objective-free exploration", "journal": "Royal Society of Chemistry (RSC)"}

continuous_hydroformylation_of_1-decene_in_an_aqueous_biphasic_system_enabled_by_methylated_cyclodex

## Abstract: For the first time, randomly methylated β-cyclodextrin was applied as the mass transfer agent in a continuous process. Considering the example of the Rh-catalyzed hydroformylation of 1-decene, process development was shown, where cyclodextrin was used together with a catalyst system that was continuously recovered and recycled using an aqueous biphasic system. In initial experiments, water-soluble and commercially available Rh/TPPTS and Rh/sulfoxantphos catalyst systems were scaled up from 50 ml into 1000 ml high-pressure autoclave systems to demonstrate their scalability. Both these systems were compared, and they afforded excellent chemoselectivity (>99%) toward the desired linear aldehyde product. In particular, higher regioselectivity (up to 31) was achieved for the Rh/sulfoxantphos system. Investigations regarding the long-term stability of the mass transfer agent and both catalyst systems were carried out in a continuously operated miniplant process. It was shown that the process could be successfully operated under the steady state for over 200 h with chemoselectivity of >97% toward the desired aldehyde product. Simultaneously, extremely low Rh leaching (total: 0.59%) was observed over the entire period of 200 h. † Electronic supplementary information (ESI) available: Reaction profiles for the Rh/TPPTS and Rh/sulfoxantphos systems are shown for both 300 and 1000 ml autoclaves. Further, the composition of the distillate stream for the continuously operated distillation is shown. See ## Introduction Hydroformylation is widely used in industry to manufacture high-value-added aldehydes and alcohols, and it is considered to be one of the largest homogenously catalyzed reactions in industry. A major issue persisting with the use of homogenous catalysts is their recovery and separation from the product mixture, particularly for higher olefins; this is important due to the cost of the employed transition metals. These days, distillation has become the most commonly used separation method in the hydroformylation of alkenes and therefore the most well-known method. 1 However, distillation conditions that are required to separate high-boiling products may lead to catalyst decomposition along with concomitant loss of metal. Hence, cheaper cobalt catalysts are used, ensuring that the processes remain economically viable. 2 However, cobalt cata-lysts are less active than rhodium catalysts, requiring higher reaction temperatures and pressures. 3 Therefore, the challenge involved in formulating techniques to convert higher olefins in hydroformylation persist, particularly in a manner that can address the separation problem and minimize environmental impact, while not compromising on catalyst selectivity and activity. Some concepts focused on this involve the intensification of the monophasic flow process for hydroformylation. 4 Other approaches have been developed in both industry and academia to tackle the main issue of catalyst separation from the product by using rhodium catalysts. Examples of these are the use of supercritical CO 2 (scCO 2 ), 5 CO 2 -expanded liquid (CXL), 6,7 supported ionic liquid phase (SILP), 7 organic solvent nanofiltration (OSN), 8 thermomorphic multiphase systems (TMS), 9 microemulsions, 10 or biphasic systems. 11 Some of them use abundant resources (such as carbon dioxide and the environmentally friendly solvent, water), but others do not. The use of water as a solvent for process applications is particularly beneficial from the economic and environmental impact viewpoints since water is fairly accessible, nontoxic, nonflammable, odorless, and has a high heat capacity and heat of vaporization. 12 Nevertheless, these benefits have to compete with its intrinsic limitation of low organic substrate solubility in the aqueous catalyst phase, which have also been confirmed by early kinetic studies. 13,14 Therefore, several approaches have been developed to tackle the issue of low space-time yield in biphasic aqueous-organic reaction systems involving poorly water-soluble substrates, namely, by using surfactants, cosolvents, thermoregulated ligands, and cyclodextrins (CDs) to achieve improved hydroformylation, as summarized in the review article by Matsinha et al. 14 From this perspective, this work further focuses on the use of CDs as the mass transfer agent to overcome mass transfer limitations in the biphasic hydroformylation of 1-decene. CDs are cyclic oligosaccharides comprising six (α-CD), seven (β-CD), or eight (γ-CD) α-D-glucopyranose units. 15 They form conical cylinders with the outer hydrophilic surface and inner hydrophobic surface. The inner surface forms a cavity and allows for the formation of an inclusion complex when it binds with a hydrophobic substrate (Fig. 1). 16 To vary their size and shape, native CDs can be modified by substituting their hydroxyl groups with various other functional groups. These are referred to as mono-or polysubstituted CDs. Due to their ability to form inclusion complexes with a wide range of compounds, native and modified CDs have applications in various fields. 17 In the field of aqueous biphasic organometallic catalysis, CDs have mainly been employed as mass transfer agents 16,18 in order to increase the solubility of hydrophobic substrates in water. For example, in 1995, Monflier et al. 16, described rhodium-catalyzed hydroformylation in an aqueous/ organic two-phase system of 1-decene in the presence of randomly methylated β-CD (RAME-β-CD; Fig. 1). 21,23 The phase separation between the organic and aqueous phases was described to be fast; further, the rhodium and phosphorus contents in the organic phase were found to be less than 0.5 and 1.2 ppm, respectively. Moreover, in 1999, Tilloy et al. proposed the reusability of the Rh/TPPTS/RAME-β-CD catalytic system by performing the rhodium-catalyzed hydroformylation of 1-decene with 5 catalytic recycle runs without any loss of activity. 24 Later, in 2004, Leclercq et al. investigated the same reaction using the sulfoxantphos ligand, which led to improved performances in terms of chemoselectivity and regioselectivity. 20 Although CDs are highly promising as mass transfer agents in aqueous biphasic systems for the selective hydroformylation of higher olefins, they have never been applied in a continuously operated process. As our research group has gained expertise in various continuously operated recycling strategies for homogeneous catalysts with regard to hydroformylation 25,26 in earlier years, we combined our knowledge with the group of Monflier to understand if CDs can be employed for such applications. Therefore, in this work, we transfer two catalytic systems (Rh/TPPTS/RAME-β-CD and Rh/ sulfoxantphos/RAME-β-CD) into one of our continuously operated miniplants for the purpose of aqueous biphasic hydroformylation. Since this has never been attempted before, both these systems were first scaled up in differently sized batch autoclaves in order to understand their performance in terms of chemo-/regioselectivity and conversion. Further, their reaction profile was determined to be the benchmark for the residence time in the subsequent continuous process. During continuous operations, the influence of various parameters (such as CD concentration and stirrer speed) on the catalytic activity was determined. Eventually, the effective recycling of the catalytic systems (supported by leaching data) with the optimized parameters was conducted via a long-term experiment, revealing the potential of CDs for such applications. ## Results and discussion Based on the studies by Monflier et al., 16, 1-decene was chosen as the model substrate for CD-based hydroformylation. The reaction network is shown in Fig. 2. The main product from the hydroformylation is undecanal, which is referred to as l-undecanal (linear undecanal). 1-Decene isomers may occur as byproducts and are referred to as iso-decene (isomerized decene). The other byproducts are n-decane (which results from the hydrogenation of 1-decene or iso-decene) and the branched hydroformylation product, namely, 2-methyldecanal (which can arise from the hydroformylation of 1-decene or isodecene). Further, 2-methyldecanal is the representative of all the branched aldehydes since it is formed as a byproduct in the highest amount. These branched aldehydes are categorized as b-undecanal (branched undecanal). All the undecanal isomers are categorized into undecanal (l-undecanal + b-undecanal). The remaining byproducts are aldol condensates, which can result from the condensation reaction of two aldehydes. All these products are summarized, and 2-nonyl-tridec-2-enal is considered to be the representative of the aldol condensates. In this work, two catalyst systems have been investigated, since they yielded worthwhile results in earlier publications with regard to high product chemo-/regioselectivity. The already published results are listed in Table 2. 18,20,24 Both these systems use water as the solvent, Rh(acac)(CO) 2 as the precursor, and RAME-β-CDs (Fig. 1) as the mass transfer agent. The only difference is the choice of phosphine ligand, namely, monodentate TPPTS and bidentate sulfoxantphos. ## Scaled up batch experiments Scaled up experiments were first performed since the original reactions were carried out in 50 ml high-pressure autoclaves. Therefore, two types of high-pressure autoclaves were used for the experiments (Fig. 3). A high-pressure autoclave with a total volume of 300 ml was used as the first reference as well as for the scaling up tests. For a facilitating comparability between the experiments, the gas-to-liquid volume ratio was maintained as per the original experiments performed by Monflier et al. 19 and Leclercq et al., 20 yielding a volume of 100 ml for the reaction solution. In this work, this reactor is also referred to as the 300 ml high-pressure autoclave. Another reactor is built into the continuous miniplant process and was used in the batch or continuous mode. This reactor had a total volume of 1000 ml and is correspondingly referred to subsequently. To maintain the gas-to-liquid volume ratio constant, the liquid volume was set to 330 ml. 2.1.1. Scaled up Rh/TPPTS catalyst systems. The first scaling up experiments were conducted for the Rh/TPPTS catalyst system. Based on the reaction data from the original experiments (Table 1), the Rh(acac)(CO) 2 precursor and TPPTS ligand were used to convert 1-decene into undecanal. The reac- Table 1 Comparison of aldehydes (chemo-)selectivity and linear-tobranched aldehyde (l/b) ratio (regioselectivity) for the rhodium-catalyzed hydroformylation of higher olefins by Rh/TPPTS and Rh/sulfoxantphos systems (adapted from Hapiot et al. 21 tion temperature was set to 80 °C at a phosphorus-to-rhodium ratio (n P /n Rh ) of 5 and CD-to-rhodium ratio (n CD /n Rh ) of 12. The reaction time was set to 6 h, since this is the maximum possible residence time in the reactor for the continuous process at the rhodium-to-substrate ratio of 496. Before the start of the reaction, preforming was conducted for 2 h. In the original experiments, pressure of 50 bar was used for the reactions to ensure sufficient gas solubility and to preclude the limitation of the reaction. Due to apparatus technology, only pressures of 40 bar could be used for the 300 and 1000 ml autoclaves. However, Mathivet et al. mentioned that reaction pressure of 10 bar was sufficient; any further increase in pressure only had a minor impact. 27 The stirrer speed was set to 800 min −1 for the scaling up experiment. The results are shown in Fig. 4. Evidently, as the reactor size increased, the catalytic activity as compared to the original 50 ml experiments decreased. Nearly full conversion could be achieved in the 50 ml autoclaves as compared to the 300 ml autoclaves; the conversion decreases to 75% and eventually to 33% when compared to the 1000 ml autoclaves. The chemoselectivity of all the three experiments are in the same range (over 95%), which indicates that the Rh/TPPTS catalyst system works as expected for all the experiments. Further, with respect to regioselectivity, the l/b ratio reached a value close to 1.8 for all the experiments. As these experiments were conducted to demonstrate the influences of different parameters as barriers for the potential transfer into a continuous process, the obtained results revealed that a transfer is generally possible when the selectiv-ities remain constant. However, in a continuous process, lower conversions have to be expected due to the potential mass transfer limitations due to an increase in the size of the autoclaves. Since this is a three-phase gas-liquid-liquid system, there may be limitations at the gas-liquid phase interface and/ or at the liquid-liquid phase interface, which is supported by the investigations of Sieffert et al.; they suggested that the reaction does not take place in the bulk water phase, but it occurs at the aqueous phase/interface layer. 28 Therefore, there would be stronger dependence between the phase interface and reaction rate and consequently the conversion rate. With scaling up and an increase in volume, the amount of substance gets proportionally scaled, but the phase interface does not. The reaction profiles for the 300 and 1000 ml systems are provided in the ESI, † revealing the potential undecanal yield for different residence times in the miniplant reactor. 2.1.2. Scaled up Rh/sulfoxantphos catalyst system. Scaled up experiments performed for Rh/sulfoxantphos catalyst systems are shown in Fig. 5. Reaction conditions do not drastically differ from the previous experiments: only the reaction temperature was increased from 80 °C to 120 °C and the preforming time was extended from 2 to 12 h. Similar to the results obtained from the Rh/TPPTS catalyst system, the conversion of 1-decene and undecanal yield generally decrease with an increase in the reactor size. The reaction profiles of the 300 and 1000 ml systems are provided in the ESI, † revealing the potential undecanal yield for different residence times in the miniplant reactor for continuous operations. However, the decrease in conversion is less as compared to that for TPPTS system because of the surface properties of sulfoxantphos. Here, the conversion decreases from 71% (50 ml autoclave) to 51% (1000 ml autoclave). The undecanal yield analogously decreases with the conversion from 70% to 46% as the chemoselectivity values for all the experiments remain over 90%. Further, the regioselectivity (l/b ratio) of around 27 for all the experiments shows the high regioselectivity toward the linear product, i.e., aldehyde. ## Continuous miniplant experiments The proofs of concept for the successful scaling up of the Rh/ TPPTS and Rh/sulfoxantphos catalyst systems have been shown in earlier batch experiments. As regio-and chemoselectivity are nearly constant for these experiments, the catalyst systems are transferred for the first time into a continuous process (Fig. 10). In this study, the long-term stability of catalytic systems in combination with CD, as well as influences of parameters such as CD concentration and stirrer speed in the reactor, were investigated. 2.2.1. Variations in CD concentrations with TPPTS. In the first continuous experiment, the Rh/TPPTS catalyst system was used to show the effect of CDs as a mass transfer agent during the reaction under biphasic conditions. The obtained results are shown in Fig. 6. The first 8 h are the startup time of the plant (shown with the grey bar). Initially, no CDs were added to the system that proved the poor reaction activity, similar to that with the aqueous biphasic mixture. As expected from hydroformylation experiments in aqueous biphasic systems as reported in the literature, 29 the observable yield of undecanal (linear and branched) was low, under 0.5% after 21 h at a stirrer speed of 500 min −1 . To increase the reaction performance, CDs were added to the system at a molar ratio of 6 with respect to the rhodium catalyst. After 29 h, the yield of undecanal increased only to 1%. Consequently, the stirrer speed was increased from 500 min −1 to 800 min −1 to increase the mixing in the reactor and potentially lowering the mass transfer limitations. Under these conditions, the yield of undecanal reached a steady value of 11.5% after 47 h. After 52 h, more CDs were added to the reactor. The molar ratio of CDs to rhodium increased to 12, which was used earlier in the batch experiments (Fig. 4 and 5). The yield increased after some induction time, which roughly equaled to the residence time; therefore, after 65 h, the undecanal yield was about 20%. As compared to the batch experiment in the 1000 ml autoclave (ESI †) under the same conditions, the undecanal yield was determined to be 10% less. This could be attributed to the recycling of the catalyst. To further increase the conversion rate, more CDs were added after 76 h, resulting in a molar ratio of CDs to rhodium of 18. The conversion increased further, reaching a steady state value of 32%. Initially, the regioselectivity (l/b ratio) was ∼2.8. After the first addition of CDs, the ratio rapidly dropped to 2.0 and then slowly reached a value within 1.8-1.9, which was the ratio determined for TPPTS in the batch experiments. The further addition of CDs had no influence on the ratio. The chemoselectivity (∼99%) was calculated over the entire duration of the experiment and therefore was higher than that in the batch experiments (Fig. 4). In Table 2, the catalyst leaching results for this experiment are listed. The samples taken from the product stream showed the total leaching (L) of the ligand of 1.2% and L value of rhodium of 0.8% after 94 h. 2.2.2. Variations in stirrer speeds with sulfoxantphos. With an increase in the CD concentration in the reactor, the mass transfer increased and hence the product yield for the Rh/ TPPTS catalyst system also increased: the same was expected for the Rh/sulfoxantphos system. Therefore, the subsequent continuous experiment does not only investigate the feasibility of the Rh/sulfoxantphos catalyst system in a continuous process, but also reveals the mixing behavior of the phase system in a more comprehensive manner. For this, the molar ratio of CD to rhodium was maintained constant at 12 and the stirrer speed at the beginning was set to 800 min −1 . The obtained results are shown in Fig. 7. The first 8 h represent the startup time of the plant (denoted by the grey bar). Thereafter, the yield of undecanal steadily increased to 32% until the 34 th Table 2 Leaching results with respect to the catalyst and ligand for continuous experiments when using TPPTS as the ligand hour, which is marginally lower than the expected from the batch reaction profile experiment (38%) (ESI †). The miniplant operation was considered to be stable at this point. To investigate if the mass transfer limitation limited the conversion, the stirrer speed was increased from 800 to 1200 min −1 to mechanically enhance mass transfer between both the phases and consequently resulting in higher conversion. However, the undecanal yield decreased to 20%. However, at the same time, foam formation increased in the reactor, which is considered to be the potential reason for the decreased yield since mass transfer between the liquid-liquid phases can be inhibited by foam. After the plant was steady again, the stirrer speed was reduced from 1200 to 500 min −1 at about 66 h. The undecanal yield slowly increased after 6 h (equal to the residence time). At the end of the experiment, the yield was 34%. Furthermore, it should be noted that the foam phase reduced considerably at 500 min −1 . However, changes in the stirrer speed and therefore in the mass transfer did not influence the chemoselectivity at all, which was maintained at >95% for the aldehyde product. The regioselectivity (l/b ratio) started at 27.5 and slowly increased to a steady-state value of 30.5 after 20 h, which are similar to the ratios achieved in the batch tests. Because of the differences in the range of measurement as well as continuous operation fluctuations, no direct influences of stirrer speed and l/b ratio are evident. In Table 3, the catalyst leaching results are represented. In the beginning, the leaching values of rhodium and ligand were slightly higher. This is most likely due to the startup procedure and lower average residence time in the decanter. Further loss corresponded approximately to the value determined for the entire operating time. Over the operating time of 93 h and with an average leaching of the ligand of 0.0122% h −1 , the total ligand leaching added to 1.14%. The average loss of rhodium was 0.0059% h −1 , and therefore, the total rhodium leaching added to 0.55%. 2.2.3. Long-term experiment with sulfoxantphos. In this experiment, the earlier results from the miniplant (which afforded the highest yields and selectivities) were used to show the long-term effects when the plant was running in the steady state. Therefore, the experiment was carried out with a sulfoxantphos catalyst system as it generally afforded higher yields and higher regioselectivities as compared to the Rh/TPPTS catalyst system. The amount of CD was set to n CD /n Rh = 18 for the experiment shown in Fig. 6 and the stirrer speed was set to 500 min −1 for the experiment shown in Fig. 7. The results of this experiment are shown in Fig. 8. At the beginning of the startup phase, the undecanal yield was ∼6%. This steadily increased until reaching a stationary value of 39% after 38 h, which was also expected from the batch reaction profile experiments (ESI †). This value was maintained stable in a steadystate operation over 200 h with maximum deviations of 1.5% until the end of the test after 241 h. The chemoselectivity of undecanal was initially >99%. Until the stationary point was Table 3 Leaching results of the catalyst and ligand for continuous variations in the stirrer speed using sulfoxantphos as the ligand reached, the chemoselectivity dropped to 97%, but retained this value with maximum deviations of over the entire operating time. This behavior of chemoselectivity was also observed in the earlier continuous experiment with sulfoxantphos (Fig. 7). The regioselectivity (l/b ratio) was 27, which increased until reaching the stationary value of 29.6 after 40 h of operation. This value was maintained with minimum deviation until the end of the experiment. In Table 4, the catalyst leaching results are presented for the entire time of operation. In general, no significant deviation in leaching was observed, which supports the evidence that CDs are very efficient as mass transfer agents. Over the operating time of 241 h, the average leaching of ligand was 0.0098% h −1 ; therefore, the total leaching of the ligand added to 2.17%. The average loss of rhodium was 0.0027% h −1 , and therefore, the total rhodium leaching added to 0.59%. The long-term stability of the system could, therefore, be demonstrated in this experiment. No makeup stream was needed to maintain the catalytic activity over the entire course of the experiment. During the experiment, a white solid deposit was observed in the decanter through a viewing window. This was noticed for the first time after ∼70 h of operation, and this deposit increased over the test period. The formation of this solid could be attributed to an increase in the CD concentration at the beginning in the sulfoxantphos system (n CD /n Rh = 18 instead of n CD /n Cat = 12 as compared to the batch experiments) and not to the long operating time itself. Solubility problems, therefore, appear to occur, which further exacerbate with an increase in the CD concentration, which could become problematic during continuous operation and should not be considered. 2.2.4. Concept of distillation and substrate recycling. The aqueous biphasic hydroformylation of 1-decene using CDs as the mass transfer agent exhibited high stability, low catalyst leaching, and high chemo-and regioselectivity in the continuous experiments under consideration; however, it lacked high turnover rates. To make it a more attractive process for commercial applications, process intensification is imperative. Consequently, several techniques have been formulated, e.g., increasing the residence time of 1-decene in the reactor by increasing the reactor volume or decreasing the feed and recycle flow rates to the reactor in order to increase conversion. Further, the increased mixing behavior can yield higher turnover, which can be achieved by changing the stirrer geometry and stirrer speed, which has already been shown earlier (Fig. 7). Further investigations in this particular field are currently ongoing. The third way to improve the efficiency of this process is to recover and recycle the nonconverted substrate after the reaction. Fig. 8 shows that up to 58% of the nonconverted substrate leaves the process after phase separation. A huge improvement could be made if the substrate could be recovered and recycled. Therefore, a potential concept of this process has been proposed, as shown in Fig. 9, by integrating (step I: hydroformylation) a second purification step into the already existing system by using a distillation column to separate the substrate from the product. To prove that this concept is feasible, a distillation column (Fig. 11) was setup, as described in the Experimental section. In particular, vacuum distillation (temperature: <150 °C) was chosen to suppress the isomerization reactions of the substrate, but also for efficient product and substrate separation. The results of continuous distillation over 8 h are shown in the ESI. † A distillate stream (top fraction) with a composition of ∼94.2 wt% 1-decene and 5.8 wt% decane/decene isomers can be formulated. The regioselectivity (l/b ratio) of decene in the feed and at the top of the column remains constant at a value of ∼6 wt%. Therefore, the isomer portion of the feed, when completely separated, is decisive for determining the quality of the distillate stream. Therefore, the general feasibility of the process concept is evident. ## Chemicals In Table 5, all the chemicals used in this work are listed. All the solvents and substrates were degassed before the reaction, but not purified further. The purity of each component was tested via gas chromatography. ## Analytics A gas chromatograph from Agilent Technologies (Type 7890A) with an HP-5 column (30 m × 0.32 mm × 0.25 m) and a flame ionization detector (FID) was used for gas chromatographic measurements of the offline samples from the batch and continuous experiments. n-Dodecane was used as the standard and isopropanol, the solvent. For measuring the online samples for continuous experiments, a gas chromatograph (Type 7890A, Agilent Technologies) was used for the gas chromatographic measurements. This system was equipped with a FID. An HP-5 (30 m × 0.25 mm × 0.25 m) was used as the separation column. A sample volume of 1 μl was injected with a split ratio of 1 : 50 with a constant supply of helium as the carrier gas. The contents of rhodium and phosphorus in the liquid phases were quantified via inductively coupled plasma emis-Table 4 Leaching results of the catalyst and ligand for continuous experimentation for 240 h using sulfoxantphos as the ligand The catalyst rates for rhodium and phosphorus were calculated on the basis of the total amount of the related product fraction, the appropriate mass fraction ( ppm) contained, and the initial mass of rhodium and phosphorus at the start of the miniplant. For example, 1% h −1 of leaching represents a loss of 1% of the initial catalyst amount via product flow in 1 h. ## Experimental setup Batch experiments. A 300 ml high-pressure autoclave (Parr Instrument Company) was used to perform the batch experiments. The autoclave was pressure-stable up to 250 bar and was equipped with a blade stirrer. The reactor was used for creating the time curves and performing kinetic experiments for scaling up the reaction. After weighing the catalyst, ligand, polar solvent, and CDs, the autoclave was closed and installed in the reaction equipment. The substrates and solvents were degassed with argon via frit for 30 min. The 1-decene substrate was subsequently transferred to a dropping funnel via a cannula. The dropping funnel, when filled, was pressurized with the system pressure from the reactor. When the predetermined time for preforming was reached, the substrate was transferred into the reactor; thereafter, it was continuously fed with syngas. The Schlenk technique was used for the preparation of all the components and the inert transfer of all the substances into the reactor at all times. A 1000 ml high-pressure autoclave (Büchi) was used to perform further scaling up experiments. The reactor was stable up to 60 bar, including a blade stirrer and baffles. The procedure for carrying out the experiments is the same as that for the 300 ml high-pressure autoclave. The reactor was also used as a continuous stirred-tank reactor (CSTR) for performing continuous miniplant experiments. Continuous experiments. The miniplant comprised a 1000 mL CSTR, which was pressure-stable up to 60 bar. In addition, the reactor was designed with a blade stirrer and baffles. The reactor was connected to a decanter (B2) unit, where the reaction mixture was separated into polar (blue) and nonpolar (yellow) phases. The catalyst phase (blue) was then fed back by a gear pump to the reactor, while the nonpolar phase was collected on a scale. The miniplant was automated and equipped with an online gas chromatograph (Fig. 10). The initial mixtures containing catalyst, ligand, water, and CDs were prepared for all the miniplant apparatuses (reactor and decanter) at once. Therefore, the substances were weighed into several bottles and flushed with argon. The miniplant was tempered and evacuated, and the reactor (B1) was filled with the mixture via a pressure-stable dropping funnel. Then, syngas was added in a stepwise manner in each apparatus and pipe section. At the beginning of the operation, the pump for 1-decene was activated. Simultaneously, continuous syngas feeding and pressure control were started. During the miniplant operation, samples from the product phase were periodically analyzed with an online gas chromatograph from the decanter. Further, offline samples were taken from the product and catalyst phases for ICP analysis, but also for validation of the results obtained from the online gas chromatograph. Distillation setup. Fig. 11 shows the general setup of the distillation column used for the recovery of the nonconverted substrate from the product stream. The column was operated at a pressure of 15 mbar by means of a diaphragm vacuum pump. The product solution from the miniplant (Fig. 10, B2) was conveyed into the column via a storage tank with a solenoiddriven diaphragm metering pump. The feed solution was heated in the column by using a heating rod; the formed steam rose and passed through 15 separation stages that were prepared using Sulzer DX (30 mm) laboratory packs (each pack was 5.5 cm high and had a HETP value of 0.085 m). At the top of the column, the distillate stream passed through a watercooled condenser in which a split was installed in order to set the reflux. The distillate stream was collected in a container. The bottom of the column was designed as a natural circulation evaporator to ensure free mixing. In addition, there is an overflow at the bottom from which the bottom product reaches another container. To protect the bottom of the column from drying out due to the condensed gas, the overflow was equipped with a gas trap. ## Conclusion The Rh-catalyzed aqueous biphasic hydroformylation of 1-decene was successfully conducted only using water as the environmentally friendly solvent and RAME-β-CD as the green mass transfer agent. Initially, two catalyst systems comprising of Rh/TPPTS and Rh/sulfoxantphos were scaled up from a 50 ml autoclave to a 1000 ml high-pressure autoclave to evaluate the feasibility of these phase systems for a potential continuous process application. Herein, both these systems performed efficiently in terms of chemoselectivity (>95%) toward the aldehyde product. However, these experiments reveal that for both these catalyst systems, the reaction rates slowed down in the larger autoclaves; therefore, conversion decreased from 99% to 30% (Rh/TPPTS) and from 70% to 50% (Rh/sulfoxantphos). However, as the chemoselectivity remained high for both these systems, they were tested in two continuous experiments (each over 90 h). These experiments showed that the reaction activity could be drastically influenced by the stirrer speed and n CD /n Rh values inside the reactor. The Rh/sulfoxantphos catalyst system not only performed better in terms of aldehyde yield (45% as compared to that the Rh/TPPTS system, i.e., 30%), but also afforded higher regioselectivities (l/b ratios), i.e., 27, as compared to the value of 1.7; therefore, this system was chosen for the long-term experiment. For a total operation time of 240 h in a continuous miniplant and that of more than 200 h for steady-state operation, the leaching of catalyst was determined to be extremely low, affording ligand loss of 0.0098% h −1 and rhodium loss of 0.0027% h −1 . compared to a commercialized hydroformylation process (Shell) and other recently proposed processes (SILP/CXL and Aqueous TMS) in the literature, a CD-based system does not yield the highest turnover rates in continuous operations, but one of the highest chemo-and regioselectivities and lowest rhodium leaching rates (Table 6). Further, the CD process necessitates very mild conditions (120 °C and 40 bar) as compared to other processes. Reaction rate improvements such as stirrer speed, stirrer type, phase ratios, and CD concentrations are still under investigation and will be discussed in an additional publication in more detail. On the other hand, process optimization was successfully described by the continuous distillation of the product mixture (undecanal, 1-decene, and iso-decene) and recovering nonconverted 1-decene substrate with 94% purity. With the possibility of continuous distillation of the nonconverted substrate, a new process concept was proposed by implementing a second recycling loop into the process. Therefore, combining an aqueous biphasic hydroformylation process using methylated β-CDs for higher olefins with extremely low catalyst leaching and high chemo-and regioselectivities toward the corresponding aldehyde product can be achieved.

chemsum

{"title": "Continuous hydroformylation of 1-decene in an aqueous biphasic system enabled by methylated cyclodextrins", "journal": "Royal Society of Chemistry (RSC)"}

selective_plasmon-driven_catalysis_for_para-nitroaniline_in_aqueous_environments

## Abstract: The plasmon-driven oxidation of amine (−NH 2 ) groups and the reduction of nitro (−NO 2 ) groups on a nanostructured metal surface in an aqueous environment have been reported experimentally and theoretically. The question of which process occurs first in the aqueous environment is an interesting question in the field of plasmon-related photochemistry. Para-nitroaniline (PNA), with both nitro (−NO 2 ) and amine (−NH 2 ) groups, is the best candidate for studying the priority of the plasmon-driven oxidation and the reduction reactions in an aqueous environment. Using surface-enhanced Raman scattering (SERS) spectroscopy, our experimental results and theoretical simulations reveal that PNA is selectively catalyzed to 4,4′-diaminoazobenzene (DAAB) through the plasmon-assisted dimerization of the nitro (−NO 2 ) group into an azo group in an aqueous environment. This indicates that the plasmon-driven reduction of the nitro (−NO 2 ) group clearly occurs before the oxidation of the amine (−NH 2 ) group in an aqueous environment. The plasmondriven reduction of PNA to DAAB is a selective surface catalytic reduced reaction in aqueous environment. Since the first report of the plasmon-driven oxidation reaction in which para-aminothiophenol (PATP, see Fig. 1a) is catalyzed to p,p′ -dimercaptoazobenzene (DMAB, see Fig. 1b) in 2010 1,2 , the realization of plasmon-driven chemical reactions has been one of most important advances in the field of nanoplasmonics , which is also called plasmon chemistry 8 . Hot electrons , which generated from plasmon decay, play an important role in the plasmon-driven catalytic reactions. In the plasmon-driven oxidation reactions, the neutral potential energy surface can become negatively charged as hot electrons temporarily attach to molecules, and then decrease the reaction barrier. Furthermore, the kinetic energy of hot electrons can be efficiently transferred to molecules to provide energy for the reduction reaction. In addition to the above advantages of plasmon-driven oxidation reactions, hot electrons can serve as the required electron source in plasmon-driven reduction reactions. It has been reported that DMAB can be catalytically produced from PATP through a plasmon-driven oxidation reaction 1,2 or from 4-nitrobenzenethiol (4NBT, see Fig. 1c) through a plasmon-driven reduction reaction 5,6 . Several recent review papers are available on this topic . However, it is not clear what role surface plasmons play in the priority of these catalytic reactions. Para-nitroaniline (PNA, see Fig. 1d) contains both nitro (− NO 2 ) and amine (− NH 2 ) groups, making it the best candidate for studying the priority of the plasmon-driven oxidation and the reduction reactions. Three new molecules (see Fig. 1e-g) could be produced from PNA according to the priority of plasmon-driven chemical reactions. If plasmon-driven oxidation reactions occur first, then 4,4′ -dinitroazobenzene (DNAB) can be produced. However, if plasmon-driven reduction reactions occur first, 4,4′ -diaminoazobenzene (DAAB) can be produced. The third possibility is the occurrence of simultaneous oxidation and reduction reactions that produce 4-nitro-4′ -aminoazobenzene (NAAB). This question is a highly interesting topic in plasmon-driven catalytic reaction in the aqueous environment. In this reports, the plasmon-driven chemical reaction of PNA in an aqueous environment was investigated experimentally using electrochemical surface-enhanced Raman scattering (SERS) spectroscopy, and simulated with density functional theory (DFT). Our results revealed that PNA was selectively reduced to DAAB with the help of surface plasmon in the aqueous environment. ## Results and Discussion The roughened Ag electrodes and less roughened Ag electrodes can be observed on the nanoscale in the SEM image in Fig. 2(a,b), respectively. Due to the larger hot sites in the roughened Ag electrodes, there are stronger electromagnetic enhancements for the detecting Raman signals for the catalytic reactions. The Raman spectrum of the PNA powder, which was measured as a reference, is shown in Fig. 3(a). The absorption spectrum of PNA can be seen from Fig. 3(b). Furthermore, we also measured the electrochemical SERS spectra of PNA on the roughened Ag substrate in the aqueous environment (see Fig. 3(c,d)). From Fig. 3(b), it can be concluded that SERS spectra excited at 532 nm are normal Raman spectra. Figure 3(c) shows potential-dependent SERS spectra of PNA in an aqueous environment, where the potential varies from 0 V to − 1.2 V. When comparing Fig. 3(a,c), the SERS profiles are significantly different. Plasmon-driven chemical reactions appeared to occur through the amine (− NH 2 ) and/or nitro (− NO 2 ) groups of PNA in the aqueous environment during the spectral measurements. To determine the reactions that occurred in Fig. 3(c), the Raman spectra of the DNAB, DAAB and NAAB powders were measured excited at 532 nm, as shown in Fig. 4. To provide more evidence, theoretical To verify the stability of the surface plasmon-assisted reduction of PNA to DAAB and the influences of the aqueous environment, two control experiments were performed. The electrode potential was scanned backwards from − 1.2 V to 0 V, resulting in very stable SERS spectra (Fig. 3d). Next, the pH value of the aqueous environment (Fig. 7) was changed, which showed that the reduction is very stable and does not prevent the PNA from being catalyzed to DAAB on the roughened Ag substrate excited at 532 nm. Thus, our experimental results suggest a new method for synthesizing stable, new molecules using surface plasmon resonance. Although PNA is not catalyzed to DNAB through the amine group or to NAAB through the nitro and amine groups, it is selectively catalyzed to DAAB through the nitro group. This clearly answers question presented in the title and demonstrates that plasmon-driven reduction occurs before plasmon-driven oxidation in aqueous environments. However, in an ambient environment, the plasmon-driven oxidation reaction via the amine group (− NH 2 ) is favored over the plasmon-driven reduction reaction via the nitro group 1,17 . Thus, the plasmon-driven oxidation and reduction reactions can be controlled by the external environment. This phenomenon is explained by the abundant O 2 on the substrate, which plays core role in the oxidation of the amine groups in ambient environments 15 . In contrast, amine oxidation is greatly restrained in an aqueous environment. The reduction of the nitro group is usually triggered by the hot electrons, which were generated from plasmon decay 5,25 , which would not be greatly affected by the reaction environment. Therefore, the reduction process of the nitro group is favored over the oxidation process of the amine group in an aqueous environment. It is interesting that the DAAB is not further catalyzed to a new polymer (DAAB)n (where n stands for the unit number), with different DAAB units via two amine groups (− NH 2 ) of DAAB in an aqueous environment. To explain this finding, we measured the potential-dependent electrochemical SERS of DAAB on the roughened Ag substrate in an aqueous environment excited at 532 nm (Fig. 8). It was found that the profile of the potential-dependent electrochemical SERS is the same as that of the DAAB powder in Fig. 4(a). Specifically, DAAB would not be catalyzed to a (DAAB) n polymer via the amine groups by surface plasmon resonance in an aqueous environment. A comparison of the calculated Raman spectra of DAAB in (Fig. 9a) and (DAAB) 2 in Fig. 9(d), together with the SERS of DAAB at − 1.2 V (Fig. 9c) and the Raman spectra of DAAB (Fig. 9b), provide further evidence that the DAAB in an aqueous environment cannot be further catalyzed to a DAAB polymer. This result could be interpreted as the prevention of DAAB polymerization into (DAAB) n due to the newly formed azo group (N = N). To reveal the relationship between the measured reduction processes and surface plasmon resonance, two additional experimental works were performed. To demonstrate contribution the roughness of substrate for the plasmon-driven catalytic reaction, we also measured SERS of PNA (see Fig. 10) on the less roughened Ag substrate (Fig. 2b) excited at 532 nm and the stronger power of 1.5 mW. It is found the intensities of electrochemical SERS spectra on the less roughened substrate are weak at different potentials, though it the plasmon-driven chemical reaction can occur when potential is less − 0.5 V. When the potentials are further increased from − 0.6 V to − 0.8 V, the Raman SERS spectra are too weak to be observed. So, the degree of roughness of substrate is a very important factor for plasmon-driven chemical reaction. The plasmon intensity is dependent on the degree of roughness of substrate. The larger roughness of Ag substrate, the stronger plasmon resonance is. So, this is an experimental evidence for the relationship between surface plasmon resonance and catalytic reaction. We, furthermore, measured the electrochemical SERS of PNA (see Fig. 11) on the roughened Ag substrate at different voltages excited at 785 nm, where the laser intensity is 3.0 mW. The sequence of measurements is from 0 V to − 1.2 V. It is found that there are no any chemical reactions, even under the help of external electric voltages. The reason is that the peak of surface plasmon resonance is far from 785 nm 13 . So, even stronger intensity of laser, the weak plasmon intensity excited at 785 nm can not drive catalytic reactions. This is the second experimental evidence for the relationship between surface plasmon resonance and catalytic reaction. Though, there is no any chemical reaction for PNA on roughened Ag substrate at different potentials, excited at 785 nm, it is a good way to ascertain the SERS of the reactant of PNA, see the comparisons between normal Raman spectra of PNA with the electrochemical SERS spectrum at − 0.6 V (see Fig. 12). Note that, the SERS spectra of the reactant of PNA excited at 532 nm (in Fig. 3a,b) can not be observed, due to the strong plasmon intensity. It is also the advantages of electrochemical SERS for ascertaining the SERS of the reactant of PNA, due to the weak surface plasmon resonance, excited at 785 nm. The priority of the plasmon-driven oxidation and the reduction reactions in an aqueous environment was experimentally and theoretically investigated, using electrochemical SERS spectroscopy and density functional theory. The electrochemical SERS spectra of PNA (with both nitro (− NO 2 ) and amine (− NH 2 ) groups) and theoretical simulations revealed that PNA is selectively catalyzed to DAAB through the plasmon-assisted dimerization of the nitro (− NO 2 ) group into an azo group in an aqueous environment. This indicates that the plasmon-driven reduction of the nitro (− NO 2 ) group clearly occurs before the oxidation of the amine (− NH 2 ) group in an aqueous environment. Two additional experimental measurements were performed to reveal the relationship between the measured reduction processes and surface plasmon resonance. ## Method The PNA, NAAB and DAAB were purchased from Aldrich Chemical Co., Sigma Co. and Alfa Co., respectively. DNAB was synthesized by Beijing Kaida Co. The SEM images of the Ag substrates were obtained using a Hitachi S-4800 microscope. The Raman spectra of the PNA, DAAB, NAAB and DNAB powders were measured using microprobe Raman system RH13325 spectrophotometer. The Ag electrode (a single-crystal silver rod of 99.99% purity) was polished with emery paper and then was carefully cleaned with the Milli-Q water in the ultrasonic bath. And then, the Ag electrode was put into the electrochemical cell, in which the solution of 0.1 M Na 2 SO 4 was used for roughening the Ag electrode. Their Raman spectra were measured using the microprobe Raman system RH13325 spectrophotometer. The voltages of working electrode were controlled by the electrochemical instrument (CHI619B). The samples were excited with 532 nm and 785 nm lasers with an effective power of 0.3 mW, respectively where the 50× objective was used. The theoretical simulations of these molecular Raman spectra and their vibrational modes were calculated using with density functional theory (DFT) 26 , the 6-31G(d) basis set, and the pw91pw91 functional 27 . The calculated Raman spectra were scaled according to ref. 28. All the calculations were done with Gaussian 09 software.

chemsum

{"title": "Selective plasmon-driven catalysis for para-nitroaniline in aqueous environments", "journal": "Scientific Reports - Nature"}

exploiting_modeling_studies_for_evaluating_the_potential_antiviral_activities_of_some_clinically_app

## Abstract: Here, we theoretically modeled the binding interaction of the Sars-CoV2 (Spike protein) utilizing molecular docking with some potential repurposed antiviral medications and two botanical products (Curcumin and Quercetin). Molecular docking between the drugs and the Sars-CoV2 proteins reflecting the pure electrostatic forces and H-bond formation is complemented with the DFT results that shed light on the electronic nature of the interactions. Moreover, DFT computations provide invaluable information about the drug reactivity indices calculated from the energies of the frontier orbitals. The DFT results indicate intermolecular electron donor-acceptor interaction besides the H-bond formation. Most of the considered medication molecules act as electron-sink candidates except EIDD-2801, the electron donor. The theoretical results show the high possibility of blocking the human cellular entry against Sare-Cov2 or weakening Sars-Cov2 activity due to the electronic donor-acceptor interactions. The findings are solely computational analysis and need to be corroborated by additional studies. ## Introduction Coronavirus 2019 (COVID-19) has become a life-threatening global concern in recent days. It causes illness and death in some patients, but no specific treatment is available. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes coronavirus, has significantly impacted human health and socioeconomic status globally . SARS-CoV2 is a positive-sense single-stranded RNA virus that belongs to the betacoronavirus genus, presumed to have zoonotic origins since it has high genetic similarity to the coronavirus RaTG13, isolated from bats . Furthermore, SARS-CoV-2 is structurally described . Morphologically, the virus consists of four structural proteins, namely Spike (S), envelope (E), membrane (M), and nucleocapsid (N), among which the former three are integral membrane proteins, and the latter remains complexed with its RNA genome. In addition, it comprises 9 or 10 accessory proteins . The spike (S) glycoprotein is a critical component of viral infection. It adheres to the host cell's surface receptor, human angiotensin-converting enzyme 2 (ACE2), allowing viral cellular entry via endosome formation and plasma-membrane fusion . The spike subunits called S1 and S2 are responsible for binding with the host-cell receptor and membrane fusion, respectively . Different variants have been detected but do not significantly influence the efficacy of the developed vaccines . In other concise words, SARS-CoV-2 is an RNA virus branded by crown-like spikes on the external surface . Any SARS-CoV-2 virus infection and replication cycle step is a potential target for antiviral interference. The binding of the viral spike protein of SARS-CoV-2 to the human cell surface receptor ACE2 is a critical step during the entry into human cells, causing infection. Hence, cell entry inhibitors could be used to prevent the SARS-CoV-2 virus and reduce the progression of COVID-19 diseases by preventing virus particles from infecting human cells. Endemic seasonal coronaviruses cause morbidity and mortality in a subset of patients, but no specific treatment is available. Here we investigate six drugs and two natural food supplements, promising antiviral medications for treating SARS-CoV-2 infection. This study evaluates the potential of repurposing the medicines for treating seasonal human coronavirus disorders. Several drug-repositioning studies and compound evaluations have been conducted to discover novel antiviral drugs against SARS-CoV-2 utilizing experimental and theoretical/computational methodologies . The drug-repurposing effort considered herein focuses primarily on agents known to be effective against RNA viruses, including SARS-CoV and MERS-CoV. Thus, we have selected the drugs mentioned hereafter: -N-Hydroxy-5'-O-isobutyryl-3,4-dihydrocytidin (EIDD2801), N-Hydroxycytidin (Eidd-2801 metabolite or EIDD-1931), also known as Molnupiravir. Both were found to be effective against COVID-19. Molnupiravir is an orally available antiviral drug candidate currently in phase III trials to treat patients with COVID-19. Molnupiravir increases the frequency of viral RNA mutations and impairs SARS-CoV-2 replication in animal models and humans. -Favipiravir (favilavir), RdRp is a purine nucleoside that acts as an alternate substrate leading to inaccurate viral RNA synthesis used in viral infections. -New favipiravir derivative, (E )-N -(4-cyanobenzylidene)-6-fluoro-3-hydroxypyrazine-2carboxamide (cyanorona-20), as the first potent SARS-CoV-2 inhibitor with very high selectivity (209-and 45-fold more potent than favipiravir and remdesivir, respectively) . -Remdesivir is a nucleotide analog that may block viral nucleotide synthesis to stop viral replication (Ebola virus infection). -Fluvoxamine compound demonstrating moderate or vigorous [Cytochrome P450 (CYP) is a hemeprotein that plays a crucial role in the metabolism of drugs and other xenobiotics] CYP inhibition by reversible or a combination of reversible plus time-dependent inhibition. The current study aims to model, via molecular docking, the binding energy and types of interactions of the Sars-Cov2 protein and ACE2 in human bodies shown in sketch (1) with the aforementioned antiviral medications, some of which are potentially tested by pharmaceuticals companies and newly suggested candidates, including natural products. Searching for more potent natural anti-COVID19 drugs is very demanding. . Medicinal plants, especially those employed in traditional Chinese medicine, have attracted significant attention because they include bioactive compounds that could be used to develop legal drugs against several diseases with no or minimal risks . We have chosen Flavonoid quercetin, an antioxidant and signaling molecule, and an effective Zinc ionophore . Moreover, Curcumin is a medical herb of a powerful antioxidant, anti-inflammatory , antimutagenic, antimicrobial , and anticancer properties . ## Sketch (1). A representation of the interaction of Sars-Cov2 protein with ACE2 in human bodies. We utilized two computational drug design methodologies to propose an efficient blocking and viable treatment for COVID-19. Moreover, molecular docking results are complemented with a simplified approach of selecting the presumed most reactive peptide-sequence (b6) of the Spike of the S-Sars-Cov2 proteins using the DFT quantum chemical method to shed light on the electronic nature of the interactions utilizing a non-expensive way of drug trusting. Furthermore, DFT computations enable us to obtain invaluable information about the drug reactivity indices. Electronic and other weak bonding factors strengthen interaction forces by introducing multiple hydrogen bonds, covalent binding, additional van der Waals forces, or multivalent binding. Here we focus on the antiviral strategy of interfering with interactions involving host-guest sites to enhance the protective immune response against Sars-Cov2. However, although drugs targeting host proteins may be potentially cytotoxic or impact the human immune system, their appeal cannot be ignored . Implementation of this medicinal chemistry strategy is expected to pave a direct way to discovering new drugs effective against current and future threats due to emerging and re-emerging viral pandemics. ## Methodology and Materials All the DFT calculations, including the optimization of ground-state geometry, are carried out using the Spartan 20 program. We applied the molecular mechanics' built-in Spartan 20 package to reach the interacting molecules' equilibrium geometry. We then applied single-point computations within the DFT theory to the molecular mechanically optimized equilibrium geometry using the long-range corrected hybrid density functionals, the wB97X-D functional , which includes empirical damped atom-atom dispersion corrections with 6-31G(d) basis set. The method used is efficient and cost-effective for seizing knowledge throughout biological hierarchies. We used the reaction field model for solvation (water) [the conductor-like polarizable continuum model (CPCM)], achieving a PCM (polarized continuum model) calculation . We used MOE 2015.01 package for docking the inhibitors to Sars-CoV2 protein and ACE2 that we obtained from Protein Data Bank (Crystal structure [6M0J]) . We downloaded chemical structures of the ligands from ChemSpider structure search or as indicated in Table 1. To find effective drugs against S-Sars-Cov2, we searched for high-scoring value in docking studies involving different drugs. The structures of the medication molecules are shown in Figure 1. ## Chemical structures of Drugs 2D chemical structures of the chosen drugs are given in Scheme 1. Moreover, we include the ID of each drug in Table 1. ## EIDD-2801 EIDD-1931 Remdesivir Fluvoxamine ## Favipiravir cyanorona-20 Quercetin Curcumin Scheme 1. 2D Chemical Structure of the medications ## Electronic factors. DFT computational cost of large proteins is challenging and prohibitive. To overcome the computational barrier of applying DFT with the limited computational cost available, and to enable us to look at the electronic factors, we will simulate the chosen drugs' influence on the Spike protein of SARS-CoV-2, which was recently identified as the beta-sheet sequence (𝛽6) that contains the topmost of the communicating protein strand SARS-CoV-2 that binds to a selected protein strand of ACE2 . Reactivity indices of the drug molecules will be first obtained from the energies of the frontier orbitals (EHOMO and ELUMO), by applying DFT method, according to the following equations: HOMO and LUMO energies are valuable in analyzing the chemical potential (µ), hardness (η), and electrophilicity (ω) that reveal the electronic donor-acceptor reactivity of the individual ligands and are crucial in assessing their reactivities toward proteins of the virus and the human receptor . The results are summarized in Table (1). The more the ligand will be reactive, the more it will participate in electronic interactions and bond formation with protein. The electrophilicity index (𝜔 = 𝜇 2 /(2 * 𝜂) measures a molecule's ability to act as an electron sink-the higher its value, the stronger the ligand's electron sink property. The electrophilicity index provides a magnitude of the energy stabilization of a molecule when it gains an extra amount of electron density during the interaction. The electrophilicity index comprises the tendency of an electrophile to earn an excessive amount of electron density, given by the chemical potential µ and the resistance of a molecule to exchange electron density during an interaction, provided by the hardness h. Therefore, a suitable electrophile is characterized by an excessive µ value and a little h value. The electrophilicity index rendered a powerful apparatus for studying the reactivity of drug molecules . 𝜔 values of all studied molecules except EIDD-2801 are higher than ACE2 and b6 sheets. Thus, electron donor-acceptor affinity of the proteins -drugs increases with increasing the value of 𝜔. However, the electrophilicity fails to explain the orbital localization in the case of proteins interacting, as shown in Table 1 and Figures 1. In this case, the hardness and chemical potential values are consistent with the donor-acceptor affinity visualized in Figure 1, where the HOMO -LUMO interactions reflect the donor-acceptor phenomenon of the proteins with each other. Figure 2 visualizes the electron acceptor property of the molnupiravir in the presence of both proteins. In other words, the medication is the electron sink agent that initially inhibits illness with covid-19. Other drugs behave similarly. This electronic situation leads us to assume a mechanism of blocking human cell entry or at least inhibits the electronic activity of the Sars-Cov2. ## Figure 1. HOMO is wholly localized on the Sars-Cov2 (reactive peptide-sequence (b6) of the Spike of the S-Sars-Cov2 protein). The LUMO (dashed) is mainly from ACE2 Sheet (red ribbon). H-bonds are shown as dotted lines. ## Docking modelling To further widen the perspective, the current study modeled the binding interaction of the S protein and ACE2 in human bodies via molecular docking with these repurposed antiviral medications, including natural products. Our molecular docking score results showed (Table 2) ## SARS-Cov2 (lower left pane), and the SARS-CoV2-ACE2 adduct (right-hand side). The tested drugs reveal that the driving force of interactions is through specific forces such as Hbond formation and the sum of the electrostatic and van der Waals energies (Table 3). The sum of all these interactions is approximated by a docking score, which represents the potentiality of binding in kCal/mol. The driving forces for these specific interactions in biological systems target complementarities between the shape and electrostatics of the binding location and the ligand or substrate. The effective site of noticeable contribution is with several amino acids of the b6 antennae of the Spike of the S-Sars-Cov2 proteins as indicated in Figures 3 and 4 and Table 3. Other proposed drugs and the food supplements Curcumin and quercetin are assumed to behave towards the virus in a similar binding and electronic way. A high docking score points to the higher potentiality of remdesivir and Merck medications in blocking S-Sars-Cov2 or ACE2, thus preventing or treating Covid-19 infections via binding with the individual proteins adduct ACE2 and S-Sars-Cov2. Good binding affinity and reactivity response suggested that the studied medications can be promising drugs to inhibit the COVID-19 invasion. As Table 3 shows, we can conclude the importance of some amino acids (highlighted) that belong to the antenna of the spike virus in binding with the target medication. Remdesivir, Molnupiravir (EIDD-2081), and Quercetin showed significant binding interactions with the antenna of the SARS-Cov2. Other drugs interact mainly with the ACE2 amino acids, blocking their receptor function. ## Conclusion The current study modeled the binding interaction of the S protein and ACE2 in human bodies via molecular docking with some well-known repurposed antiviral medications, including natural products. We utilized two computational drug design methodologies to identify the origins of the The findings are solely computational analysis and need to be supported by additional clinical studies. Moreover, the results of both computational methods applied, and the simplifications assumed in the case of DFT application are consistent to a great extent.

chemsum

{"title": "Exploiting Modeling Studies for Evaluating the Potential Antiviral Activities of some Clinically Approved Drugs and Herbal Materials Against SARS-CoV-2: Theoretical Studies Towards Hindering the Virus and Blocking the Human Cellular Receptor", "journal": "ChemRxiv"}

a_unified_topology_approach_to_dot-,_rod-,_and_sheet-mofs

## Abstract: We offer insights into how network topology influences stability of metal-organic frameworks and suggest the application of rare sheet-MOFs, where metal ions and linkers form infinite 2D units (SBUs) as a strategy for achieving higher stability. We also demonstrate a unified topology approach to MOFs exemplified by the dot-, rod-, and sheet-MOFs reported. These MOFs are based on vicinal dicarboxylates, which we propose as a way to prepare rod-MOFs. Cyclic voltammetry suggests Ce(IV)-MOFs as being more stable than expected with potential electrochemical applications. ## INTRODUCTION The assembly and structure-directing role of the multi-metallic secondary building unit (SBU) is central to the synthesis and understanding of metal-organic frameworks (MOFs). 1 In most cases, the metal-containing SBU can be reduced to a single point, 0D (for example {Zn 4 O(OCO) 6 } in MOF-5) with a certain geometry (octahedral in MOF-5) and connectivity (6-c) to the organic SBUs. We will refer to this as a dot-MOF. In some MOFs, however, carboxylates, imidazolates, or other Lewis base groups bridge the metal ions in an infinite pattern where it is not possible to define a 0D SBU (Figure 1 center and right). Most notably, we find these in the rod-MOFs, where the metal ions are connected to form an infinite 1D SBU, such as {Mg 2 (O-CO) 2 (O) 2 } N in MOF-74, then further connected into a 3D framework by the organic SBUs. 2 In this report, we introduce the concept of infinite 2D SBUs. To date, these are rare in MOF chemistry, but our findings indicate that they are important. We will refer to them as sheet-MOFs, and the three classes are compared in Figure 1. For some MOF applications, the stability, chiefly thermal or resistance to degradation in solution 3 (i.e., water at high or low pH) as well as mechanical stability, will be a crucial parameter for real-world devices. Although there are no comprehensive systematic stability studies of a large sampling of different MOFs, indications are that rod-MOFs are more thermally stable than dot-MOFs. 4 We note that the rod-MOF topology was recently credited with the hydrolytic stability of the water-harvesting MOF-303, 5 and that the rod-MOF MFM-300(Cr) was recently branded as ''ultrastable.'' 6 By analogy, we suspect that sheet-MOFs are even more stable than rod-MOFs, principally because the higher the dimensionality of the SBU, the less likely it is to have a strained structure especially around the metal sites. ## The bigger picture Metal-organic frameworks is an emerging class of materials with wide-ranging capabilities due to tunability, high surface area, porosity, and large internal void spaces. However, stability may be an issue. Here, we argue for increased stability going from discrete multi-metal entities to infinite rods to sheets as building units and propose a unified way to classify them. We present a rare sheet-MOF based on non-dense 2D net, opening the way to even more stable MOFs while preserving the potential of high porosity and surface area. Cyclic voltammetry on this Ce(III)-based MOF CTH-15 indicates a remarkable stabilization of the Ce(IV) oxidation state and that more Ce(IV) MOFs may be feasible. Potential applications of this, or similar MOFs, in electrochemical applications related to drinking water purification are discussed. For thermal tolerance, the sheet-MOF performance is indicated by the remarkable stability of ULMOF-1, up to 610 C under a nitrogen atmosphere. 7 In this compound, sheets of edge-and corner-shared lithium-oxygen tetrahedra connect to a dense sheet bridged by 1,4-benzene dicarboxylate (bdc) organic SBUs. Another example is the Na + -based MOF-705 and MOF-706. Their sheet-MOF properties were used to explain the unusual stability. 8 There are good reasons to be observant of this general classification. However, a more precise description of topologies is essential in reticular chemistry synthesis planning, 1 analysis, and communication of the resulting structures. Properties such as porosity and flexibility have recently been investigated using the topology approach, 14 and a special case of folding topologies has been identified. 15 However, rod-MOFs, and now sheet-MOFs, pose a particular problem because these are commonly analyzed in a different way from dot-MOFs. 2,16,17 A unified approach would be advantageous both from a synthesis and an understanding perspective. Note, however, we do not in any way seek to supersede or diminish current efforts to generate solid algorithms for the automatic analysis of the large amount of crystallographic data now available 18,19 ; we find these absolutely essential. We approach the topology question from the vantage point of three MOFs prepared with benzene-1,2,4,5-tetracarboxylate, H 4 btec: a dot-MOF in H 2 NMe 2 [Y(btec)(H 2 O)] CTH-14, a sheet-MOF in [Ce 3 (btec)(Hbtec)(OAc)(HCO 2 )] CTH-15, and a rod-MOF in 4,4 0 -azopyridinium[Gd 2 (btec) 2 ] CTH-16. From the MOF design point of view, we note that btec, with its vicinal 1,2-dicarboxylates, seems prone to form rod-MOFs, and this synthesis strategy will be elucidated further. As for rare-earthelement (REE)-based MOFs, these are fascinating for their luminescence properties, and the high coordination numbers of the REE ions make for diverse geometries and unprecedented network topologies. 20 Moreover, H 4 btec is luminescent on its own, and we report on how this property is modulated by the REEs in these solids. More importantly, MOFs have been identified as catalysts for different reactions related to environmental issues. 21 Specifically, Ce-MOFs were demonstrated as effective catalysts for combustion of volatile organic compounds in air, 22 and Ce(IV/ III)-MOF catalysis was recently proposed for hydrolysis of nerve agents. 23 We have investigated the redox properties of the Ce-and Gd-MOFs CTH-15 and CTH-16 by cyclic voltammetry, and our results suggest a stabilization of Ce(IV), normally a strong oxidant. Benzene-1,2,4,5-tetracarboxylate, H 4 btec (Figure 2) is a somewhat odd MOF linker because the two carboxylate groups on each side are in close proximity. This looks inviting for synergistic coordination in which the metal-ion attachment to one carboxylate influences the coordination of the second one. There are 350 such MOFs or coordination networks reported in the Cambridge Crystallographic Database (CSD) to date, starting with a Ce(III)btec compound in 1997. 24 ## RESULTS AND DISCUSSION We start with synthesis and structures, then move on to thermal characterization, chemical stability, photoluminescence, and electrochemical properties. In the final part, the discussion draws from all these parts, touching on synthesis, flexibility, and topology. Figure 3 gives an overview of the structures of CTH-14-CTH-16. ## Synthesis and structures of CTH-14, CTH-15, and CTH-16 The solvothermal reaction of H 4 btec, 4,4 0 -azopyridine and Y(NO 3 ) 3 $6H 2 O in a 1:1:2 (v/v/v) DMF/H 2 O/glacial acetic acid solution in a pyrex tube at 120 C produced colorless single crystals of CTH-14, H 2 NMe 2 [Y(btec)(H 2 O)] after 24 h. A similar method was applied in the making of CTH-15, [Ce 3 (btec)(Hbtec)(OAc)(HCO 2 )] and CTH-16, 4,4 0 -azopyridinium[Gd 2 (btec) 2 ] with the exception that the metal salts were Ce(NO 3 ) 3 $6H 2 O and Gd(NO 3 ) 3 $6H 2 O, respectively. In CTH-15, colorless crystals were obtained after 3 days, and pale yellow crystals of CTH-16 were acquired on the second day of the solvothermal reaction. CTH-14 and CTH-15 could also be prepared without the presence of 4,4 0 -azopyridine. Table S1 details the crystallographic data and structure refinement parameters of CTH-14, CTH-15, and CTH-16, and we also comment on the refinement processes of CTH-15 and CTH-16. A short summary of the crystallography is provided below. Figures S1-S5 give some more detail of the structures. When the structures are discussed, it will be clear that neither of these compounds are expected to show any remarkable surface areas or porosities. The gas sorption results are therefore only briefly presented and discussed after each structure. More details are provided in Table S6. Note the angles between the carboxylate group and the benzene plane (dashed lines) that are further investigated in Figure 6. The (H 4 btec) ligand offers two carboxylate groups in close vicinity and therefore may be prone to the formation of infinite SBUs. As expected, this compound shows no porosity in the gas sorption experiment. [Ce 3 (btec)(Hbtec)(OAc)(HCO 2 )]: CTH-15 All tested crystals of CTH-15 had similar lattice constants and were potentially pseudomerohedral twinned. This compound, whose structure solved in the monoclinic space The MOFs were obtained using solvothermal conditions from H 4 btec and mononuclear metal precursor salts in DMF with or without the 4,4 0 -azopyridine (except CTH-16). This gives MOFs with the same organic SBU (btec part shown in green) and three distinct carboxylate-bridged coordination entities (shown in cyan), which in turn can be thought of as a 0D metal SBU and infinite 1D or 2D metal SBUs. This gives one dot-MOF, one rod-MOF, and one sheet-MOF whose network topologies were analyzed. Green and yellow spheres indicate nodes in the 0D SBUs, and black spheres nodes in the network using the points-of-extension approach (vide infra). More details are given below and in Figures S1-S4. group P2 1 /c with a very small monoclinic angle of 90.011(2) , has three symmetry-independent Ce atoms, bonding 10-12 oxygen atoms each, and two btec ligands in different protonation states, a common occurrence for Ln-btec-MOFs. Formate is a common breakdown product of the DMF solvent and acts as a non-bridging, coordinating anion. This Ce-MOF has an unusual network for a MOF in that an infinite 2D SBU is formed (Figure 3). The calculated porosity of CTH-15 with solvent removed is 38% with spherical cavities of diameter 4.2 A ˚using CrystalMaker (see supplemental information for details) or 20.7% using the more common Platon approach. However, neither the porosity by N 2 sorption at liquid N 2 temperature nor CO 2 and N 2 sorption at ambient temperature (20 C) could be verified possibly because channel diameter may be too narrow to detect, 25 or, as indicated by a solvent accessible void calculation, the voids are more isolated cavities than channels (see supplemental information for details). Powder diffraction indicates that CTH-15 retains its network structure after activation in vacuum and the gas sorption experiment. 4,4 0 -azopyridinium[Gd 2 (btec) 2 ]: CTH-16 4,4 0 -azopyridinium[Gd 2 (btec) 2 ] CTH-16 crystallizes in the triclinic space group P1 with its asymmetric unit containing one nine-coordinated Gd center, two half btec linkers, and one half 4,4 0 -azopyridium. In analogy with the isoreticular series (4,4 0 -bipyridinium)[Ln 2 (btec) 2 ] Ln = Pr, Eu, Gd, 26 we assign the missing charges to protonated pyridine (O4.H-N4 2.394 A ˚) rather than a protonated btec ligand based on their presumed pK a values although the proton cannot be identified from the electron density map. The 4,4 0 -azopyridinium ion itself is situated around a special position in the channels of CTH-16. It was isotropically modeled as a fragment obtained from CSD structure KUGJUS 27 because it is disordered, which is frequently the case. Of 152 structure determinations in the CSD with organic co-crystals of 4,4 0 -azopyridine, more than 40% display disorder or other problems. The identity of the 4,4 0 -azopyridinium guest was moreover confirmed by luminescence measurements (Figure 4). There are no apparent voids in the structure, Article and gas sorption measurements did not show any porosity for this compound. As shown in Figure 3, CTH-16 is a typical example of a rod-MOF. Having established the structures, we now investigate some properties of these materials. Thermal analysis, thermal stability, and powder X-ray diffraction CTH-14-CTH-16 were studied by thermogravimetric analysis (TGA) under air and exhibited different thermal behaviors. In the thermogravimetric curve of CTH-14 (Figure S12), the first two mass losses were associated to the loss of the dimethylamine (HNMe 2 ) (found 10.4%, calculated 11.2%); the rest corresponded to the loss of the coordinated water molecule and the decomposition of the linker (btec). The CTH-15 TGA curve is illustrated in Figure S13 and shows a first weight loss at 8.80%, which is the release of mass corresponding to solvent (calculated 8.1%). The second weight loss indicates the decomposition of the linker, acetate, and formate to give a residue of 47.7% CeO 2 (calc. 46.0 %). TGA analysis of CTH-16 (Figure S14) gave one weight loss, which means that 4,4 0 -azopyridinium is tightly held in the CTH-16 pores. Although constant weight during a TGA experiment is no proof of architectural stability, it is still an indication of the material's overall thermal stability. 4 For our compounds, the onset of thermal degradation is clearly some 20 lower for CTH-14 compared with CTH-15-CTH-16 that show close to identical starting points of the mass loss at around 400 C. This is consistent with the idea that rod-and sheet-MOFs are more stable than dot-MOFs such as CTH-14. The PXRD patterns of CTH-14-CTH-16 were in good agreement with their calculated PXRD patterns, as demonstrated in Figures S15-S17. PXRD patterns of these three MOFs were collected after N 2 sorption experiments, indicating that all samples remain intact after activation except for CTH-16. PXRD was further used to explore the thermal architectural stability of the materials (Figures S18-S20). Heating up to just under the decomposition temperature as indicated by the TGA experiment, 350 C, shows the dot-MOF CTH-14 losing architectural integrity and transforming to another unknown crystalline phase already after the 120 vacuum activation for the gas sorption experiment. The sheet-MOF CTH-15 and rod-MOF CTH-16 on the other hand are stable up to at least 350 C. The compounds do differ in some substantial ways other than the dimensionality of the SBUs, so to make the case for a stability trend dot < rod < sheet, we need to look at more data. Recently, Bennett and co-workers 4 reported a study of MOF thermal stability. They based their analysis on the onset of decomposition from the plateau phase in the TGA, and more than 200 MOF TGA experiments were taken into account. We picked out and analyzed the topology of the metal SBUs for the most stable MOFs in this study with a T d exceeding 500 C (see Table S4). We found that of these 10 thermally stable MOFs, there were one dot-MOF, six rod-MOFs, and three sheet-MOFs, with the sheet-MOF ULMOF-1 taking the record, being stable up to 610 C as already mentioned. Although this corroborates our idea, we acknowledge that this cannot be the only factor deciding the thermal stability. In this respect, we note that the only MOFs stable enough to melt are a couple of ZIFs. ZIF-4 melts at 567 C and ZIF-62 at 410 C and is stable for another 100 C. ## Article This seems understandable in light of the above discussion as ZIFs connected only through their imidazolate units can be thought of as having three-dimensional SBUs. ## Chemical stability Chemical stability of CTH-14, CTH-15, and CTH-16 were probed by soaking the MOFs in water at pH 7, aqueous 5-M HCl, and aqueous 5-M NaOH at room temperature for 1 h (Figures S21-S23). All three MOFs were stable in water for at least 1 h, as their corresponding PXRD patterns remained the same as the as-synthesized products. In 5-M HCl, CTH-14, CTH-15, and CTH-16 maintained their crystallinity although some peak broadening was observed, but their final products did not match their as-synthesized diffraction patterns but rather H 4 btec on its own (Figure S24). Lower chemical stability of these three MOFs were found when using very basic (5-M NaOH) media. There is loss of intensity in the PXRD patterns of CTH-14 and CTH-16 and a rather amorphous final product in CTH-15. This is likely because CTH-15 has a relatively good leaving group (formate, pK a 3.7), which is easily attacked and may provoke the decomposition of the framework (the exchange of formate for other anionic molecules has been demonstrated in, for example, MOF-520 29 ). Ding et al. 3 recently surveyed solution-stability studies of MOFs, and again, we picked the most stable examples, materials reported to survive a week or more under various conditions (see Table S3). Of these, eight are dot-MOFs, another eight rod-MOFs, and one is a ZIF, thus with a 3D SBU. We note that of these, only ZIF-8, MIL-177-LT, and two rod-MOFs were stable under as harsh conditions as we tested CTH-14, CTH-15, and CTH-16. Again, given the predominance of dot-MOFs, this supports that the dimensionality of the metal SBU is a factor also in solution stability. The importance of chemical stability in many chemical applications of MOFs makes it somewhat surprising to note the scarcity of thermodynamic solubility product data. 30 ## Photoluminescence properties Lanthanoid-based MOFs have become popular in the past years because of their potential useful magnetic and photoluminescence properties. These properties are attributed to their geometries, higher coordination numbers compared with transition metals, inherent luminescent abilities, and high number of unpaired electrons. The luminescence properties of CTH-14, CTH-15, and CTH-16, H 4 btec, and 4,4 0 -azopyridine were investigated in the solid state at room temperature using an excitation wavelength of 300 nm. Detailed assignments are discussed in the supplemental information. The emission spectra of the free ligands are displayed in Figure 4. H 4 btec exhibits somewhat structured emission at 370 nm in line with previous studies, 34,35 while 4,4 0 -azopyridine on the other hand shows a less structured emission band at 414 nm. 36 The photoluminescence spectrum of CTH-14 shows emission at 351 nm, whereas CTH-15 has a more intense emission at 400 nm, both consistent with similar compounds 37,38 (Figure S25). The CTH-16 spectrum (Figure 4) is dominated by an emission band at 414 nm, which can be attributed to the 4,4 0 -azopyridine guest. Weaker features at around 370 and 500 nm are also observed that may originate in the btec ligand (Figure S26). This ## Article indicates that 4,4 0 -azopyridine is present in the CTH-16 structure, corroborating the findings from the X-ray investigation. However, these measurements do not appear to be sensitive to whether the 4,4 0 -azopyridine is coordinated to a metal ion or protonated. 39 Compounds CTH-14, CTH-15, and CTH-16 exhibit photoluminescent behaviors that may warrant further investigations in light of the need for devices stable to moisture, air, and thermal degradation, 40,41 especially given the architectural stability of CTH-15 up to 350 C as shown by powder diffraction and the insolubility in common polar and non-polar solvents. ## Electrochemistry The electrochemical behavior of CTH-15 and CTH-16 was characterized by cyclic voltammetry in the solid state using a carbon paste electrode. For CTH-15, the voltammogram shows a quasi-reversible electron transfer reaction, probably related to the Ce(IV)/Ce(III) redox couple in the MOF structure, see Figure 5 (as no corresponding couple was shown for CTH-16 with the non-redox active Gd 3+ ). The electron transfer kinetics is slow as expected for a carbon paste electrode. The redox potential is about 0.2 V versus Ag/AgCl, taken as the mean value between the anodic and cathodic peak potentials. For the Ce(IV)/Ce(III) in aqueous solution, the reversible potential is 1.72 V versus SHE, i.e., 1.52 V versus Ag/AgCl. 42 The large decrease in oxidation potential for CTH-15 means that Ce(III) is destabilized in the structure, i.e., it is more easily oxidized, whereas Ce(IV) is stabilized and more difficult to reduce than its corresponding water complexes. This is in line with previous solution studies indicating a substantial span in redox potentials for the Ce(IV)/Ce(III) couple and in particular stabilization of the Ce(IV) by anionic oxygen ligands. 43, 44 Notably, with the Article octadentate 1-hydroxy-2-pyridinonate ligand 3,4,3-li-(1,2-hopo), the Ce(IV)/Ce(III) potential was measured to be 0.175 V versus Ag/AgCl, 45 close to the Ce(IV)/Ce(III) potential measured for CTH-15. It is also in line with very recent studies on Ce(IV)-MOF-808. 46 Cerium is well established in various electrochemical processes, 47 but studies of electrochemical applications of MOF have been mostly concentrated on batteries and supercapacitors or water-splitting reactions. 48 Given the affinity of cerium ions to nitrate ions, 47 it would be interesting to explore CTH-15 or related systems in catalytic electrochemical nitrate reduction, an important issue due to a pressing need to remove nitrate from drinking water. 49,50 We note that the general requirements for a nitrate reduction cathode as pointed out by Zeng et al. 50 ''liquid channels for reactant delivery and gas channels for product separation'' seem to fit nicely with the MOF concept. 50,51 In this context, the low toxicity of cerium and its relative high earth abundance, on the level of copper, is noteworthy. More comprehensive studies of the electrochemistry of CTH-15 and related systems are underway. Gd(III) in CTH-16 is not redox active, but the guest molecule 4,4 0 -azopyridine could in principle be reduced to the hydrazine. However, the two small redox couples that can be observed in the cyclic voltammogram of CTH-16 (Figure S28) we tentatively attribute to the btec linker 52 because they have no resemblance to the electrochemical behavior of 4,4 0 -azopyridine, also investigated in the solid state (Figure S27), the latter showing two clear peaks for the oxidation and reduction, forming 1,2-di(pyridine-4-yl)hydrazine corresponding to the solution chemistry. This indicates that the guest is shielded in the structure and the reduction, requiring water or protons, cannot take place. We will now discuss these results in a broader context. ## Implications for btec-based MOF synthesis Flexibility The guests, or counter ions, are of slightly different shape and size in CTH-16 and in the isoreticular (4,4 0 -bipyridinium)[Ln 2 (btec) 2 ], meaning that the network has to have some flexibility. For the unit cell, this is expressed by 7% volume increase in CTH-16 and a visibly less tilted network (Figure S4). There are many flexible, or ''breathing'' MOFs, but the classic example is MIL-53, 58 a rod-MOF just like CTH-16 with an sra-net. While there are different topological discussions of flexibility, 14,15 for MIL-53 and its derivatives, a kneecap movement of the metal carboxylates has been identified in the crystal structures. This involves bending of the M 2 O 2 plane of the carboxylate versus the benzene plane, so that for a ''closed'' (GUSNEN01) version of MIL-53 the average angle is 30 and for the ''open'' (MINVOU) 10 (Figure S6 left). In contrast, the out-of-plane turning of the carboxylates (Figure 2) is very slight for all such compounds, 85% laying within 22 from being co-planar and 10% being exactly co-planar (Figure S6 right). The situation is very different for the btec-MOFs, where there is a larger variety in angles and these angles are pairwise correlated (because of steric effects) (see Figure 6). We suggest that both of these mechanisms are in play simultaneously to render these frameworks flexible. We also suggest that the flexibility of the btec ligand depicted in Figure 6 is partly responsible for the variety of structures observed with the REE ions. We count 17 different topologies in the 43 different 3D btec coordination polymers reported to date (Table S5). 14 ## Article for systems likely to form compounds that should be very similar or even isostructural. The devil is in the details, different solvents, stoichiometries, and methods yielding these diverse results. ## Rod-and sheet-MOF synthesis Rod-MOF structures may be a strategy for increased stability of MOFs, and btectype ligands may be a way to achieve that. The vicinal positions of the carboxylates induce chain formation, and the concerted movement of the CO 2 units (see flexibility discussion) may also be a factor. Experimental indications come from the prominent featuring of btec-MOFs in the 2016 review by Schoedel et al. 2 and also from our analysis of the 43 different REE 3D btec coordination polymers (Table S5). Of these, almost 80% are rod-MOFs, and two are sheet-MOFs. However, due to the shortness of the btec linker, we cannot expect any great porosity or surface area for these. However, extended btec-type ligands, bridging aromatic ligands with double (such as btec) or triple vicinal carboxylate groups are rare in MOF chemistry. So, while [1,1 0 :2 0 ,1 0 '-terphenyl]-4,4 0 ,4 0 ',5 0 -tetracarboxylates with one vicinal dicarboxylate are relatively common ligands (36 MOF structures in the CSD), the one benzene group extension of H 4 btec to 2,3,6,7-anthracenetetracarboxylic acid has but four known MOF structures. 59 We see here an opportunity for new families of MOFs: for example, one candidate metal ion being Ce(IV), which we have shown here to be substantially stabilized compared with its aqueous solution chemistry. Ce(IV) MOFs are previously known, 46, but it is not a large class of MOFs, possibly because of the use of Ce(IV) as a well-known oxidant in organic synthesis. However, Ce(IV) may be more stable than we think, 43,44 . 73 To answer the more general question of whether there are specific conditions favoring rod-and sheet-MOF formation is more difficult. The same parameters that lead to the formation of compounds with non-connected rods and sheets should be good, if these are known. Then, we might just speculate that one difference is that precursors of the dot-MOF may be coordinately saturated discrete 0D metal SBUs that are potentially stable, or meta-stable, in solution. Depending on the stability constants, these may be obtained under high ligand-to-metal ratios. No corresponding solution species are possible for the infinite 1D and 2D metal SBUs of rod-and sheet-MOFs, suggesting that lower ligand-to-metal ratios may be a viable strategy. Having discussed our results in a broader structural context, we will now look into some specific topology issues. ## Network topologies: A unified approach to MOFs There are three basic approaches for the discussion of the network topologies. The ''all-nodes'' approach with identification of SBUs, the ''standard simplification,'' which we will simply call ''standard,'' which takes the organic linkers as one node and each metal atom as a separate node thus splitting multinuclear metal SBUs into several nodes, or the ''cluster simplification,'' decomposing the structure into pieces with high connectivity, which may again generate metal SBUs. 14,74,75 The former lends itself more easily to synthesis planning for chemists, whereas the latter two approaches have been coded for, and databases such as the Cambridge Structural Database (CSD) have been searched and analyzed. 14 The topologies reported will be related in a net relation graph 76 and for single-metal SBUs often identical. For rod-MOFs, the differences are significant (Figure 7). The all-nodes approach has been modified to use the carboxylate carbons as branching points, which has the Article merit that it will also reflect the shape of the rod. This ''points-of-extension'' approach 2 is shown in the center left of Figure 7. Making connections only through the organic ligand does not consider the rod at all (see standard method, Figure 7, left). Recently, alternative approaches to rod-MOFs have been discussed, 16,17 and we will also compare them with a simplified version of the straight rod representation (STR) (Figure 7 center right). We also note the method recently proposed to use both points of extension and metal centers (PE&M, Figure 7 right). 16 This helps avoid some ambiguity in the points-of-extension approach, but it also introduces more nodes. ## Topology of H 2 NMe 2 [Y(btec)(H 2 O)]: Dot-MOF CTH-14 The coordination network CTH-14 is a straightforward dot-MOF with the metal SBU as the bridged dimer in the form of two edge-sharing square pyramids (Figure S1), and the structure can be described as eight-and four-connected scu-net, 77 with additional space filled with dimethylammonium ions (see Figure 8 right). Apart from the SBU approach using the dimer as the metal node giving the high symmetry scu-net, we could also consider the carboxylates binding to single-metal ions, the standard approach, where we now need to consider Y 3+ binding to five btec and the btec binding to either 4 or 6 Y 3+ (Figure S7), but no connection between the two Y 3+ in the dimer. This gives the 4,5,6-connected net, crg (or 4,5,6T11, Figure S8). The points-of-extension method applied to a dot-MOF Yet another way of looking at the CTH-14 structure is to consider the points-ofextension method as used for rod-MOFs. The merit of this would be to have the same description for dot-, rod-, and sheet-MOFs. We first note that this method has analogies with the augmented net. The augmented net is the net that is formed by replacing nodes by their corresponding polyhedra (Figure 9). 78 This is a common way of depicting, for example, MOF-5 and the pcu-net. Thus, the SBU {Zn 4 O(OCO) 6 } in MOF-5 forms an octahedron if we connect the six carbonyl Article carbons. In the process, we will also split the 6-connected node of the pcu-net into six 5connected nodes, giving the network topology cab or pcu-a (-a for augmented). For the scu-net, this means that the scu-a is formed by squares and cubes with 3-and 4-connected nodes, respectively. However, other polyhedral shapes can form 8connected nodes, for example, the two fused square pyramids (Figure S1) that form the metal SBU in CTH-14. The points-of-extension net formed in this way is a 6-nodal 4-, 5-, and 6-connected net (with ToposPro designation loh3). Although this may only seem to add unnecessary complications, pcu is a perfectly adequate description of MOF-5 and scu of CTH-14; what the points-of-extension method supplies is a topology including also the metal SBU geometry. This is exactly equivalent to the reported method of analyzing rod-MOFs 2 and also the way we propose to analyze sheet-MOFs. The points-of-extension method thus puts dot-MOFs, rod-MOFs, and sheet-MOFs on the same topological footing. Topology of [Ce 3 (btec)(Hbtec)(OAc)(HCO 2 )]: CTH-15 Although MOFs with layered infinite SBUs have been observed before, notably for [Fe 2 (btec)] MIL-62, 79 most of these tend to be dense layers, pillared by linkers. Some are based on the corresponding hydroxide structures, such as [Eu 2 (OH) 4 (bdc)], MIL-51, or the earlier-mentioned ULMOF-1 with a lithium-oxide-based layer. However, the overall shape of a rod is a line, a 1D object, but a 2D infinite SBU may have many overall motifs, like a square planar grid or a honeycomb net. In Figure 10, we can assign a 3-connected honeycomb (hcb) motif for the 2D SBU {Ce 3 (OCO) 7 } 2 N in CTH-15. These motifs can, similar to the rods in the rod-MOFs, be constructed in different ways. For example, edge-sharing octahedra may give a honeycomb net. Using the reported approach for rod-MOFs 2 that we call the points-of-extension method, we take the carboxylate carbons as the points defining the shape of the metal SBU (just as the form of the SBU of MOF-5 is described by an octahedron in Figure 9). This yields chains of edge-sharing trigonal prisms bridged by rectangles The augmented net is the net that is formed by replacing the nodes by the corresponding coordination figure. This is a common way of depicting, for example, the pcu-net and MOF-5. In the picture, we see the six-connected pcu-net in black and the augmented five-connected pcu-a or cab-net in red. We can generalize the augmented net concept and instead use the coordination figure of the actual SBU, which for MOF-5 is identical to the coordination figure of the net. In CTH-14, the approach leads to a points-of-extension net. and a three-nodal 4,5,6-connected topology loh2 (ToposPro designation) with point symbol {3 2 .4 4. 5.6 3 } 2 {3 2 .6 4 }{3 3 .4 5 .5.6 6 } 2 (Figure 11). A similar approach as the standard in Figure 7, linking only metal ions to the centers of the ligands instead gives the edge transitive two-nodal 4-and 6-connected stpnet based on square planes and trigonal prisms if we exclude the acetate and formate that can be seen as merely attachments to the network (Figure S10). The straight rod representation can also be extended to 2D with the hcb-nets (Figure 10) connected by the btec ligands. If we further simplify by considering the btec ligands as just a link between these sheets, we get the well-known five-connected bnn-net (Figure S10). Compared with the rod-MOFs, these sheet-MOFs seem rare and should by no means be confused with the case of having a flat 0D SBUs forming layers in a structure of a dot-MOF. Two other examples of sheet-MOFs are [Cd 3 (suc) 2.5 (dpa) 2 ]ClO 4 (suc = succinate, dpa = 4,4 0 -dipyridylamine) with a 8-metal 2D square grid SBU, 80 and [Pb 5 (1,3-bdc) 5 (H 2 O) 2 ] 2 with a 16-metal 2D honeycomb grid SBU with a roughly 5 A ˚aperture. 81 Because of the extra oxygen enabling the formation of these short 83 Topology of 4,4 0 -azopyridinium[Gd 2 (btec) 2 ]: CTH-16 CTH-16, 4,4 0 -azopyridinium[Gd 2 (btec) 2 ] is a typical example of a rod-MOF. The rods are formed by face-sharing cubes, bridged by 4-connecting btec 4 ligands, and the points-of-extension approach gives the bi-nodal 4,5-connected topology cjm with point symbol {3 2 .6 2 .7 2 }{3.4 5 .5 2 .6 2 } 4 (Figure 12). The rod-MOF structure CTH-16 can also be treated with the standard method, yielding the likewise bi-nodal htp-net based on hexagons (btec) and trigonal prisms (Gd), the assignment given to the isoreticular 4,4 0 -bipyridinium[Ln 2 (btec) 2 ] compounds in a recent database survey (Figure S4). 14 If we make straight rods and let the btec just be a linear connector, the structure reduces to the dia-net in an embedding with seesaw geometry of the nodes (Figure S11). ## The merits of the different approaches In Table 1, we compare some different topology approaches for the compounds presented in this work. Comparison of the different topology approaches for the materials presented in this article and MOF-5 for comparison. For each topology, we also present in parenthesis p = number of nodes and q = number of edges, as a crude indication of the complexity of each net. ## Article All topology methods have their merits, but it is essential to know the difference, especially for infinite SBUs, and be clear on which one is used. The standard method gives a conceptually simpler way of describing the net (Table 1) but normally has no links in the direction of the rod or the sheet, and thus these become ''invisible.'' The points-of-extension approach, on the other hand, can easily give very complex rod-SBUs 85 and, we suspect, even more complex sheet-SBUs. One advantage is the preservation, by definition, of the connectivity of the organic SBU. For example, as long as all carboxylate groups of the btec ligand are engaged in metal bonding, btec will always be described as a rectangular 4-connected SBU. Another advantage is the explicit description of the metal SBU that is no longer reduced to a single point. In addition, the metal SBUs are assembled during the reaction so that planning for a particular network to form, knowledge of what nets are possible with that particular metal SBU, or choosing what metal SBU to aim for will be essential for synthesis planning. However, one has to be sure that the increased complexity really gives an increased understanding for the points-of-extension approach to be worthwhile. So, although visually picturing MOF-5 as the points-of-extension uninodal net cab is closer to the real structure, the topology adds little. The same is true if describing MOF-5 with the three-nodal fff topology with the bridging dicarboxylates described as 4-connected nodes in the standard method. On the other hand, if easy-to-recognize rod or sheet patterns as ladders or fused polyhedra that could inspire the chemist are not what we are after, but instead a numerical comparison of large number of structures, then the PE&M may be a way forward. This gives for MOF-5 the three-nodal mof-net. A more detailed description also makes it easier to find closely related materials in various databases. ## Conclusions We have shown that in addition to the common rod-nets there are analogous infinite 2D SBUs forming sheets. We propose that the three classes of MOFs emerging from the designation of the metal SBUs as 0D, 1D, and 2D be called dot-MOFs, rod-MOFs, and sheet-MOFs, respectively, in order for us to be able to discuss them easily. Bennett and co-workers noted recently that there is a ''huge gap in our understanding of the thermal stability of this vast class of materials.'' 4 Literature data we have assembled, nevertheless, supports the idea that the thermal, but also chemical, stability of MOFs follow in the trend dot < rod < sheet. Certainly, for our materials, the sheet-MOF CTH-15 was the most stable thermally. Having said that, we want to stress that this is not the only factor determining the stability of a MOF. The network topologies of all three classes can be analyzed using the points-ofextension method. This does not invalidate or supersede other methods but provides a way of describing all MOFs in the same theoretical framework. Notably, it takes into account the geometry of the metal SBU and preserves the connectivity of the organic SBU. We suggest that implementing a search algorithm for the points-of-extension method and applying it to the MOF subset in the CSD may yield further insights into MOF structures and properties along the lines of recent studies, with the Article standard method, cluster method, or PE&M approaches perhaps more suitable for numerical comparisons. 19 We have attributed the flexibility, or potential breathing behavior, of CTH-16 and its isoreticular analogs partly to the large conformational space of the vicinal carboxylates of the btec linker, shown by analysis of data in the CSD. The photoluminescent behavior suggests potential for further investigations, but most importantly, this was used to identify the guest molecule in CTH-16. Finally, cyclic voltammetry shows a remarkable stabilization of Ce(IV) compared with solution chemistry, and we suggest Ce(III/IV) MOFs to be interesting catalyst candidates for various redox reactions. More experiments are planned to explore the potential for electrochemical redox catalysis with CTH-15. ## EXPERIMENTAL PROCEDURES Resource availability Lead contact Further information should be directed to and will be fulfilled by the lead contact, Lars O ¨hrstro ¨m (ohrstrom@chalmers.se). ## Materials availability All chemicals were purchased from Sigma-Aldrich and were used without further purification. Samples of CTH14-16 may be available on request to the lead contact. Single-crystal X-ray diffraction Single-crystal X-ray diffraction data were collected on a Rigaku XtaLAB Synergy-S, Dualflex diffractometer equipped with an AtlasS2 detector at 173 C using Cu Ka radiation (l = 1.54184 A ˚) and a Rigaku XtaLAB Synergy-S, Dualflex diffractometer equipped with a HyPix-6000HE detector using Mo Ka radiation (l = 0.71073 A ˚). Diffraction data were acquired and processed with CrysAlisPro software package. 86,87 Direct or structure expansion methods were used for all structures, and the refinements were established by full-matrix least squares with SHELX-2018/3 88 using X-seed 89 and Olex2 90 software as a graphical interface. Details of structure refinements are found in the supplemental information. ## Data and code availability Volumetric gas adsorption and surface area analysis N 2 adsorption isotherms were recorded on a Micromeritics ASAP2020 surface area analyzer at liquid N 2 temperature (196 C). The samples were pre-treated at 120 C under dynamic vacuum (1 3 10 4 Pa) for 6 h before the analysis. The relative pressure range of 0.05-0.15 was used to estimate the Langmuir and BET surface area of the samples. Additionally, CO 2 and N 2 adsorption isotherms were recorded at 20 C (with a temperature-controlled water bath) using the same instrument. ## Cyclic voltammetry All electrochemical measurements were made in a three-electrode cell with an Ag/ AgCl reference electrode and a Pt mesh counter electrode. The working electrode consisted of carbon paste (Metrohm), with and without electroactive material. The composition of the electroactive part was 50/50 wt % of carbon paste and electroactive material. A phosphate buffer with pH = 7 and concentration 0.1 M was used as electrolyte. The electrolyte was purged with nitrogen before each experiment, and a stream of nitrogen was kept over the solution during the experiments. The scan rate was 10 mV/s. ## Emission measurements Room temperature steady-state emission spectra of the compounds were obtained with a Spex Fluorolog 3 from JY Horiba; the excitation wavelength was 300 nm for all samples. For H 4 btec, 4,4 0 -azopyridine, CTH-15, and CTH-16, the samples were loaded into 1-mm cuvettes, but for CTH-14, the sample was deposited on a glass substrate, as there was not enough to fill a cuvette. ## SUPPLEMENTAL INFORMATION Supplemental information can be found online at https://doi.org/10.1016/j.chempr. 2021.07.006. ## ll OPEN ACCESS

chemsum

{"title": "A unified topology approach to dot-, rod-, and sheet-MOFs", "journal": "Chem Cell"}

carbazole-functionalized_hyper-cross-linked_polymers_for_co₂_uptake_based_on_friedel–craf

## Abstract: To systematically explore the effects of the synthesis conditions on the porosity of hyper-cross-linked polymers (HCPs), a series of 9-phenylcarbazole (9-PCz) HCPs (P1-P11) has been made by changing the molar ratio of cross-linker to monomer, the reaction temperature T 1 , the used amount of catalyst and the concentration of reactants. Fourier transform infrared spectroscopy was utilized to characterize the structure of the obtained polymers. The TG analysis of the HCPs showed good thermal stability. More importantly, a comparative study on the porosity revealed that: the molar ratio of cross-linker to monomer was the main influence factor of the BET specific surface area. Increasing the reaction temperature T 1 or changing the used amount of catalyst could improve the total pore volume greatly but sacrificed a part of the BET specific surface area. Fortunately changing the concentration of reactants could remedy this situation. Slightly changing the concentration of reactants could simultaneously obtain a high surface area and a high total pore volume. The BET specific surface areas of P3 was up to 769 m 2 g −1 with narrow pore size distribution and the CO 2 adsorption capacity of P11 was up to 52.4 cm 3 g −1 (273 K/1.00 bar). ## Introduction HCPs get more attention in recent years due to their high BET specific surface area , made under mild reaction conditions , used nonprecious materials as catalyst and wide applications , etc. The synthesis methods of HCPs include solvent knitting methods , Scholl coupling reaction , the knitting method with formaldehyde dimethyl acetal (FDA) , functional group reactions and so on. Among these methods, the knitting method with FDA as external cross-linker is the most time-efficient approach . FDA was first used as cross-linker to knit aromatic building blocks . Researchers used this method to knit triptycenes , triphenylphosphine , benzimidazole, 1,3,5-triphenylbenzene , carbazole , naphthol-based monomers etc. with FDA to obtain various HCPs, which exhibited outstanding porous properties. Studies of the effects on porosity of HCPs are significant, but almost all investigations focused on the role of monomer length and geometry on the porosity [6, . Researchers have synthesized a series of carbazole-based microporous HCPs and came to the conclusion that 2D and 3D-conjugated architectures with nonplanar rigid conformation and dendritic building blocks were favorable for getting a high BET specific surface area [6, . Qiao synthesized five microporous materials using carbazole with different flexible chains, proving that the flexible chain length was an important factor for the porosity . Different phenyl-based structures were also synthesized to explore the effect of the monomer structure on the porosity . However, researchers seldom cared about the effect of reaction conditions on the porosity, which is of farreaching significance in preparation of HCPs. In this work, a series of HCPs was synthesized from 9-PCz with FDA as the external cross-linker, the porosity was tuned by variation of the reaction conditions such as the molar ratio of cross-linker to monomer, the reaction temperature T 1 , the amount of used catalyst and the concentration of reactants. Additionally, the CO 2 uptake of the obtained polymers was explored. ## Results and Discussion The synthesis of HCPs is shown in Scheme 1 and Table 1. Using the Friedel-Crafts reaction, 11 samples (P1-P11) have been synthesized. To study the effect of the synthesis conditions on the porosity the molar ratio of building unit to crosslinker (P1-P5), the reaction temperature T 1 (P3, P6, P7), the amount of the catalyst used (P3, P8, P9) and the concentration of reactants (P3, P10, P11) were varied. ## Chemical structure analysis FTIR spectra were measured to verify the structure of HCPs (Figure 1). The peak at 3100-3000 cm −1 correspond to the C-H stretching vibrations of the aromatic rings, which declined obviously in P1-P5 compared to monomer 9-PCz. The peak of the disubstituted phenyl ring in the 9-PCz monomer at near 725 cm −1 disappeared while the peak of the trisubstituted phenyl ring near 800 cm −1 was dominant in polymers . C-H stretching vibration at about 2920 cm −1 belongs to the structure of -CH 2 -in the HCPs . The FTIR spectra of P6-P11 (Supporting Information File 1, Figure S1) were similar to the ones for P1-P5. P1-P11 are polymers with very similar chemical structure, which have been proved by FTIR. In addition, we performed solid state 1 H NMR and solid state 13 C NMR on P3 as a representative sample. The solid state 1 H NMR showed peaks in the range of 1.5-3.5 ppm for the saturated protons (Supporting Information File 1, Figure S2). Also, the solid state 13 C NMR (Figure 2) showed peaks between 25-50 ppm, indicating sp 3 carbons . The peaks about 139 ppm belong to the substituted aromatic carbon, the peaks about 128 ppm were attributed to the unsubstituted aromatic carbon. Based on the above peaks in the solid state NMR, the Friedel-Crafts polymerization product was confirmed. ## TGA analysis The thermal stability of HCPs was investigated by TGA tests (Figure 3 and Supporting Information File 1, Figure S3). A slight weight loss at 100 °C was observed for P2, P4, P5, and P7, due to the solvent wrapped in the hyper-cross-linked networks, which could not be removed even in vacuum . Except this, the TGA curves of P1-P11 exhibited similar de-composition behavior. The highest decomposition temperature of P1-P11 was up to 594 °C, with ca. 70% mass residues even when the temperature raised up to 800 °C (Supporting Information File 1, Table S1), demonstrated the splendid thermal stability of P1-P11 as reported for microporous polymers . ## Morphology analysis The morphology of P1-P11 was investigated by SEM images (Figure 4), which showed that HCPs were composed of rough surface particles. The particles had different size and agglomerated to loose aggregates. There were plentiful pores randomly distributed among the particles. X-ray diffraction (XRD) of obtained HCPs exhibited similar diffraction patterns, only a round peak at 10°, hinting that P1-P11 were amorphous polymers (Figure 5). a Surface area and pore volume were obtained by the Brunauer-Emmett-Teller (BET) method in the pressure range of 0.05-0.35 P/P 0 , the standard deviation of the porosity is 0.1%; b microporous surface area calculated from the adsorption branch of the nitrogen adsorption-desorption isotherm using the t-plot method; c surface area calculated from the nitrogen adsorption branch based on the Langmuir model; d microporous volume calculated from the adsorption branches using NLDEF methods. ## Porous properties The permanent porous nature was subsequently studied by subjecting the polymers to nitrogen adsorption-desorption experiments at 77 K. The porosity data of P1-P11 are listed in Table 2. The main influence of the porous properties is the reaction degree, as well as the C/N ratio, which has been confirmed by elemental analysis. Elemental analyses were measured to compare the degree of crosslinking of P1-P11 (Table 3), because the C/N ratio of the polymer will increase as the degree of crosslinking increases. The effect of molar ratio of cross-linker to monomer on the porosity of HCPs (P1-P5) To explore the effect of molar ratio of cross-linker to monomer on the porosity of HCPs, five polymers (P1-P5) were synthesized. The reaction conditions of P1-P5 were similar except the gradually increasing molar ratio of FDA to 9-PCz from 1 to 5. Except P1, P2-P5 exhibited a rapid nitrogen adsorption ability at low pressures (P/P 0 < 0.05, Figure 6a), which indicated that the micropores exist in the networks . The sorption isotherm of P2-P4 exhibited a combination of type I and IV nitrogen sorption isotherms according to the IUPAC classification . The hysteresis between adsorption and desorption of P2-P4 indicates that the polymers contain mesopores . There is no sharp rise at high relative pressures (P/P 0 > 0.9) of P1-P5, which means that scarcely macropores exist in the networks . The pore size distribution (Figure 6b) was calculated from the adsorption branches using nonlocal density functional theory (NLDFT) methods. P2-P5 exhibited narrow pore size distribution in the micropore scopes (<2 nm), while P1 showed a wide pore size distribution. The surface areas of P1-P5 ranged from 350 to 769 m 2 g −1 (Table 2). The lowest specific surface area of P1 was due to the less FDA which reduced the crosslinking density. Largely exaltation of BET surface area from P1 to P3 was due to the increasing FDA/9-PCz ratio improved the crosslinking density which could be confirmed by the increasing C/N ratio (Table 3). However, further increasing the molar ratio of FDA to 9-PCz could not result in a higher BET specific surface area, because the high steric hindrance prevented further crosslinking reaction , the raised C/N ratio (Table 3) maybe because of the tail end groups of FDA. The BET specific surface area of P3 was much higher than the polymer CZB (with similar carbazole monomer) . All that reveal that enough and suitable cross-linker amount was the premise of superior specific surface areas. The effect of reaction temperature T 1 on the porosity of HCPs (P3, P6, P7) The reaction temperature can mainly influence the polymerization process; to explore the effect of the reaction temperature on the porosity, we have synthesized P6 and P7 by increasing the reaction temperature T 1 from rt (P3) to 40 and 50 °C. When comparing the sorption isotherms of the HCPs (Supporting Information File 1, Figure S4), it was envisioned that those of P6 and P7 were a combination of type I and II nitrogen sorption isotherms, which emerged two steep N 2 adsorption abilities at the low pressure region (P/P 0 < 0.1) and the high pres-sure region (P/P 0 > 0.9), indicating that micropores and macropores appeared in the polymers . The BET specific surface area of P3, P6 and P7 (769 m 2 g −1 , 659 m 2 g −1 and 599 m 2 g −1 , respectively) decreased with increasing the reaction temperature T 1 , microporous surface area and the microporous volume presented the same trend. According to the C/N ratio, the crosslinked degree of P3 is also better than P6, P7. But the total pore volume (0.63 cm 3 g −1 , 1.24 cm 3 g −1 and 1.12 cm 3 g −1 , respectively) increased heavily with increasing T 1 . The total pore volume of P6 and P7 is much bigger than the carbazole-based HCPs such as CPOP-13 (890 m 2 g −1 , 0.468 cm 3 g −1 ), CPOP-14 (820 m 2 g −1 , 0.416 cm 3 g −1 ), Cz-POF-4 (914 m 2 g −1 , 0.6 cm 3 g −1 ) . This may be the result of that excessive temperature caused excessive crosslink at the beginning of the reaction, the plethora network cocooned a part of the reaction center, and prevented it from further cross-linking (micropores), hence it formed macropores. All that indicated that improving the reaction temperature T 1 could enhance the total pore volume of HCPs but lowers the specific surface area. ## The effect of the amount of catalyst on the porosity of HCPs (P3, P8, P9) To explore the effect of the amount of catalyst on the HCPs porosity, P8 (3 mmol FeCl 3 ), P3 (4 mmol FeCl 3 ), P9 (5 mmol FeCl 3 ) were made by varying the catalyst amount. The nitrogen sorption isotherms (Supporting Information File 1, Figure S5) of P8 and P9 were similar as the one of P7; this means that micropores and macropores exist simultaneously in the polymers. The specific surface area for P8 (612 m 2 g −1 ) and P9 (671 m 2 g −1 ) was inferior to P3 (769 m 2 g −1 ), this trend was similar to the C/N ratio. The total pore volume of P9 (1 cm 3 g −1 ) was much higher than P3 (0.63 cm 3 g −1 ), P8 (0.64 cm 3 g −1 ). The influence of the amount of catalyst used on the porosity can result in a high specific surface area, when applying a suitable amount of catalyst. The effect of the concentration of reactants on the porosity of HCPs (P3, P10, P11) The effect of the concentration of reactants on the porosity was studied by changing the FDA concentration in the synthesis of HCPs P3, P10, and P11. As shown in Supporting Information File 1, Figure S6 and in Table 2, the sorption isotherms of P10 and P11 are similar to the one of P7, which signifies the presence of permanent micropores and macropores in the polymers . While the P3 porosity was composed of micropores and mesopores as above-mentioned, the pore size distribution of the three HCPs were similar and showed a narrow distribution in the micropores region and a pore size center at ca. 0.7 nm. The BET specific surface areas of P3, P10, and P11 were about the same (769 m 2 g −1 , 755 m 2 g −1 , 760 m 2 g −1 , respectively), the microporous surface area and the microporous volume were also similar. However, there were wide disparities in the total pore volume among the obtained polymers. The volume of P10 (1.11 cm 3 g −1 ), P11 (1.27 cm 3 g −1 ) was about twice that of P3 (0.63 cm 3 g −1 ) which is owing to the extra generated macropores, and it is higher than many carbazole-based HCPs with similar BET specific surface area . We conjecture that the concentration of FDA can affect the formation process and morphology of the polymers, when polymer particles agglomerate and stack together (Figure 4). We reached the conclusion that when varying the concentration of reactants slightly, a great increase of the pore volume can be accomplished without sacrificing the BET special surface area. ## CO 2 uptake behavior The presence of many CO 2 -philic sites (N-bearing substituents) and narrow pore distribution in the networks could improve the molecular interaction with CO 2 . Hence, three polymers (P3, P10, and P11) were selected as representative samples to conduct CO 2 adsorption experiments up to 1 bar at both 273 and 298 K (Figure 7). HCPs showed a similar and moderate CO 2 uptake (Table 4). P11 displayed the optimal CO 2 storage of 52.4 cm 3 g −1 (10.4 wt %) at 1.0 bar/273 K. which was higher than that of the carbazole-based microporous polymers PBT-C1 (46 cm 3 g −1 ) , CMPSO-1B3 (46.8 cm 3 g −1 ) , CPOP2-4 (7.8-9.0 wt %) , tetraphenylmethane-based CPOP10 (S BET = 3337 m 2 g −1 , 9.1 wt %, at 298 K/1.00 bar) or the melamine-based microporous PAN-NH-NH 2 (9.7 wt %) . There was no saturation observed when the pressure reached to 1 bar, indicating that a higher CO 2 capacity could be obtained by further increasing the pressure. The isosteric heat (Q st ) of each polymer was calculated based on the adsorption data at different temperature using the Clausius-Clapeyron equation (Supporting Information File 1, Figure S7). At the zero CO 2 gas surface coverage, the limiting enthalpies of adsorption of the three samples was similar (P3:30 kJ/mol, P10:28 kJ/mol, P11:29 kJ/mol) and within the scope of physical adsorption , which was beneficial to the materials reuse. ## Conclusion Using the Friedel-Crafts reaction, 9-PCz microporous polymers (P1-P11) were prepared by varying the molar ratio of cross-linker to monomer (P1-P5), the reaction temperature T 1 (P3, P6, and P7), the amount of catalyst used (P3, P8, and P9) and the concentration of reactants (P3, P10, and P11). The systematic study showed that the molar ratio of cross-linker to monomer was the main way to influence the BET specific surface area. A sufficient amount of cross-linker was the premise of a superior BET specific surface area. Increasing the reaction temperature T 1 or the amount of catalyst used could increase the pore volume greatly but sacrificed in part the BET specific surface area. Changing concentration of reactants could remedy this situation. When slightly varying the concentration of reactants simultaneously, a high surface area and high total pore volume could be obtained. Those provided a reference for preparing HCPs using Friedel-Crafts polymerization. The BET specific surface area of the prepared HCPs was up to 769 m 2 g −1 , and the CO 2 uptake capacity was up to 10.4 wt % at 273 K/1 bar. ## Characterization methods Fourier-transform infrared (FTIR) spectra of HCPs were obtained by using a Nicolet 6700 spectrometer over a wave number range of 4000-400 cm −1 by scanning 32 times at a resolution of 4 cm −1 . TG analysis of the polymers were conducted with a NETZSCH TG 209F1 TG analyzer for 40-800 °C at a heating rate of 10 °C min −1 under a nitrogen flow of 50 mL min −1 . The X-ray diffraction (XRD) patterns of the as prepared polymers were collected using a PANalytical X'pert Pro MPD diffractometer with Cu Kα radiation at room temperature, with step size of 0.0202°, 2θ ranging from 5.0 to 60°. Scanning electron microscope (SEM) measurements of obtained samples were carried out using a Hitachi SU1510 microscope. The nitrogen adsorption and desorption and the CO 2 adsorption and desorption isotherms of HCPs were obtained using a GAPP V-Sorb 2800P BET surface area and pore volume analyzer. Polymers were degassed at 100 °C for over 10 h under vacuum before all gas analysis experiments. ## Synthesis HCPs The synthetic illustration of HCPs is depicted in Scheme 1. Using the Friedel-Crafts reaction, P1-P11 have been made by changing the molar ratio of building unit to cross-linker (P1-P5), the reaction temperature T 1 (P3, P6, and P7), the amount of catalyst (P3, P8, and P9) and the amount of solvent used (P3, P10, and P11). The synthesis of P3 as representative procedure is given in detail: Under a nitrogen atmosphere, 9-PCz (0.67 mmol, 0.163 g), FDA (2 mmol, 0.152 g) were dispersed in DCE (18 mL) and then anhydrous FeCl 3 (4 mmol, 0.64 g) was added to the dispersion; the mixture was allowed to react at room temperature for 5 h, then at 80 °C for 19 h with vigorous stirring. Then the mixture was cooled to room temperature and quenched by using 20 mL of CH 3 OH. Then the solid product was separated by filtration, and the solid product was washed with first methanol, followed by THF, HCl/H 2 O 2:1 (v/v) and distilled water successively, further purified by Soxhlet extraction with MeOH for 24 h and then THF for another 24 h. Finally, the product was dried in a vacuum oven at 100 °C for 24 h. The obtained polymer material was obtained as a brown solid. The synthesis of other polymers was similar as P3, only the monomer amount or other experimental conditions were varied as shown in Table 1. Although washed excessively, the yield of the polymers still exceeded 100% which was due to the adsorbed catalyst or solvent in the pore structure . All obtained samples were colored ranging from pale brown to dark brown.

chemsum

{"title": "Carbazole-functionalized hyper-cross-linked polymers for CO₂ uptake based on Friedel\u2013Crafts polymerization on 9-phenylcarbazole", "journal": "Beilstein"}

assessing_accuracy_of_an_analytical_method_in_silico:_application_to_"accurate_constant_via_transien

## Abstract: Analytical methods may not have reference standards required for testing their accuracy. We postulate that accuracy of an analytical method can be assessed in the absence of reference standards in silico if the method is built upon deterministic processes. A deterministic process can be precisely computersimulated thus allowing virtual experiments with virtual reference standards.Here, we apply this in silico approach to study "Accurate Constant via Transient Incomplete Separation" (ACTIS), a method for finding the equilibrium dissociation constant (K d ) of protein-small molecule complexes. ACTIS is based on a deterministic process: molecular diffusion of the interacting protein-small molecule pair in a laminar pipe flow. We used COMSOL software to construct a virtual ACTIS setup with a fluidic system mimicking that of a physical ACTIS instrument. Virtual ACTIS experiments performed with virtual samples -mixtures of a protein and a small molecule with defined rate constants and, thus, K d of their interaction -allowed us to assess ACTIS accuracy by comparing the determined K d value to the input K d value. Further, the influence of multiple system parameters on ACTIS accuracy was investigated. Within multi-fold ranges of parameter values, the values of K d did not deviate from the input K d values by more than a factor of 1.25 strongly suggesting that ACTIS is intrinsically accurate and that its accuracy is robust. Accordingly, further development of ACTIS can focus on achieving high reproducibility and precision. We foresee that in silico accuracy assessment, demonstrated here with ACTIS, will be applicable to other analytical methods built upon deterministic processes. Accuracy is one of the key performance parameters of quantitative analytical methods. Many methods do not have reference standards and, thus, their accuracy cannot be experimentally tested. We suggest that accuracy of an analytical method can be assessed in the absence of reference standards in silico if the underlying physicochemical processes are deterministic, i.e. can be precisely simulated. Detailed computer simulation should facilitate virtual experiments in which a virtual sample can be perfectly defined and, thus, can serve as a virtual reference standard. In this work, we applied this in silico accuracy assessment to "Accurate Constant via Transient Incomplete Separation (ACTIS)". ACTIS is a separation-based method for finding K d of complexes (PL) between proteins (P) and small-molecule ligands (L): where K d is defined via equilibrium concentrations of P, L, and PL: [P] eq / [PL] eq ( 2) There is no reference standard for the determination of K d , i.e. there is no PL with a known K d , thus, accuracy of ACTIS cannot be assessed experimentally. On the other hand, ACTIS is built upon a deterministic process of molecular diffusion of P, L, and PL, interacting as in eq 1 while moving within a laminar pipe flow. This process can be precisely described by a set of partial differential equations with precisely defined initial and boundary conditions and precisely defined input values of the rate constants in eq 1: k on,inp and k off,inp . The input values of the rate constants define the input value of the equilibrium constants: which can serve as a virtual reference standard in virtual ACTIS experiments. Accordingly, ACTIS is suitable for in silico accuracy assessment. In ACTIS, a short plug of an equilibrium mixture of P and L in a buffer solution is injected into a capillary pre-filled with the pure buffer solution. The plug is then propagated inside the capillary by a pressure-driven flow of the buffer solution. Different rates of transverse diffusion of PL and L in laminar flow cause their transient incomplete separation (TIS) (Figure 1A) resulting in a non-diffusive peak for PL and a diffusive peak for L (Figure 1B). To determine K d , TIS is performed for a series of equilibrium mixtures with a constant concentration of L and varying concentration of P producing a set of curves termed separagrams (Figure 1B). The cumulative signal of protein-bound L and unbound L is taken at the time corresponding to the maximum of the diffusive peak for each curve. Subsequently, a and a cumulative signal from L and PL is measured at time  L , which is the characteristic time of transverse diffusion of L. The signal is measured at a constant concentration of L and varying concentrations of P. C: A binding isotherm "signal-at- L vs concentration of P" is built, and K d is found as the concentration of P, which corresponds to the signal in the middle between the maximum and minimum signals. classical binding isotherm "signals vs concentration of P" is built to reveal the value of K d (Figure 1C). ACTIS is a uniquely deterministic method as it relies on molecular diffusion in a pressure-driven flow which can be described by a system of partial differential equations with fully-defined initial and boundary conditions. 1-3 As such, it is perfectly suited for computational assessment of its accuracy. A physical ACTIS instrument has a minimum fluidic system with a pump, an injection loop, a separation capillary, and a multiport valve (Figure 2A). The valve serves as two connectors: one from a pump tube to the injection loop and the other one from the injection loop to the separation capillary. Accordingly, the instrument has five essential fluidic components which can be presented as a series of coaxial pipes of different radii (Figure 2B). We used COMSOL to construct a five-component ACTIS setup depicted in Figure 2B with dimensions identical to those of a physical ACTIS instrument (described in the figure legend). We found that K d determined in this setup deviated from K d,inp , defined by eq 3 and used as a virtual reference standard, by a factor of 1.02. This setup was utilized to study how the accuracy of K d was affected by variations in: (i) the radius of the injection loop while keeping its volume constant, (ii) the radius of the separation capillary, (iii) shape of the initial (prior to start of TIS) plug of the equilibrium mixture, and (iv) ramp time in flow-rate onset (after the start of TIS). The variations of parameter values used in this study exceeded markedly the ones that are expected in a physical ACTIS setup. We found that despite multi-fold variations in values of a number of parameters, the maximum deviation of K d from K d,inp was less than a factor of 1.25, suggesting robust intrinsic accuracy of ACTIS. The intrinsic accuracy of ACTIS has a great practical importance as it allows ACTIS developers to vary instrument configuration without raising concerns about the influence of such variations on accuracy of K d . Efforts and resources can thus be focused on optimizing instrument configuration to achieve the best reproducibility and the highest precision, the second most important performance parameter of this quantitative analytical method. ## ■ THEORETICAL BACKGROUND Here we describe theoretical aspects of ACTIS that are essential for understanding this method at the conceptual level and help establish the theoretical background required for understanding our current work. In particular, we explain basic principles of K d determination and TIS of L from PL; the corresponding two sections are a close reiteration of our previously published explanation. Then, a standard way of finding K d is to determine R exp for a wide range of [P] 0 at a constant [L] 0 and to plot a binding isotherm: R exp vs [P] 0 . Finally, this binding isotherm is fitted with eq 7 using K d as a fitting parameter and the best fit reveals the sought value of K d . 1 Fundamentally, finding R exp with eq 6 requires that S L and S PL be measurable and that S L ≠ S PL . The latter inequality of signals from pure L and pure PL requires that L and PL be "separated" either spectrally or physically. 5 Importantly, complete separation is not required if a signal from the mixture of L and PL is a superposition of signals from individual components L and PL comprising the mixture (see eq 5), and if pure P does not contribute to the cumulative signal from L and PL. As a result, spectral methods that provide only incomplete separation of signals from L and PL (optical spectra of L and LP do overlap typically) are common in finding R exp . 5−7 ACTIS facilitates finding R exp via incomplete physical separation of L and PL. Transient Incomplete Separation of L from PL. TIS of L from PL will occur always when a short plug of their mixture is propagated within a Hagen-Poiseuille laminar flow in a long capillary. Such a flow is established by a pressure difference between the capillary ends and has a characteristic parabolic profile of flow velocity: the velocity ranges from zero at the capillary walls to its maximum in the capillary center. 8 TIS of L from PL in the longitudinal direction is possible due to the difference in rates of transverse diffusion between small-size L and large-size PL. PL that is near the capillary center will diffuse to the capillary wall slower than L and, thus, will be displaced longitudinally by the flow more than L. PL located near the capillary wall will diffuse to the capillary center slower than L and will be displaced longitudinally by the flow less than L. As a result, during a short transitional stage, a bulk of PL moves faster than a bulk of L, while a tail of PL moves slower than that of L. The separation is incomplete, i.e. the longitudinal concentration profiles (concentration vs position in the capillary) of L and PL do overlap, even during the transitional stage. Further, this separation gradually dissipates, i.e. the longitudinal concentration profiles of L and PL become symmetrical around the same symmetry axis, after the transitional stage. The after-TIS stage is described by the well-known Taylor dispersion. 9 Tracking TIS is viewed to be optimal with a flow system in which the optimum distance from the starting position of the equilibrium-mixture plug to the detector (l opt ) is linked with the average flow velocity (v av ), characteristic time of transverse diffusion of L from the center of the capillary to its inner wall ( L ), the volumetric flow rate (Q), and diffusion coefficient of L (µ L ), as follows: 1,10 where v av relates to Q and the inner capillary radius (r) as: The characteristic time of transverse diffusion across the capillary is defined in general as: Longitudinal concentration profiles of L and PL are partially separated in the time domain as shown in Figure 1B. Further, if a signal can be measured for each of L and PL, with the abovementioned detector, inside the capillary or at its exit and is proportional to the average cross-sectional concentration of each of them, S L  [L] and S PL  [PL], then, the cumulative signal S satisfies eq. 5 and can, thus, be used to determine R exp with eq. 6. Finally, R exp can be measured for a wide range of [P] 0 to construct a classical binding isotherm R exp vs [P] 0 , which, in turn, can be used to find K d via fitting the isotherm with eq. 7. ## ■ DESIGNING A VIRTUAL ACTIS EXPERIMENT Five-Component Virtual ACTIS Setup. The five-component physical ACTIS system depicted in Figure 2A operates as follows. The injection loop is filled with the sample. The sample is slowly transferred into the separation capillary at a distance equal to the sample-plug length from the capillary entry. The sample is then propagated fast through the separation capillary to cause TIS of L from PL. We used COMSOL to construct a virtual five-component ACTIS setup with default dimensions identical to those of a physical ACTIS instrument (Figure 2B). The dimensions of the two connectors are identical and can be considered fixed as well as the dimensions of the pump tube. The dimensions of the injection loop and the separation capillary may need to be changed and, thus, were varied in this study. Of course, the virtual system also contains representations of a virtual pump (a source of a hydrodynamic flow characterized by its volumetric flow rate Q) and a virtual detector (a detection volume inside the capillary), which are also associated with some parameters that can vary and potentially cause inaccuracy in K d . A default initial configuration in the virtual five-component setup is the one in which the loop and the loop-capillary connector are filled with the sample, while the remaining components are filled with the buffer solution. This situation mirrors the physical ACTIS setup. In a default virtual ACTIS experiment, the sample is slowly transferred from the injection loop to the separation capillary (Q = Q inj ) and then propagated fast inside the capillary as in the physical ACTIS experiment (Q = Q TIS ). Further in this work, the five-component setup was used first to test the accuracy of K d for the default geometry (Figure 2B). Then, in a more rigorous study, it was used to investigate how the accuracy of K d determined with ACTIS was affected by variations in: (i) the radius of the injection loop while keeping its volume constant, (ii) the radius of the separation capillary, (iii) shape of the initial (prior to start of TIS) plug of the equilibrium mixture, and (iv) ramp time in flow-rate onset (after the start of TIS). Below, we provide details on how these four parts of the study were set up. ## Geometry of the Injection Loop and Separation Capillary. There are four parameters that characterize the geometry of the injection loop and the separation capillary: the radius (r 1 ) and length (l 1 ) of the injection loop and the radius (r 2 ) of the separation capillary and the distance from the beginning of the separation capillary to the centre of the virtual detector (l 2 ). We conducted two sets of virtual ACTIS experiments in which r 1 and r 2 were varied. In one set of virtual experiments, r 1 was varied from 5 µm to 5000 µm (l 1 was changed accordingly to keep the volume of the injection loop constant), while r 2 and l 2 were kept constant at default values linked through eqs 8-10. In the other set of virtual experiments, r 2 was varied from 5 µm to 5000 µm while holding constant r 1, l 1 , and l 2 . The rest of the parameters were default (see Figure 2B). Geometry of the Initial Sample Plug. While the ideal shape of the initial plug of the equilibrium mixture in the beginning of the separation capillary is cylindrical, in a real experiment, it will be distorted to some degree due to the imperfections of the injection process. To assess how such distortions can affect the accuracy of K d , we examined square, Gaussian, and half-Gaussian longitudinal distributions of concentrations of L, PL, and P in the initial plug (Figure 3A); cross-sectional concentration profiles were uniform. Here, as opposed to a default way of injecting a 3.0 cm-long sample plug from the injection loop, the same-length initial plug was defined near the entrance of the separation capillary at a distance equal to the plug length, and no injection process was simulated. To model this initial plug shape, the longitudinal (along axis x) distribution of concentrations was defined as a square wave function: ) where H(kx) = 1/(1 + e −2kx ), which is an analytical sigmoidal approximation of the Heaviside step function. 11 The shape of the plug was varied by varying both k 1 and k 2 . The condition of k 1 = k 2 corresponds to symmetric plug shapes, e.g. for k 1 = k 2 = 2 the plug shape is Gaussian and for k 1 = k 2 = 200 the plug shape is a close approximation of the square function. For k 1 ≠ k 2 the plug shape is asymmetric, e.g. half-Gaussians. The rest of the parameters were default (see Figure 2B). Ramp Time in Flow-Rate Onset. Achieving and maintaining the ideal plug shape requires the injection of a sample plug from the injection loop into the separation capillary at a slow flow rate of Q inj for a time period of t inj . TIS, on the other hand, requires the propagation of the injected sample plug though the separation capillary at a high flow rate of Q TIS . In a virtual ACTIS experiment, the transition from Q inj to Q TIS can be very fast (e.g. sub-seconds) or even instant. In a real experiment, however, this transition is a process that requires a second-scale time interval of t ramp , i.e. the ramp time of the flow rate onset (Figure 3B). In ACTIS, the runtime is relatively short (peaks of PL and L are detected within 30 s), i.e. the ramp time of a few seconds may significantly influence TIS of L from PL. Therefore, we varied t ramp to examine the impact of this variation on the profiles of PL and L and, in turn, on the accuracy of K d determined with ACTIS. The rest of the parameters were default (see Figure 2B). ## ■ SETTING UP A VIRTUAL ACTIS EXPERIMENT IN COMSOL To simulate TIS in the five-component setup depicted in Figure 2B we used COMSOL Multiphysics software, version 5.4, with the "Transport of Diluted Species" and "Laminar Flow" modules, which incorporate equations for both mass transfer and reversible binding of P and L in an equilibrium mixture. Computation time depends on the dimensions of the simulated geometries; thus, to reduce this time, the lengths of all of the five fluidic components can be scaled down along with the length of the detection window. 1 In such a case, the values of Q inj and Q TIS must also be scaled down to keep the l 2 / Q TIS ratio constant. The scaled down dimensions used in COMSOL computation are reported in SI. Moreover, the mesh size was chosen to minimize computational time and memory requirements. For all the simulations the "finer mesh" setting in COMSOL was used with a computational time of ≈ 3 h (2 × Intel® Xenon® CPU X5690@3.47 GHz, 96 GB RAM); finer settings such as "extra fine mesh" resulted in some cases in a better accuracy (up to 10 times better); however, they required an order of magnitude longer computational time (≈ 20 h) (SI). In all virtual ACTIS experiments, we considered a smallmolecule L with typical diffusion coefficient of µ L = 500 µm 2 /s and a large P with a typical diffusion coefficient of µ P = 50 µm 2 /s. The remaining parameters used as COMSOL settings were: 50 µm 2 /s, [L] 0 = 0.5 µM, T = 300 K. [P] 0 was varied from 1 nM to 1 mM using 11 different non-zero concentrations plus zero concentration. Three other values of k off,inp i.e. 10 4 , 10 5 , and 10 6 s 1 with respective values of K d,inp , 10 −7 , 10 −8 , and 10 −9 M, were also used in COMSOL. Testing several k off values allowed us to verify the independence of the simulations on a particular value of k off , and, therefore, K d ; the simulations were shown to be consistent with the default value of k off,inp = 10 3 s 1 (Figure S1). The sample plug is fully injected during t inj / 2 (t inj = 24 s). Subsequently, the sample plug is displaced from the entrance of the separation capillary inside the capillary at a distance equal to the plug length during another t inj / 2. This displacement is meant to place the initial plug away from the junction between the loop- is set to Q TIS and the plug is further propagated through the separation capillary. The time interval required to reach the propagation flow rate Q TIS is t ramp . capillary connector and the separation capillary; flow disturbance is expected to be the greatest near the junction. The virtual detector was meant to be at the end of the virtual separation capillary, and virtual detection was performed by averaging concentrations across the capillary within a cylindrical detection window (dimensions: 5 mm × r 2 ). Experimentally, the cumulative signal S from protein-bound L and unbound L is obtained by averaging points within a time window around the second-peak maximum, i.e. the diffusive peak (around time  L in Figure 1B). 1 Using a finite-length time window compensates for noise in experimental data; accordingly, we used a finite time window to find the cumulative signal S for all our virtual experiments. Unless otherwise stated, for all the separagrams, the position of the time window was chosen by default by selecting the time at which S L > S PL and S L − S PL = max. All the five fluidic components were modeled as 2D axisymmetric shapes to further reduce computation time in COMSOL. To study how the variation in the ramp time of flow onset affects the accuracy of K d , a piecewise function was defined in COMSOL for the linear transition from Q inj to Q TIS , and the length of this transition time, t ramp , was varied. To study how the variation in plug shape influences the accuracy of K d , a concentration distribution (eq 11) was defined as an analytical function in COMSOL and, then, varied as explained in the previous section. Note that neither COMSOL nor any other computational calculation software possesses infinite (numerical) accuracy. COMSOL's inaccuracy will contribute to the assessment of ACTIS's inaccuracy making it an assessment of the lower limit of ACTIS accuracy. ## ■ RESULTS AND DISCUSSION General Considerations. It is difficult to setup a benchmark for K d accuracy, but taking into account that variations in K d determined experimentally often reach orders of magnitude, we consider deviation by less than a factor of 1.25 as acceptable. Accordingly, for the purpose of this work, we consider K d accurate if 0.80 < K d / K d,inp < 1.25. We assume that the virtual ACTIS system is stable (this was confirmed by repeating virtual runs); therefore, we do not provide random errors. Note that we do not consider the detector as a source of imprecision in the following consideration because highly precise detectors for different detection methods, such as fluorescence and mass spectrometry, are available. In fact, any detector that can measure a cumulative cross-sectional average signal from L and PL satisfying eq 5 is suitable for accurate ACTIS measurements provided that it satisfies the following requirements. The detector must have a sufficiently high signal readout speed and a concentration limit of quantitation below K d values of the studied complexes; the limit of quantification (LOQ = 10 × S/N, where N is the level of noise) for L on the used detector should be at least equal to [L] 0 . 4 Accuracy of the Default ACTIS Setup. The virtual fivecomponent ACTIS setup, described in the "Setting up a Virtual ACTIS Experiment" section, was first examined for accuracy under default conditions shown in Figure 2B. A full set of separagrams was calculated and K d was determined from the respective binding isotherm (Figure 4). We found that K d determined in this setup deviated from K d,inp by a factor of 1.02. The proven accuracy of the default five-component ACTIS setup opened a route for performing a more stringent accuracy test with variations in (i) the radius of the injection loop while keeping its volume constant, (ii) the radius of the separation capillary, (iii) the shape of the initial (prior to start of TIS) plug of the equilibrium mixture, and (iv) ramp time in the flow rate transfer from Q inj to Q TIS . ## Variation of Injection-Loop Radius. In this examination, r 1 was varied from 5 to 5000 µm (the default value for r 1 was 50 m). This variation in r 1 corresponded to variation in r 1 / r 2 from 0.05 (= 5/100) to 50 (= 5000/100) for the default value of r 2 = 100 m. To keep the volume of the injection loop constant (which is the purpose of the injection loop), l 1 was changed accordingly for each r 1 : l 1 ~ 1 / r 1 2 . The rest of the parameters were default (see Figure 2B). For each value of r 1 , a full set of separagrams has been computed, and the K d value was determined from a respective binding isotherm (Figure S2). All determined K d values were compared to K d,inp (Figure 5A). We found that K d was accurate according to our criteria (0.8 < K d / K d,inp < 1.25) for r 1 / r 2 ≤ 2.5 but not for r 1 / r 2 ≥ 5. At r 1 / r 2 = 0.25 we obtained a binding isotherm with anomalous shape (Figures S2) and K d / K d,inp = 1.27. This anomaly was due to a numerical artifact from meshing in COMSOL; by using a more refined mesh we obtained K d / K d,inp < 1.02. The improvement in accuracy by mesh refinement may suggest that the large deviations in K d at r 1 / r 2 ≥ 5 are also due to too coarse meshes as well. Thus, more refined and optimized meshes (in particular, for boundary regions between small and large areas) could improve K d determination in a virtual ACTIS experiment. We confirmed this for the extreme value of r 1 / r 2 = 50 and found an optimal K d / K d,inp = 1.00 at the expense of excessively increasing the computational time (≈ 72 h instead of ≈ 3 h) and the potential risk of overfitting (SI). In order to keep studies consistent, comparable and in a reasonable time frame as well as to avoid overfitting, we chose the "finer mesh" setting from COMSOL without manual optimization for all following studies. It is instructive to analyze the shapes of the separagrams (Figure S2) to understand the influence of varying r 1 on the TIS process. First, for separagrams at r 1 / r 2 of up to 2.5 (= 250/100) they show the two expected peaks -the first for the nondiffusive species (PL) and the second for the diffusive species (L) as in Figure 1B. However, for greater ratios, i.e. r 1 / r 2 ≥ 5, the two peaks start merging into one (see shape plots in Figure 5A) followed by a long tail. At these greater ratios, the radius of the injection loop is much larger than that of the separation capillary, and the injection-loop length is very short (Table S1). This large difference between the radii causes the sample plug to take more time to move out of the corner regions of the injection loop that are far removed from the center of the axis (Figure S3) during sample injection. Since the motion of the sample plug out of the injection loop is driven not only by convective motion but also by diffusion, more PL is retained in the injection loop as the sample plug is injected into the separation capillary, since PL diffuses much slowly than L. This retention causes a significant decrease in the signal of PL (Figure S2). In summary, K d could be accurately determined for ratios r 1 / r 2 ≤ 2.5. The ratios r 1 / r 2 ≥ 5 resulted in large deviation in K d likely due to a combination of The dashed lines indicate the time window within which the average signal was taken to calculate the R exp values using eq 6. The binding isotherm was fit with eq 7 with K d being a fitting parameter. mesh size limitations and non-optimal TIS conditions; however, it is to be noted that these ratios correspond to r 1 > 500 µm and l 1 < 500 µm which are impractical parameters for a physical ACTIS instrument. Variation in Separation-Capillary Radius. In this examination, the value of r 2 was varied from 5 to 5000 µm. Thus, the ratio r 2 / r 1 (note the ratio is opposite to r 1 / r 2 from the previous study) varied from 0.1 (= 5/50) to 100 (= 5000/50) for a fixed value of r 1 = 50 m. The rest of parameters were default (Figure 2B). Note that while the ratios between r 1 and r 2 change in this and the previous examinations, these two sets of virtual experiments are not identical as decreasing r 2 while keeping r 1 , l 1 , and l 2 constant lead to situations when the plug length is comparable to the length of the separation capillary, which is a greatly suboptimum condition. For each ratio r 2 / r 1 , a full set of separagrams has been calculated, and the K d value was determined from a respective binding isotherm (Figure S4). All determined K d values were compared to K d,inp (Figure 5B). K d values were accurate according to our criteria (0.8 < K d / K d,inp < 1.25) for r 2 / r 1 ≤ 5 (= 250/50). For r 2 / r 1 > 5, the deviation of K d from K d,inp was beyond the acceptable range; specifically, in this case the values of K d were overestimated: Again, it is instructive to analyze the shapes of the separagrams (Figure S4). They show two classical peaks corresponding to the non-diffusive species (PL) and diffusive species (L) (as in Figure 1B) for r 2 / r 1 between 1 (= 50/50) and 20 (= 1000/50). For r 2 / r 1 < 1, however, the non-diffusive peak becomes less pronounced (see the left shape plot in Figure 5B), which can be attributed to the plug-length (l plug ≥ 52 cm) being greater than the distance from the capillary inlet to the detection window l 2 = 50 cm). Consequently, most of PL and L in the equilibrium mixture cannot be separated before they reach the detector. Obviously, reducing the plug length by adjusting the injection loop would solve this issue and return these cases to the optimum conditions. Still, K d deviated from the input K d,inp by less than a factor of 1.25 implying that K d could be accurately determined from these curves. For r 2 / r 1 > 5, both peaks merged into a single diffusive one (see the right shape plot in Figure 5B). Here, the time of separation (≥ 500 s) is comparable with or greater than the characteristic time of complex dissociation (≈ 1 / k off = 1000 s); therefore, the injected sample plug is no longer in the state of equilibrium. Moreover, the transfer flow between the small-radius injection loop and the large-radius separation capillary results in the phenomenon of "flow separation". 12 In this situation, the flow lines are no longer parallel across the whole cross section of the separation capillary. Indeed, recirculation of the flow occurs near the capillary walls (Figure S5) leading to remixing, which affects TIS of L from PL. Additionally, the Péclet number Pe for large r 2 (> 500 µm) becomes smaller than 40 and suggesting that longitudinal diffusion along the separation capillary becomes nonnegligible at r 2 / r 1 > 5 (Pe ≈ 42, Table S1 and S2). 13 All the above non-optimal conditions, i.e. the departure from equilibrium, the "flow separation" phenomenon, and significant longitudinal diffusion, likely contribute to the large deviation in K d from K d,inp at r 2 / r 1 > 5. In summary, K d could be accurately determined for r 1 / r 2 ≤ 5. The ratios r 1 / r 2 ≥ 10 resulted in large deviation in K d due to the different non-optimal conditions described previously and likely due to sub-optimal meshing. However, if one works within the range of optimal conditions, which correspond to r 1 / r 2 ≤ 5 in our case, it can be concluded that the accuracy of K d determination by ACTIS is invariant with respect to changes in the separation capillary radius. Variation in the Shape of the Initial Plug. For this study, the plug shape was varied ranging from an ideal cylindrical shape to conical shapes and to asymmetrical shapes. Here, as opposed to injecting the plug from the injection loop, the initial plug was defined at the entrance of the separation capillary; no injection process with Q inj was simulated; only the TIS-causing propagation at Q TIS was computed. The variation in concentration distributions in the initial plug was done by changing parameters k 1 and k 2 of the Heaviside function as illustrated in Figure 3A. The rest of the parameters were default (see Figure 2B). For each plug shape, a full set of separagrams has been calculated, and a K d value was determined from a respective binding isotherm (Figure S6). All determined K d values were compared to K d,inp (Figure 5C). For all examined plug shapes, K d could be accurately determined, i.e. 0.99 ≤ K d / K d,inp ≤ 1.08. The shapes of the separagrams show the two peaks corresponding to the nondiffusive species (PL) and diffusive species (L) (as in Figure 1B) for all plug shapes. All separagrams were only minimally affected by the variations in the plug shape demonstrating that ACTIS was robust towards large variations in the plug shape and proving that moderate imperfections in plug shape expected in real experiments would not affect the accuracy of K d measurements. Variation in Ramp Time of Flow Onset. The ramp time t ramp was varied from 0.10 to 40 s as illustrated in Figure 2C. The total run time was 60 s with the diffusive peak (L) arriving to the detector between 15 and 30 s. The rest of the parameters were default (see Figure 2B). For each value of t ramp , a full set of separagrams has been obtained, and a K d value was determined from a respective binding isotherm (Figure S7). All determined K d values were compared to K d,inp (Figure 5D). For all examined t ramp , K d could be accurately determined, i.e. 0.97 ≤ K d /K d,inp ≤ 1.17. The shapes of the separagrams show the two peaks corresponding to the non-diffusive species (PL) and diffusive species (L) (as in Figure 1B) for all t ramp . Overall, separagram shapes change only slightly; only positions of the non-diffusive and diffusive peaks move from 10 to 25 s and from 17 to 35 s, respectively (Figure S7). These results show that accuracy of K d determination with ACTIS is robust towards changes in ramp time of the flow onset. ## ■ CONCLUSION In this work, we examined and proved the robustness of ACTIS, in a virtual fluidic system resembling the one of a real ACTIS instrument, towards large variations in the parameters characterizing both the fluidic system and the flow. We used COMSOL to construct a virtual five-component ACTIS setup exactly mimicking the geometry of a physical assembly (Figure 2). K d determined in this setup deviated from K d,inp , by a factor of only 1.02. This five-component ACTIS setup was also used to study how accuracy of K d is affected by variations in: (i) the radius of the injection loop while keeping its volume constant, (ii) the radius of the separation capillary, (iii) shape of the initial (prior to start of TIS) plug of the equilibrium mixture, and (iv) ramp time in flowrate onset (after the start of TIS). The variations used in the test exceeded markedly the ones that could be caused by undesirable variations expected in a physical ACTIS setup. The tested variations provide wide ranges for desirable experimental instrumental adjustments which would not make ACTIS inaccurate. In general, the values of K d did not deviate from the input K d values by more than a factor of 1.25 upon variations in the above parameters. Regarding the variation of the injection loop radius, we found that K d could be accurately determined for 0.05 ≤ r 1 / r 2 ≤ 2.5. Larger ratios, i.e. r 1 / r 2 ≥ 5, corresponding to injection loop radii much greater than the separation capillary radius resulted in deviations greater than by a factor of 1.25 likely due to a combination of mesh-size limitations and non-optimal TIS conditions. Accuracy of K d was also not affected greatly by variation of the separation capillary radius if r 2 / r 1 < 10. Note that for r 2 / r 1 ≤ 0.5 the signal variation in the separagram becomes much smaller caused by the plug length being comparable to the separation capillary length (Table S2). A small signal variation will make it difficult to evaluate separagrams containing noise, e.g. from experimental data. Great deviations in K d were observed for r 2 / r 1 ≥ 10 due the departure from equilibrium of the injected sample plug, the "flow separation" phenomenon, and significant longitudinal diffusion, with deviation in K d of K d / K d,inp > 1.25. The mesh-size limitations likely also contribute to the observed deviations. Moreover, deviations in plug shape from the ideal cylindrical shape do not influence the accuracy of K d . If plug injection is reproducible from run to run, the determined K d will be accurate. Finally, large variations in the pump ramp time for flow onset do not influence the accuracy in K d , i.e. cost-effective pumps with slow ramp times can be used for ACTIS. From the above results and our experience, we recommend the following strategy to setup an ACTIS instrument. First, one should choose the propagation flow rate (Q TIS ) as well as the separation capillary dimensions (r 2 and l 2 ) so that the ligand L reaches the detector at time τ L (see eqs. 8-10) which is much shorter than anticipated lifetime of the complex (1 / k off ). Our choice of τ L ≈ 10 s should suffice most of the stable PL complexes. Second, the injection loop radius r 1 should be selected to be 0.5-1.0 times the separation capillary radius r 2 . For instance, a separation capillary radius of 100 µm corresponds to a range in injection loop radius of 50-100 µm. Third, the ACTIS instrument should be assembled using the straightforward scheme provided in Figure 2; an elaborate setup proposed in our preliminary study 1 is not needed since parameters such as the plug shape or ramp time have only minimal impact on K d accuracy according to the results of our current study. Our findings and recommendations will allow ACTISinstrumentation developers to change (simplify) instrument configuration drastically without raising questions about how such changes may affect accuracy of K d . By relying on ACTIS intrinsic accuracy, developers can focus their efforts and resources on optimizing instrument configuration to achieve the highest precision. Finally, to the best of our knowledge, this work provides the first example of a comprehensive proof of accuracy of an analytical method performed in-silico. We foresee that the in-silico accuracy-assessment approach will be used for a similar task on other methods built upon processes with the deterministic nature. ## ■ ASSOCIATED CONTENT Supporting Information The Supporting Information is available free of charge on the ACS Publications website and on ChemRxiv (DOI: 10.26434/chemrxiv.12345644). Theoretical background for computer simulation and data evaluation; Simulation of separagrams; Figure S1, Variation in k off,inp -separagrams and binding isotherms; Figure S2, Variation in injection loop dimensions -separagrams and binding isotherms; Figure S3, Variation in injection loop dimensions -sample-plug distribution; Figure S4, Variation in separation capillary radii -separagrams and binding isotherms; Figure S5, Velocity streamlines at different separation capillary radii; Figure S6, Variation in the initial plug shape -separagrams and binding isotherms; Figure S7, Variation in ramp time of flow onset -separagrams and binding isotherms; Table S1+S2, Péclet number and plug length (PDF) Evaluation files with raw data (ZIP) Model files of Numerical simulations in COMSOL (ZIP)

chemsum

{"title": "Assessing Accuracy of an Analytical Method in silico: Application to \"Accurate Constant via Transient Incomplete Separation\" (ACTIS)", "journal": "ChemRxiv"}

how_silicene_on_ag(111)_oxidizes:_microscopic_mechanism_of_the_reaction_of_o2_with_silicene

## Abstract: We demonstrate, using first-principles molecular-dynamics simulations, that oxidation of silicene can easily take place either at low or high oxygen doses, which importantly helps clarify previous inconsistent reports on the oxidation of silicene on the Ag(111) substrate. We show that, while the energy barrier for an O 2 molecule reacting with a Si atom strongly depends on the position and orientation of the molecule, the O 2 molecule immediately dissociates and forms an Si-O-Si configuration once it finds a barrier-less chemisorption pathway around an outer Si atom of the silicene overlayer. A synergistic effect between the molecular dissociation and subsequent structural rearrangements is found to accelerate the oxidation process at a high oxygen dose. This effect also enhances self-organized formation of sp 3 -like tetrahedral configurations (consisting of Si and O atoms), which results in collapse of the two-dimensional silicene structure and its exfoliation from the substrate. We also find that the electronic properties of the silicene can be significantly altered by oxidation. The present findings suggest that low flux and low temperature of the oxygen gas are key to controlling oxidation of silicene.The two-dimensional (2D) structure of silicon (Si), which is a graphene analog, is called "silicene". It has attracted intense interest since its successful synthesis in the laboratory on a variety of substrates [1][2][3][4][5][6][7][8][9] and it has been argued that silicene could have unique electronic properties such as massless Dirac Fermion behavior, as graphene does 10 . Silicene is thus expected to possess great potential for future applications such as nanoscale electronic devices. In contrast to graphene, silicene is known to take a variety of 2D configurations on a substrate (such as Ag or Al), including those with a periodicity of 4 × 4, × 13 13 11 , and even a configuration consisting of 3-, 4-, 5-, and 6-sided polygons 8 . This variety of structures gives silicene more flexible properties than graphene.Silicene also has the great advantage of easy integration into existing circuitry that is already based on Si technology. The application of silicene to nanoscale devices is, however, currently hindered by the existence of unsaturated (dangling) bonds on its surface, which makes it highly reactive under atmospheric conditions. Because of this, pristine silicene can only be grown under vacuum on a substrate; silicene is only found to be free from contamination or oxidation in air when capped by e.g., organomolecules [12][13][14][15][16] . It is, however, crucial to be able to control the stability of silicene under various conditions in order to fabricate silicene-based nanodevices. Elucidating the stability of silicene under atmospheric conditions, and especially its resistance toward oxidation, is thus urgently required.Only a few theoretical studies have examined oxidation of silicene, however, these have focused on free-standing silicene, and not silicene on a substrate [17][18][19][20] . These studies, including the work by Wang et al. 17 , Ozcelik et al. 18 and Liu et al. 19,20 , reported that free-standing silicene is unstable in O 2 as O 2 will readily dissociate on silicene, in a barrier-less process, to form stable Si-O bonds. The resulting silicene oxides may retain the honeycomb lattice of silicene or distort it. For silicene on a substrate, in contrast, previous experimental studies have reported contradictory observations on the oxidation of silicene exposed to oxygen gas (O 2 ). De Padova et al. 21 and Molle et al. 22 claimed that silicene (in the form of a nanoribbon and nanosheet, respectively) possesses low reactivity towards O 2 up to 1000 Langmuir (L). In contrast, more recent experiments by Du et al. 23,24 demonstrated that silicene on Ag(111) exhibits a high reactivity towards oxygen, observing incorporation of oxygen atoms into the silicene honeycomb network even at low gas coverages (e.g. 10-20 L). Furthermore, they obtained highly oxidized silicene under an oxygen exposure of 60 L, and even amorphous-like silicene oxides at 600 L, at which crumpling of the silicene overlayer was observed, suggesting its exfoliation from the underlying substrate 23 . This glaring contradiction highlights the need to illuminate the surface morphology and dynamical processes occurring at an atomistic level in the early stage of silicene oxidation. Such detail, however, is extremely difficult to achieve in experiment. What is needed, therefore, is a computational approach that can unveil the dynamical changes of the atomic configuration during the oxidation process. In our previous work with Xu et al. 24 we performed DFT calculations of oxidized silicene on the Ag(111) surface, that represented the structure determined from the STM experiments, and its stability at a simulation temperature of 300 K. We did not, however, model the oxidation process itself, but instead started with the oxidized overlayer. Here, we report a first-principles molecular-dynamics (FPMD) study of oxidation of 4 × 4 silicene on Ag(111) with low (~0.1 monolayer (ML)) and high (~0.44 ML) oxygen coverages. To the best of our knowledge, this is the first attempt to reveal the dynamical process of oxidation of the silicene overlayer on the Ag(111) substrate. We show that an O 2 molecule can easily react with the Si atoms in the silicene overlayer on Ag(111). In particular, an O 2 molecule is able to find barrier-less pathways to oxidation when it is within a distance of ~3 from a substrate-side Si atom (i.e. a Si atom that is located closest to a substrate atom). Sequential displacements or movements of specific atoms, which we call a "chain-like reaction", are found to play a significant role in the oxidation process at high oxygen coverages. Also shown in our work, is that the electronic properties of silicene can be significantly altered by oxidation, which gives us some hints for the potential applications of silicene oxides. We note that there are two types of 4 × 4 superstructures of silicene on the Ag(111) surface, namely 4 × 4-α 3,25 and 4 × 4-β 25 . Since oxidation of the former silicene overlayer was examined in most of the previous experiments , we use the same 4 × 4 structure for silicene on Ag(111) in our simulations reported here. ## Results Oxidation at a low oxygen dose. A single O 2 molecule was introduced into the system of 18 Si and 80 Ag atoms to model the reaction of oxygen with the 4 × 4 silicene overlayer at a low oxygen coverage. This corresponds to an O 2 coverage of 0.11 ML. Firstly, we constructed the energy profile of the O 2 /silicene/Ag(111) system as a function of the distance between the silicene overlayer and the O 2 molecule. The following three arrangements are considered (Fig. 1): (A) an O 2 molecule approaching an outer Si atom of the silicene overlayer, keeping the molecular axis parallel to the silicene surface, (B) an O 2 molecule approaching an outer Si atom, keeping the molecular axis perpendicular to the surface, and (C) an O 2 molecule approaching an Ag-side Si atom, keeping the molecular axis parallel to the surface. At each distance, the positions of the O atoms are allowed to relax (while keeping the molecular orientation fixed), in addition to relaxing the positions of the Si and Ag atoms. The distance between the silicene overlayer and the O 2 molecule was measured from the Ag-side Si atoms (denoted by the red lines in the insets of Fig. 1. In Case B, the distance is measured from the Ag-side Si atom to the Ag-side O atom). The energy profile for Case A and Case B in Fig. 1 exhibits a monotonic increase in energy as the O 2 molecule approaches the silicene overlayer while keeping its molecular orientation. The energy barriers are estimated to be 0.16 eV for Case A at a distance of 3.1 , and 0.20 eV for Case B at a distance of 2.9 . The sharp drop in energy seen at shorter distances is attributed to the formation of Si-O bonds, where the outer Si atom is lifted upward to form the Si-O bonds. In contrast to Case A and Case B, such an energy drop is not seen for Case C, wherein an Si-O bond is not formed. In fact, the O 2 molecule moved horizontally as if it experienced a repulsive force from the Ag-side Si atom when the distance was decreased to ~2.6 . This indicates that the reactivity with oxygen is quite different between the outer-and Ag-side Si atoms, which would come from the fact that the dangling bond on the latter is effectively removed by the Ag atoms while that on the former is left intact. It is clear from Fig. 1 that there exists an energy barrier when an O 2 molecule approaches the silicene overlayer while keeping its orientation vertical or parallel to the surface. The barrier, however, may be reduced by approaching the surface in different orientations. We thus performed FPMD simulations (under the NVE condition) to see if the O 2 molecule itself was able to find other pathways with lower (or no) barriers to react with the Si atoms. Figure 2 shows the snapshots from the FPMD simulations runs that were performed for Case A and Case B, along with the time evolution of the energy from the same FPMD runs. The simulations were started with the O 2 molecule located at a distance of 3.26 (for Case A) or 3.10 (for Case B). We found that, even with a zero initial velocity, the O 2 molecule reacted with the Si atoms to form an Si-O-Si configuration in both cases. The Si-O bond length in both the Case A and Case B final configurations ranged from 1.63 to 1.76 , which is slightly longer than the Si-O bond length in crystalline SiO 2 (1.62 ), but is in good agreement with the previous theoretical results of free-standing silicene . The slightly longer Si-O bond length in the oxidized silicene is due to the high buckling of the honeycomb lattice incorporating the O atoms. The energy significantly decreases in both reaction processes, in which an energy equivalent to ~270 K is released indicating this is an exothermic process. Our FPMD simulations, therefore, clearly show that there are low-energy barrier pathways for the reaction of O 2 with silicene. The formation of a configuration where an outer Si atom has a tilted O 2 molecule adsorbed on it is the key to finding a pathway to the oxidation reaction. The formation of such a configuration could be rationalized as follows; electron donation from the Si atoms takes place when an O 2 molecule dissociates, which stabilizes the Si-O bond. (This is consistent with a previous calculation of the oxidation of Si(001) 26 , and will also be demonstrated in our calculations later.) As an O 2 molecule is considered to have a double bond, such a transfer of electrons may easily occur when an O atom of the molecular oxygen adsorbs close to an outer Si atom. Therefore, the surface bound Si-O 2 configuration as in Fig. 2A:(1) is favored in the oxidation process. This, in turn, makes the hollow site of the Si 6 ring less favorable to the O 2 molecule. Note that the center of the O 2 molecule, on the other hand, does not favor the outer Si atom, because it is unlikely that electron transfer will take place between them due to electron repulsion. This is consistent with the fact that there exists an energy barrier when the center of an O 2 molecule approaches the outer Si atom orientated parallel to the surface, as in Case A (see Fig. 1). Also note that a barrier-less reaction leading to the formation of the Si-O 2 configuration has also been observed in a previous DFT calculation of free-standing silicene 20 , suggesting that the Ag(111) substrate does not greatly affect the reactivity of the silicene overlayer toward an O 2 molecule. We also performed a FPMD run for Case C with the O 2 molecule initially located at a distance of 2.64 (see Fig. 1) from the silicene. The O 2 molecule immediately started moving away from the silicene overlayer even with a zero initial velocity. This markedly contrasts with Case A and Case B. We repeated the FPMD simulation with a slightly different initial position of the O 2 molecule (where it started at a distance of 3.1 and a different orientation of the molecular axis, while still keeping it parallel to the silicene overlayer). The molecule again moved away from the surface without adsorbing or reacting with the silicene. It thus appears that there is not a barrier-less pathway to oxidation around the Ag-side Si atoms, or the pathway, if any, is very narrow for O 2 to enter without a guiding force. This finding is consistent with the recent experimental observation that oxidation always starts with the outer Si (top-layer Si) atoms 23 . The different behavior observed in Case A, B and C indicates that the energy landscape felt by the O 2 molecule above the silicene overlayer is highly rugged and depends on the surface morphology, so that exploring the landscape is necessary for O 2 to react with the silicene. This means that the O 2 molecule should have a certain amount of kinetic energy to be able to explore the landscape. Oxidation may thus be possibly suppressed if the temperature can be kept very low. We note that such a dependence of chemical reactivity on the surface morphological details is also observed in the oxidation process of the Si(001) surface 26 . While the resultant configurations obtained in Case A and Case B are slightly different, the Si-O-Si configuration is formed in both cases, and is consistent with the previous experimental result 23 . The energy difference between these configurations at 0 K is only ~0.17 eV/system (i.e. ~17 K); the "B configuration" has lower energy. Because of this small difference, either configuration could be formed in experiment at low oxygen coverages. Note that we confirmed that formation of the Si-O-Si configuration is not accidental by repeating the FPMD simulations multiple times for Case A and Case B, as well as for Case C. This suggests that the reactions we present here are not a statistical anomaly. The electronic properties of the O 2 molecule before and after its reaction with the silicene (corresponding to the B:final structure) are presented in Fig. 3. The upper panel of Fig. 3 shows the density of state (DOS) for the p electrons of the O 2 molecule before the reaction, in the configuration displayed in the inset panel. The molecule is clearly spin-polarized, as is expected for the ground state of an O 2 molecule (triplet state). In contrast, the DOS for the dissociated O atom, corresponding to the O-Si-O adsorbed configuration (inset panel of the lower DOS plot) shows a spin-unpolarized characteristic; the up-spin and down-spin bands are fully degenerate due to the formation of the Si-O-Si configuration. The lower panel of Fig. 3 shows a plot of the electron localization function (ELF) 27 that gives an indication of the probability of electron localization as measured between 0 and 1, with 1 showing a high probability of covalent bonding. The subpanels (a), (b1), and (b2) display the color plot of the ELF for the slices indicated in the insets of the DOS panels of Fig. 3 [Slice (a) includes the whole O 2 molecule, while Slices (b1) and (b2) include the Si-O and Si-Si bonds, respectively, with the O and Si atoms that the slice passes through denoted by the arrows]. The notable point is that the electron localization in the Si-Si bond (b2) remains high, while that in the Si-O bond (b1) is lower. This indicates that the Si-O bond is rather ionic compared to the Si-Si bond. In fact, our Bader charge analysis shows that each of the O atoms gains electrons of ~1.5|e| from its adjacent Si atoms after forming the Si-O-Si configuration. Thus the electronic property associated with the intermediate sp 2 /sp 3 bonding in the silicene overlayer is significantly altered by the oxidation. Oxidation at a high oxygen dose. Our FPMD calculations show that sequential reactions involving more than one O 2 molecule are the key to understanding the oxidation process of silicene at high oxygen coverages. Specifically, the oxidation process is dominated by a "chain-like reaction" involving multiple O 2 molecules, which will be detailed below. In the FPMD simulation performed at a high oxygen coverage, 16 O 2 molecules are introduced into the system of 72 Si and 320 Ag atoms, corresponding to an oxygen coverage of 0.44 ML on Si/Ag(111). This coverage is consistent with our previous preliminary calculations 24 that corresponded to a coverage of 0.5 ML, and is sufficient to follow the dynamical changes of the atomic configuration in the early stage of the oxidation. All the O 2 molecules were initially evenly positioned across the lateral directions of the surface, above the silicene overlayer (See Fig. S1 of the Supporting Information). The distances between the O 2 molecules and the silicene overlayer were in the range of 4.42-6.43 . The initial velocities at ~300 K were assigned to each O 2 molecule, and were adjusted to move the O 2 molecules toward the silicene overlayer. The 4 × 4 structure, 3 without any defects, was employed as the initial structure for the silicene overlayer on the Ag(111) surface. The initial velocities of the Si and Ag atoms were also set to have a kinetic energy of 300 K, and the temperature of the whole system was then kept at 300 K throughout the FPMD run. We note that the effective oxygen flux in the simulation would be extremely high compared to experiment because of the short simulation time (3.5 ps). A typical "chain-like reaction" process is displayed in Fig. 4, wherein a series of reactions proceeds from A through to F. The white circles in the panels of Fig. 4 denote the O 2 reactions or structural rearrangements focused on in each panel. The same oxygen atoms are included in the white circles throughout A to F. We found, that by looking at the process in detail, the following three steps play an important role in the oxidation at high O 2 coverages: the structural rearrangements [E, F] (for example, an under-coordinated Si atom is lifted up toward the vacuum space in the substantial structural rearrangement as in Fig. 4(F′), which is a highly reactive site towards O 2 molecules). When an O 2 molecule approaches an outer Si atom, the O 2 molecule immediately dissociates forming Si-O bonds; this step is followed by structural rearrangements that create new reaction sites. This series of actions enhances the subsequent O 2 reactions, since available reaction sites are repeatedly formed, resulting in a "chain-like" oxidation process. (In fact, we found that a "chain-like reaction" similar to Fig. 4 took place simultaneously on a different part of the silicene overlayer. See Fig. S3.) This reaction leads to collapse of the original honeycomb 2D structure and formation of 3D-like bonded configurations [e.g. see Fig. 4(F′)]. Since highly oxidized silicene showed a tendency towards exfoliation from the Ag substrate in the previous experiment 23,24 , we suggest that the growth of such 3D-like configurations may be responsible for the exfoliation. It is not surprising that the 3D-like structural arrangements are formed during the oxidation process, considering that an O-Si-O configuration favors the tetrahedral bonding that is seen in the SiO 2 crystal. Figure 5 shows the tetrahedrality of the structural configuration composed of a Si atom with its four neighboring atoms (either Si or O). Here, the order parameter q t is calculated to estimate the tetrahedrality 28,29 , q k t is defined for each Si atom (kth atom) at each MD step as ( ) , where θ k ij is the angle between the vectors that join a central Si atom with its ith and jth nearest neighbors ( ≤ j 4). q t is then obtained by averaging q k t over all the Si atoms as , where N k is the number of the Si atoms. The upper panel of Fig. 5 shows the time evolution of q t during the oxidation process. q t takes a value of 0 (by definition) before the O 2 reactions start since all the Si atoms have three neighboring atoms only. Once the reactions start (at a time step of ~200), however, O atoms become incorporated into the Si honeycomb network, forming Si-O bonds and inducing the structural rearrangements. This results in the sharp increase and subsequent steady growth of q t , indicating that the number of tetrahedral configurations monotonically increases. That is, the degree of sp 2 character of the bonding in the silicene 2D structure is gradually reduced, and a tetrahedral 3D structure having an enhanced sp 3 character grows instead. Interestingly, the growth of q t does not show a direct correlation with that of the coordination number of Si (N c ). The lower panel of Fig. 5 shows the time evolution of N c [counting both Si and O atoms (red lines), or O atoms only (green lines) as the neighboring atoms]. Similarly to q t , N c sharply increases and then grows steadily. The monotonic growth, however, ceases at a time step of ~1100, followed by only a slight increase. The number of O atoms counted as the neighboring atoms (green line) is rather constant after the growth stage. This reflects the fact that most of the O 2 molecules in the system had already reacted with the silicene overlayer and had been incorporated into the Si-Si bond network at the growth stage. As can be seen in the upper panel of Fig. 5, however, q t continues increasing even after the growth stage (at a time step > 1100). It is thus considered that the structural rearrangement recovering the tetrahedrality can proceed without the supply of O 2 molecules, once it is triggered by oxidation. This implies that a high oxygen flux within a short timeframe would be sufficient to enhance the formation of the 3D oxidized configurations and thus exfoliation of the silicene overlayer. The total N c (counting both Si and O atoms as the neighboring atoms) shows even a slight decrease at time step 1100-1500. This indicates that under-coordinated Si atoms are newly generated in the structural rearrangements, as has been discussed before [e.g. Fig. 4(F′)]. The total N c then exhibits a slight increase again after a time step of ~1500, induced by further rearrangements or capture of O 2 molecules that were still intact. It is worth noting that, because of this high oxygen dose, some of the O 2 molecules could form dimers or trimers via direct intermolecular interactions before reacting with the silicene overlayer. Such O 2 aggregates facilitate the occurrence of the chain-like reaction. This indicates that uniformly oxidized silicene may not be easily obtained at a high oxygen coverage, but instead would result in exfoliation and emergence of the bare Ag surface. A low flux, with a long time exposure to oxygen may thus be needed to form uniformly oxidized silicene and hence formation of a silicene oxide sheet. The atom-resolved DOS for the oxidized Si atoms at the high O 2 coverage (Fig. 6) clearly show the change in electronic properties induced by the oxidation. Figure 6(a) shows the atom-resolved DOS for a Si atom having three neighboring Si atoms, as in the silicene honeycomb lattice. It is clear that the electronic bands near the Fermi energy have a high intensity and are dominated by the p z electrons from the dangling bond on the Si atom, and show a metallic nature. In contrast, the atom-resolved DOS for a four-coordinated Si atom (with two O atoms and two Si atoms) having a highly tetrahedral configuration [Fig. 6(b)] shows completely different characteristics with the electronic bands near the Fermi energy being substantially reduced, especially those from the p z electrons, due to capping the dangling bond with O atoms. We thus conclude that the metallic nature of silicene is reduced as oxidation proceeds. This tendency has also been observed in the recent experimental study 23 , which reports that a semiconducting nature for silicene could be realized by the oxidation process. ## Discussion The present FPMD calculations reveal that there exist barrier-less oxygen chemisorption pathways around the outer Si atoms of the silicene overlayer. Though the pathways are not significantly wide for the O 2 molecule, oxygen can easily react with a Si atom to form an Si-O-Si configuration, once the molecule finds an entrance to the pathway on the rugged energy landscape provided by the silicene overlayer. The resultant Si-O bond is ionic, rather than covalent, because of the charge transfer from the Si atom to the O atom. As a result, the nature of the intermediate sp 2 /sp 3 bonding is substantially degraded, and a tetrahedral 3D configuration as seen in SiO 2 crystals locally forms as the oxidation proceeds. In the oxidation process involving multiple O 2 molecules (representing a high oxygen dose), a synergistic effect between the molecular dissociation and the subsequent structural rearrangement is the key to understanding the atomistic mechanism of the oxidation process. A notable point is that the structural transformations resulting in highly tetrahedral configurations composed of Si and O atoms can proceed without a supply of O 2 molecules, once the silicene overlayer is covered by oxygen of > ~0.5 ML. This self-organized rearrangement should be one of the driving forces to accelerate exfoliation of the silicene overlayer from the Ag substrate at a high oxygen dose. Careful control of the oxygen flux is thus necessary to produce an oxidized silicene sheet maintaining its 2D morphology. Also, suppression of the oxidation process might be possible by maintaining the oxygen gas temperature at a low value. Significantly, our results give some hints to help explain why the differences seen experimentally for the oxidation of silicene on the Ag(111) surface arose. We suggest from our work that a number of factors, such as oxygen coverage or dose, as well as reaction temperature, may alter the degree of oxidation of silicene. In particular, our results indicate that a different flux (or pressure) of oxygen gas could induce different oxidation processes. As the unit of Langmuir is the pressure of the gas times the time of exposure, different pressure conditions may lead to the same value in L. This means that experiments reporting the reaction of O 2 with the silicene at exposures with the same Langmuir value may actually be using different pressure conditions. It is therefore highly possible that oxidation could proceed differently in the experiments given the same exposure. Further, since the electronic properties can be altered by oxidation, as demonstrated in this work, control of the process is highly desirable so as to obtain non-oxidized, partially-oxidized, or fully-oxidized silicene. The present results are thus of great help to realize such control and to extend the potential range for the use of silicene in nanoscale devices under a variety of conditions, including metal/oxide semiconductor devices. ## Methods The present FPMD calculations were performed within the framework of density functional theory as implemented in the Vienna Ab Initio Simulation Package (VASP) 30 . The exchange-correlation functional in the Perdew-Burke-Ernzerhof form 31 was used and the ion-electron interaction was described by the projector augmented wave method 32 . Two systems were modeled based on the unit structure of 4 × 4 silicene on the Ag(111) surface (3 × 3 honeycomb silicene lattice on the 4 × 4 Ag(111) surface) with dimensions of 11.6496 × 11.6496 × 30.0 . One system is for a low oxygen coverage (0.11 ML) composed of the 1 × 1 unit structure (consisting of 2 O, 18 Si, and 80 Ag atoms), while the other system represents a high oxygen coverage (0.44 ML) composed of a 2 × 2 unit structure (consisting of 32 O, 72 Si, and 320 Ag atoms). The Ag substrate consists of five atomic layers with the bottom layer fixed. This model has been validated previously 8,24 . A plane-wave basis set with an energy cutoff of 400 eV was used with the following k-point mesh; 11 × 11 × 1 for the DOS calculations and 2 × 2 × 1 for the MD calculations of the This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

chemsum

{"title": "How silicene on Ag(111) oxidizes: microscopic mechanism of the reaction of O2 with silicene", "journal": "Scientific Reports - Nature"}

base-triggered_self-amplifying_degradable_polyurethanes_with_the_ability_to_translate_local_stimulat

## Abstract: A new type of base-triggered self-amplifying degradable polyurethane is reported that degrades under mild conditions, with the release of increasing amounts of amine product leading to self-amplified degradation.The polymer incorporates a base-sensitive Fmoc-derivative into every repeating unit to enable highly sensitive amine amplified degradation. A sigmoidal degradation curve for the linear polymer was observed consistent with a self-amplifying degradation mechanism. An analogous cross-linked polyurethane gel was prepared and also found to undergo amplified breakdown. In this case, a trace amount of localized base initiates the degradation, which in turn propagates through the material in an amplified manner. The results demonstrate the potential utility of these new generation polyurethanes in enhanced disposability and as stimuli responsive materials. ## Introduction "Smart" polymers that degrade in response to external triggers have found applications in many felds, including drug delivery, transient electronics, encapsulation and sensing. More recently, chain-shattering degradable polymers and selfimmolative degradable polymers have attracted considerable attention because their backbones can be completely degraded into small fragments with high sensitivity to different types of triggers including pH, light, and redox agents. However, both types of polymers require a stoichiometric amount of triggering agent and degradation rates are constant at best. Some selfimmolative polymers suffer from slow or incomplete breakdown because side reactions occur as the degradation proceeds down the polymer backbone. An alternative approach uses autocatalytic degradation chemistry wherein a specifc catalytic trigger causes chain cleavage and generation of additional triggers for acceleration of the degradation. Phillips and coworkers reported that ROMP polymers with appropriate pendant chains could exhibit dramatic changes in macroscopic properties through amplifed, self-propagating side-chain reactions. 22,23 In particular, a global switch in hydrophobicity and a change in the optical properties of a flm occurred with local stimulation. In an effort to develop polymeric materials that might degrade with accelerated rate profles and inspired by acid amplifer small molecules, we recently reported poly(3-iodopropyl)acetals that breakdown liberating HI. 24 In essence such polymers carry the seeds of their own destruction, 25 with liberated acid catalyzing further cleavage in an autocatalytic loop. It is important to determine the generality of the autocatalytic polymer degradation strategy by developing breakdown pathways using other triggers such as base, light and redox agents. Herein we report a new polyurethane that undergoes self-amplifed degradation mediated by base and further show that in analogous gels, a small localized addition of base leads to rapid long-range breakdown (Fig. 1). ## Results and discussion As recently noted, base-degradable polymers are underdeveloped relative to acid-degradable polymers. 26,27 In designing auto-catalytic base-degradable polyurethanes, the base ampli-fers reported by Ichimura and others were considered. Within this class of small molecules, the Fmoc protected carbamate offered a convenient aromatic scaffold for functionalization and the potential for conventional polyurethane synthesis. The actual polyurethanes studied, 1 and 1c, were prepared in six steps as shown in Scheme 1. Functionalization of the fluorene ring was achieved through a Friedel-Crafts acylation after protection of the alcohol group with acetic anhydride, thus affording 4 or 5. Acidic deprotection and reduction with BH 3 $THF produced intermediate 6 and 7, which were further converted to diol monomer 8 and 9, respectively, by selectively reducing the benzylic alcohol group with Et 3 SiH. Traditional polycondensation was performed with a 1 : 1 ratio of diol monomer and hexylmethylene diisocyanate to afford polymer 1 and 1c. As illustrated in Scheme 2, the addition of base can abstract the weakly acidic fluorenyl methine proton on the polymer 1 backbone, followed by E1cB elimination and decarboxylation to generate a dibenzofulvene and stoichiometric amine that can catalyse additional cleavage reactions before or after addition to the dibenzofulvene unit. The two polyurethanes 1 and 1c are identical structurally except that control polymer 1c is unable to undergo base-triggered degradation because the additional methylene group prevents the E1cB elimination from occurring. Both 1 and 1c were characterized by gel permeation chromatography (GPC) with DMF as the eluent (Fig. S6 and S7 †). Polymer 1 has a M n ¼ 22 kDa (Đ ¼ 2.1) and control polymer 1c has M n ¼ 11 kDa (Đ ¼ 2.6). The 1 H NMR was consistent with the expected structure of 1 (Fig. S3 †) and 1c (Fig. S4 †). Thermal gravimetric analysis (TGA) of polymer 1 and polymer 1c revealed the onset of thermal degradation to occur around 120 C and 280 C respectively (Fig. S25 and S26 †) and the onset thermal temperature at 120 C of polymer 1 correlates well to what Simeunovic and his coworkers reported. 33 The T g of polymer 1 and 1c were determined to be 61 C and 45 C, respectively, the latter value measured by differential scanning calorimetry (DSC) (Fig. S27 †). Several bases were found to trigger the autocatalytic degradation of 1 (Fig. S11 †), with hexylamine chosen for further study because its basicity and steric hindrance is most similar to the amplifed amine species. Thus, the base-triggered degradation of polymers 1 and 1c in DMF solution was initiated by the addition of hexylamine and monitored by gel permeation chromatography (GPC). When 1 was exposed to 5 mol% hexylamine (per repeat unit), it showed a progressive and signifcant decrease in molecular weight over a 12 h period. As seen in Fig. 2a, the reduction in polymer size over time is nonlinear. Thus, the retention time of the 1 shifts only 1 min during the frst 2 h but between 6 h and 9 h signifcantly broadens and shifts to longer times. In contrast, under the same conditions, the GPC of polymer 1c remained unchanged over 24 h (Fig. S8 †). 1 H NMR was used to monitor the molecular details of the degradation of polymers 1 and 1c in the presence of hexylamine in DMSO-d 6 solution. Consistent with the GPC study, no change in the NMR of 1c was observed over 24 h with 5 mol% hexylamine (Fig. S10 †). In the case of 1, addition of 5 mol% of hexylamine led to the simultaneous disappearance of the methine and methylene protons labelled a and b at d 4.33 and 4.16 ppm, respectively and the appearance of alkene protons at d 6.25 ppm from the dibenzofulvene elimination product (Fig. 2b and S9 †). To determine how the concentration of the base trigger affects the rate of the degradation, quantitative 1 H NMRmonitored kinetics were carried out in the presence of 0.5 mol%, 1 mol%, 5 mol%, 20 mol% and 100 mol% hexylamine. As seen in Fig. 2c, a stoichiometric amount of hexylamine induced complete polymer degradation at room temperature within 1 h. The rate profle and time for complete degradation correlated with the amount of base trigger. Thus, with no added base the polymer was stable, whereas for 0.5 mol%, 1 mol%, 5 mol%, 20 mol% and 100 mol% hexylamine the degradation reached 90% at ca. 15 h, 12 h, 10 h, 2 h, and 47 min, respectively. Most exciting was the observation that the three lowest concentration hexylamine experiments (0.5 to 5 mol%) exhibited obvious induction periods and sigmoidal conversion curves indicative of autocatalytic degradation. Additional support for the autocatalytic, base amplifcation mechanism came from ftting the degradation data of polymer 1 to an autocatalytic kinetic model (eqn (S3) †). 23,24, In this model, rate constants k 1 and k 2 separately represent the nonautocatalytic and autocatalytic, amine-accelerated rate constants (see ESI † for details). Consistent with the mechanism shown in Scheme 2, ftting the sigmoidal curves seen in Fig. 2c, led to k 2 values that were quite close and k 2 c 0 values that are larger than the k 1 values (Table 1). The latter is especially true for the 0.5 mol% hexylamine run, in which the k 2 c 0 (6.7 10 3 min 1 ) is 30 times larger than k 1 (2.1 10 4 min 1 ). This larger k 2 c 0 value is characteristic of an autocatalytic reaction (Table 1 and Fig. S14-S16 †). To further characterize the degradation of 1, liquid chromatography coupled mass spectrometry (LC-MS) was utilized to identify the major degradation products, and further indicate the chemical structure of the polymer repeating units. Analysis of the degradation products from polymer 1 through LC-MS revealed two major peaks, degradation product 1 with higher intensity appearing at 6.4 min (m/z ¼ 393.4) and degradation product 2 with a lower intensity appearing at 9.3 min (m/z ¼ 669.4) (Fig. S12 †). These products are consistent with two types of repeating units in 1 (Fig. 3) and degradation product 1 demonstrates the ability of polymer 1 to form amine products via Fmoc deprotection. Polyurethanes are important and widely used polymeric materials commonly found in plastics, adhesives and coatings. 37,38 Unlike polymer 1, these materials are usually prepared from a polyol that produces cross-linking. The combination of cross-linking and the stability of the urethane linkage makes polyurethanes highly durable but also limits their end-of-life breakdown. To examine whether the base-amplifed degradation might be applicable to bulk materials, triol 6 was prepared (see ESI †) and polymerized with hexamethylene diisocyanate and dibutyltindilaurate (DBTDL) as catalyst in N-methylpyrrolidone (NMP) with bromothymol blue present to visualize the gel and provide a pH indicator (Fig. 4b and S17 †). The polymerization was performed at room temperature in a circular Teflon mold for 24 h to give a polyurethane flm of 11 with a 500 mm thickness. To characterize the polymer flm, it was immersed in additional NMP which induced signifcant swelling, but did not dissolve the gel. This observation is consistent with a crosslinked gel. To demonstrate the urethane network, the polymer flm was dried under high vacuum and characterized by attenuated total reflection infrared spectroscopy (ATR-FTIR). The absorption peaks at 1694 cm 1 and 1252 cm 1 were assigned to the urethane structure and the absorption peak at 3326 cm 1 was assigned to unreacted hydroxyl groups in the polymer network (Fig. S18 †). 39 Degradation study of the polymer flm was performed with flm being swelled by NMP solution. In the degradation study, the centre of the polymer flm changed from yellow to blue after 2 mL of a 180 mM hexylamine NMP solution was added in the centre and photographs were acquired over time (Fig. 4c). It was observed that the degradation area kept increasing, producing a deep blue colour, suggesting the formation of increasing numbers of terminal amino groups with conversion of the bromothymol blue pH indicator to its blue coloured ring open form. Quantifcation of the degradation area using the blue colour change for the polymer flm was assisted by Image-Pro Plus (Fig. 4d). An increase in degradation area also simulates a sigmoidal curve, with a nonlinear increase from 10% at 100 min to 90% at 300 min, which is consistent with an autocatalytic degradation process for the crosslinked gel. The complete degradation of the polymer flm required 420 min and over this period the yellow solid flm became a blue solution The degradation process was also monitored by rheology. The storage modulus of the gel was measured and no major rheological change was observed from the polymeric network without addition of the base trigger (blue curve, Fig. 4e and S23 †). However, the bulk polymeric network underwent a rapid decrease in storage modulus from about 5300 Pa to nearly 0 Pa upon addition of a very small amount of a dilute hexylamine solution in NMP at room temperature (red curve, Fig. 4e and S23 †). In this case autocatalytic equations (eqn (S4) and (S5) †) that relate the storage modulus to degradation time were utilized to quantify the gel breakdown kinetics as described in more detail in the ESI. † In particular, these equations relate the storage modulus decrease to the concentration of crosslinks, thus enabling inference of apparent chemical rate constants. The ftting of triplicate runs (Fig. S24 †) gave k 2 c 0 ¼ 15.9 AE 5.3 min 1 , which is much larger than the k 1 ¼ 2.1 10 3 AE 1.1 10 3 min 1 . These observations are consistent with an autocatalytic degradation process. ## Conclusion In conclusion, we developed a new type of self-amplifying degradable polymer with self-accelerating degradation properties using the well-developed base-sensitive Fmoc protecting group used in peptide synthesis. The incorporation of Fmoc in every repeating unit provides extremely sensitive polymeric materials with a small amount of base leading to rapid and amplifed degradation. The base amplifcation process may be useful in applications where rapid production of an amine base is desirable. The crosslinked gel provides a rare example where a tiny local stimulation generates long range, rapid macroscopic degradation. In principle such a degradation might propagate over very large distances. Our current efforts are focused on generalizing this self-amplifed degradation process to other kinds of triggers such as light, ions, and ROX agents. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Base-triggered self-amplifying degradable polyurethanes with the ability to translate local stimulation to continuous long-range degradation", "journal": "Royal Society of Chemistry (RSC)"}

synthesis_of_phosphorothioates_using_thiophosphate_salts

## Abstract: Reactions of O,O'-dialkyl thiophosphoric acids with alkyl halides, in the presence of a base, provide a direct synthetic route to phosphorothioates via O,O'-dialkyl thiophosphate anion formation. Studies on the reaction of ambident nucleophile ammonium O,O'diethyl thiophosphate with benzyl halides and tosylate in different solvents show that only S-alkylation is obtained. Reaction of this ambident nucleophile with benzoyl chloride (a hard electrophile), gave the O-acylation product. A simple, efficient, and general method has been developed for the synthesis of phosphorothioates through a one-pot reaction of alkyl halides with the mixture of diethyl phosphite in the presence of triethylamine/sulfur/and acidic alumina under solvent-free conditions using microwave irradiation. ## Introduction Organophosphorus compounds have found a wide range of application in the areas of industrial, agricultural, and medicinal chemistry owing to their biological and physical properties as well as their utility as synthetic intermediates. The synthesis of phosphate esters is an important objective in organic synthesis, since they have found use in the preparation of biologically active molecules, and also versatile intermediate in synthesis of amides and esters. Among the phosphate esters, phosphorothioate derivatives are of interest as effective pesticides. In recent years a number of phosphorothioates have been introduced as potential chemotherapeutic agent. Despite their wide range of pharmacological activity, industrial and synthetic applications, the synthesis of phos-phorothioates has received little attention. The following methods, not generally applicable, have been reported in the literature: (i) reaction of dialkyl phosphites with sulfenyl chlorides, sulfenyl cyanides, thiosulfonates, disulfides, and sulfur, (ii) condensation of phosphorchloridate with thiols and (iii) redox-type reactions of phosphorus triesters with thiols in the presence of tellurium (IV) chloride. However, all of these methods have problems, including drastic reaction conditions and also some severe side reactions. Surface-mediated solid phase reactions are of growing interest because of their ease of set-up, workup, mild reaction conditions, rate of the reaction, selectivity, high yields, lack of solvent and the low cost of the reactions in comparison with their homogeneous counterparts. The application of microwave energy to accelerate organic reactions is of increasing interest and offers several advantages over conventional techniques. Synthesis of molecules that normally require long reaction times, can be achieved conveniently and very rapidly in a microwave oven. As a part of our efforts to explore the utility of surface-mediated reactions for the synthesis of organophosphorus compounds, we report a new method for the preparation of phosphorothioates by reaction of diethyl phosphite with alkyl halides in the presence of a mixture of ammonium acetate/sulfur/alumina under solvent-free conditions using microwave irradiation which produces high yields of phosphorothioates (Scheme 1). ## Results and Discussion Recently we have found that ammonium O,O'-diethyl thiophosphate can be obtained by reaction of diethylphosphite in the presence of a mixture of ammonium acetate/sulfur/acidic alumina under solvent-free conditions using microwave irradiation. This reagent can be used as an efficient reagent for the conversion of epoxides to thiiranes. This ambident nucleophile has two potentially attacking atoms (S or O) and can attack with either of them, depending on conditions, and mixtures are often obtained in the reaction with electrophilic centers (Scheme 2). Scheme 2: Ambident nucleophile ammonium O,O'-diethylthiophosphate We have found that the reaction of diethyl phosphite with alkyl halides in the presence of a mixture of ammonium acetate/ sulfur/alumina under solvent-free conditions using microwave irradiation produces high yields of phosphothioates (S-alkylation, Scheme 1). We decided to investigate the reaction of this ambident nuclophile under different conditions (different leaving groups and solvents). Firstly, we introduce a novel method for large-scale synthesis of ammonium O,O'-diethyl thiophosphate. The reaction of sulfur with diethylphosphite in the presence of ammonium hydrogen carbonate under reflux condition in a solvent mixture of ethyl acetate and diethyl ether (1:1) gave ammonium O,O'-diethyl thiophosphate in quantitative yield (Scheme 3). ## Scheme 3: Synthesis of ammonium O,O'-diethyl thiophosphate The results of the reaction of this reagent with benzyl bromide, chloride and tosylate in different aporotic and protic solvents show that S-benzyl O,O'-diethyl phosphorothioate (S-alkylation) was formed as sole product (Scheme 4). ## Scheme 4: Solvent and leaving group effects on the synthesis phosphorothioates We conclude here that changing of leaving group and use of different media gives no O-alkylation product (i.e. changing from soft to hard leaving group and aprotic to protic solvent). Although ammonium O,O'-diethyl thiophosphate is a potential ambident nucleophile, only its soft center is reactive in this case. Recently the synthesis of S-thioacyl dithiophosphates has been reported as an efficient and chemoselective thioacylating agent using the reaction of acyl chlorides with dithiophosphoric acid in the presence of pyridine or triethylamine. In another study we decided to investigate the reaction of the ambident nucleophile ammonium O,O'-diethyl thiophosphate salt with acyl chlorides. Reaction of ammonium O,O'-diethyl thiophosphate with benzoyl chloride, as a model compound, in acetonitrile gave benzamide as the major product (Scheme 5). We conclude that replacement of benzyl with benzoyl group (hard electrophilic center) gives the O-acylation product. As a part of our efforts to explore the utility of surface-mediated reactions for the synthesis of organophosphorus compounds, herein we report a new method for the preparation of phosphorothioates by reaction of diethyl phosphite with alkyl halides in the presence of a mixture of triethylamine/sulfur/alumina under solvent-free conditions using microwave irradiation. We found that a mixture of alumina, sulfur, diethylphosphite and triethylamine under microwave irradiation gave triethylammonium O,O'-diethyl thiophosphate that can be used for the synthesis of phosphorothioates under solvent free conditions (Scheme 7, Table 1). As shown in Table 1, a wide range of alkyl halides in the presence triethylamine/ sulfur/alumina reacted with diethyl phosphite, giving the required products 2 in moderate to good yields. Scheme Synthesis of phosphorothioates using triethylammonium O,O'-diethyl thiophosphate using microwave irradiation. In summary, a simple work-up, low consumption of solvent, fast reaction rates, mild reaction conditions, good to excellent yields, relatively clean reactions with no tar formation make these methods an attractive and a useful contribution to present methods for the preparation of phosphorothioates. Studies on the reaction of ambident nucleophile ammonium O,O'-diethyl thiophosphate with benzyl halides and tosylate in different solvents show that only S-alkylation will be obtained as sole product. Reaction of this ambident nucleophile with benzoyl chloride (hard electrophilic center), gave the O-acylation product. ## Supporting Information Supporting Information File 1 The additional file contains full experimental details [http://www.beilstein-journals.org/bjoc/content/ supplementary/1860-5397-2-4-S1.doc]

chemsum

{"title": "Synthesis of phosphorothioates using thiophosphate salts", "journal": "Beilstein"}

lithium-decorated_borospherene_b40:_a_promising_hydrogen_storage_medium

## Abstract: The recent discovery of borospherene B 40 marks the onset of a new kind of boron-based nanostructures akin to the C 60 buckyball, offering opportunities to explore materials applications of nanoboron. Here we report on the feasibility of Li-decorated B 40 for hydrogen storage using the DFT calculations. The B 40 cluster has an overall shape of cube-like cage with six hexagonal and heptagonal holes and eight closepacking B 6 triangles. Our computational data show that Li m &B 40 (1-3) complexes bound up to three H 2 molecules per Li site with an adsorption energy (AE) of 0.11-0.25 eV/H 2 , ideal for reversible hydrogen storage and release. The bonding features charge transfer from Li to B 40 . The first 18 H 2 in Li 6 &B 40 (3) possess an AE of 0.11-0.18 eV, corresponding to a gravimetric density of 7.1 wt%. The eight triangular B 6 corners are shown as well to be good sites for Li-decoration and H 2 adsorption. In a desirable case of Li 14 &B 40 -42 H 2 (8), a total of 42 H 2 molecules are adsorbed with an AE of 0.32 eV/H 2 for the first 14 H 2 and 0.12 eV/H 2 for the third 14 H 2 . A maximum gravimetric density of 13.8 wt% is achieved in 8. The Li-B 40 -nH 2 system differs markedly from the previous Li-C 60 -nH 2 and Ti-B 40 -nH 2 complexes.Due to its merits of cleanness, renewability, abundance in nature, and high energy density per unit mass, hydrogen has been recognized as an appealing energy carrier for the future world. It has the potential to reduce our dependence on fossil fuels, which are limited in resource and harmful to the environment 1-4 . One bottleneck of using hydrogen for vehicular applications is the lack of safe and efficient hydrogen storage materials 5-7 that store molecular H 2 reversibly with high gravimetric density and fast kinetics for adsorption, as well as desorption, under the conditions of moderate temperature and pressure 8,9 . An ideal H 2 storage system would be one that binds hydrogen in molecular form and with an adsorption energy (AE) in the regime of 0.1-0.5 eV per H 2 , that is, intermediate between the physisorbed and chemisorbed states 10,11 . Although advances have been made towards meeting the U.S. DOE's targets for hydrogen storage, an ideal system is yet to be designed and synthesized. Therefore, seeking novel hydrogen storage materials has remained an important issue.Previous experiments and theoretical calculations have shown that metal-decorated carbon fullerenes and nanotubes [12][13][14][15][16][17][18][19] , as well as their boron-, nitrogen-and beryllium-substituted nanostructures [20][21][22] , might be good candidates for the storage of H 2 molecules. For instance, Zhang and co-workers showed that the reversible hydrogen storage of transition-metal-coated C 60 and C 48 B 12 may be as high as 9 wt% 21 . Yildirim et al. revealed that Ti-coated single-walled carbon nanotubes can store 8 wt% of H 2 23 . To avoid the clustering problem of transition metal atoms on the surface of carbon nanostructures, Yoon and co-workers 18 found that Ca can achieve homogeneous monolayer coating, which is superior to other metal elements. They concluded that up to 8.4 wt% of hydrogen can be stored in Ca 32 C 60 with an AE of 0.2-0.4 eV/H 2 . Through first-principles computations, Sun et al. 13 predicted that Li-decorated fullerene C 60 (Li 12 C 60 ) can store up to 9 wt% of H 2 , albeit with a rather weak AE of 0.075 eV/H 2 . Furthermore, Yoshida et al. 17 measured the hydrogen absorption of Li 9 C 60 based on experiments and confirmed that up to ~2.6 wt % H 2 can be stored at 250 °C and 30 bar H 2 . For lithium-doped fullerenes (Li x -C 60 -H y ) with a Li:C 60 mole ratio of 6:1, a reversible uptake of 5 wt% H 2 at 350 °C and 105 bar H 2 and desorption onset temperature of ~270 °C was observed 15 . Subsequently, another experimental results 16 showed that up to 9.5 wt % deuterium (D 2 ) are absorbed in Li 12 C 60 under a pressure of 190 bar and a temperature below 100 °C. Boron is the lighter neighbor of carbon in the periodic table, which possesses the similar merit as carbon in terms of light weight and potential applications for hydrogen storage. For this purpose, its chemical hydrides were studied, as were relevant model nanostructures, such as boron monolayer sheets, fullerenes, and nanotubes . In particular, following the proposal of the celebrated I h B 80 buckyball 30 , which is built upon the C 60 motif by capping all 20 surface hexagons, a number of papers were devoted to hydrogen storage using B 80 coated with metals (M = Li, Na, K, Be, Mg, Ca, Sc, Ti, and V) 27, . However, B 80 was subsequently found to favor core-shell type structures at various theoretical levels 34,35 . It is thus not feasible to pursue any realistic technological applications of B 80 buckyball as hydrogen storage materials. Very recently, the first all-boron fullerenes or borospherenes, D 2d B 40 and D 2d B 40 − , were observed in a combined experimental and theoretical study 36 , marking the onset of the borospherene chemistry, whose future development may be envisioned to parallel that of the fullerenes. Endohedral M@B 40 (M = Ca, Sr) and exohedral M&B 40 (M = Be, Mg) metalloborospherenes were also predicted, which further support the structural, electronic, and chemical robustness of the B 40 borospherene 37 borospherenes were also studied 38,39 , which expand the borospherene family and may eventually lead to new boron-based nanomaterials. Borospherene B 40 possesses a cube-like cage structure, whose six hexagonal and heptagonal holes each occupy a face of the cube. It also has eight triangular, close-packing B 6 structural blocks, each on an apex of the cube. All B atoms are on the surface of the cage, which is an ideal, well-defined system for chemistry. B 40 differs from carbon fullerenes in terms of structure and bonding, and the pursuit of borospherene-based nanomaterials for hydrogen storage is thus intriguing from a fundamental point-of-view. Furthermore, borospherenes are lighter than carbon fullerenes, which make the former systems better candidates to reach a higher gravimetric capacity for hydrogen storage. Relevant to this topic, Dong et al. 40 predicted on the basis of density-functional theory (DFT) calculations that Ti-decorated B 40 fullerene (Ti 6 B 40 ) is capable of storing up to 34 H 2 molecules with a maximum gravimetric density of 8.7 wt% and a reversible storage capacity of 6.1 wt%. To our knowledge, the U.S. DOE has set a target of 7.5 wt% for hydrogen storage capacity for the year of 2015 41,42 . In this work, we choose to study lithium-decorated borospherene B 40 as a potential candidate for hydrogen storage via extensive DFT calculations. Since boron-based nanomaterials are also candidates for lithium storage, the current ternary B-Li-H system is quite unique 28,29,43,44 . Compared to transition metal, Li as the lightest metal definitely will facilitate the improvement of hydrogen storage capacity for the metal-decorated B 40 system. The Li m -B 40 -nH 2 system differs markedly from Li m -C 60 -nH 2 or Ti 6 -B 40 -nH 2 , which have an AE value that is either rather small (0.075 eV) 13 or too large (up to 0.82 eV) 40 . Even the recently proposed Li 8 -B 6 -nH 2 system 44 only has an AE of less than 0.1 eV. Our computational data show that Li-decorated B 40 appears to be a promising medium for hydrogen storage. The Li atoms readily attach to the top of hexagonal and heptagonal holes on B 40 , forming a series of charge-transfer complexes from C s Li&B 40 (1), C 2v Li&B 40 (2), up to D 2d Li 6 &B 40 (3). The Li m &B 40 complexes can adsorb three H 2 molecules per Li site with a moderate AE of 0.11-0.25 eV/H 2 . The Li 6 &B 40 (3) complex stores up to 34 H 2 with an average AE of 0.10 eV/H 2 . The first 18 H 2 of these possess ideal AEs, which suggest a gravimetric density of 7.1 wt%. Furthermore, the eight close-packing, triangular B 6 corner sites of B 40 are also suitable for Li-decoration and H 2 adsorption. In a desirable Li 14 &B 40 (7) complex, up to 42 H 2 molecules can be stored with AEs of 0.12-0.32 eV/H 2 , which corresponds to a gravimetric density of 13.8 wt%. ## Results and Discussion Isolated B 40 Borospherene for H 2 Adsorption. The first all-boron fullerene called as borospherene 36 , D 2d B 40 ( 1 A 1 ), possesses a very large HOMO-LUMO gap of 3.13 eV at the PBE0 level that indicates its overwhelming stability, which is comparable to that of I h C 60 ( 1 A g ) (3.02 eV) calculated at the same level. All the valence electrons in B 40 are either delocalized σ or π bonds and there is no localized 2c-2e bond, unlike the C 60 fullerene. In fact, the surface of B 40 is not perfectly smooth and exhibits unusual heptagonal faces which may play a role that release the surface strains, in contrast to C 60 fullerene whose surface makes up of pentagons and hexagons and presents the least strain. And the diameter of B 40 is 6.2 , slightly smaller than the value of C 60 (7.1 ), which makes B 40 more comfortable to accommodate a range of small molecules inside the cage. We initially studied H 2 adsorption on the isolated B 40 borospherene. The optimized structures of B 40 H 2 , H 2 @ B 40 and 2H 2 @B 40 are shown in Fig. 1. In the C 2 B 40 H 2 dihydride, the H 2 molecule tends to form two B-H covalent bonds with the tetracoordinate-B sites, which dissociate H 2 . The dissociative AE of a single H 2 is up to 1.30 eV. For H 2 storage inside the cage, only one H 2 molecule can be encapsulated into B 40 , which is marginally exothermal with an AE of 0.24 eV. Interestingly, once such an encapsulation is completed, the H 2 molecule cannot escape due to substantial energy barriers (> 3 eV). The AE of a second H 2 inside the cage is found to be endothermic by 1.32 eV, which is thus not feasible experimentally. In short, the above results show that an isolated B 40 borospherene is not a good candidate for hydrogen storage directly. The B 40 -H 2 interactions appear to be different from the case of C 60 . The latter is known to interact with H 2 via weak van der Waals forces 45 . As a comparison, our calculation results show that the dissociative AE of a single H 2 for C 60 H 2 is only ~0.18 eV at the same level. However, similar to B 40 , only one hydrogen molecule can reside inside the C 60 cage with a negative AE value of ~0.22 eV. ## Configurations of Li-Decorated B 40 . As shown in Fig. 2, we start with a single Li atom interacting with B 40 . Relative stability of Li atom bound on heptagonal and hexagonal holes were considered. The exohedral C s Li&B 40 (1), in which Li caps a heptagon, turns out to be more stable by 0.20 eV with respect to C 2v Li&B 40 (2). In the latter species, Li caps a hexagon. The BE for Li is 3.08 and 2.88 eV in 1 and 2, respectively. Thus, Li prefers to bind on top of the heptagonal hole of B 40 . The Li-B distance in 1 is 2.34 , compared to 2.33 in 2 (Table 1). Clearly, the BE of Li on the center of heptagon or hexagon in B 40 is substantially higher than those on the pentagon of C 60 (1.80 eV), in Li 2 dimer (0.95 eV), and in the Li bulk (1.63 eV) 13 . This should help suppress the potency of Li aggregation to form clusters on B 40 surface, suggesting that Li is a suitable adsorbate to decorate B 40 . As shown in Table 1, electron transfer occurs from Li to borospherene B 40 cage in 1 and 2, resulting in a positive charge of 0.87-0.88 |e| on Li as revealed in the Bader charge analysis. The ionized Li atom hints a possibility for H 2 adsorption via the polarization mechanism 18 . To increase the coverage of Li on B 40 , we place one Li atom on top of every hexagon and heptagon hole and reach exohedral D 2d Li 6 &B 40 (3) (Fig. 2). In complex 3, six Li atoms remain isolated on the surface holes, resulting in a highly symmetric D 2d geometry. The average BE of Li in 3 is 3.07 eV/Li, which is comparable to that in Li&B 40 (1) (3.08 eV) and is slightly larger than that in Li&B 40 (2) (2.88 eV). There appears to be a collective effect for Li adsorption because six Li atoms in 3 have a higher total BE (18.48 eV) than six individual Li atoms in 1 and 2 combined (18.08 eV). This fact suggests that Li 6 &B 40 (3) is a favorable configuration for Li-decoration. Remarkably but not surprisingly, our computational data indicate that 3 is at least 6.29 eV more stable than B 40 attached by a compact Li 6 cluster (Fig. S1). Therefore, surface aggregation of Li for island clusters is unlikely in the When one H 2 molecule is introduced to 1, due to the polarization interaction between the charged Li atom and the H 2 molecule, the Li-B bond distance is slightly enlarged (by 0.01 ) to 2.35 . The H-H distance is found to be 0.76 . The equilibrium Li-H distance is 1.97 , which is comparable to the value of 2.04 in the case of a free Li + ion interacting with H 2 46 . The AE of the first H 2 to 1 is 0.25 eV, which is in quantitative agreement with that in Li + H 2 (0.25 eV) 46 . With more H 2 molecules being attached to 1, the average AE of H 2 , consecutive AE of H 2 , the Li-B distance, and the distances between H 2 and Li change accordingly. As shown in Table S1 and Fig. 5a, a single Li atom in 1, coated on a heptagonal hole, can adsorb up to six H 2 molecules with an average AE of 0.11 eV/H 2 . From one to six H 2 , the average AE decreases from 0.25 to 0.11 eV/H 2 , whereas the consecutive AE decrease from 0.25 to 0.05 eV/H 2 . This effect may be partially due to the steric repulsion 47 when the number of H 2 molecules increases. In line with this trend, the Li-B distances are elongated gradually from 2.35 to 2.43 . However, the H-H bond distance is nearly constant in the range of 0.75-0.76 , which is the value of free H 2 molecule, consistent with the nature of molecular adsorption. The Li-H distances span a rather wide range from 1.97 to 2.91 . Notably, there is an abrupt increase in the Li-H distances from 1-3 H 2 to 1-4 H 2 , so that the first three H 2 are closer to the Li site than the next three. In other words, the adsorption of the first three H 2 molecules forms an inner core with Li, upon which the additional H 2 molecules adsorb loosely. The data of consecutive AE confirm this to be indeed the case: The first three H 2 possess an AE of 0.25-0.11 eV, in contrast to 0.04-0.05 eV for the next three (Table S1). In fact, the structure of 1-4 H 2 can be constructed on the basis of 1-3 H 2 by adding one H 2 on the top of Li. However, when the fifth and sixth H 2 are put on successively in 1-5 H 2 and 1-6 H 2 , they flee away after structural optimization as shown in Fig. 5a. Therefore, the Li site in 1 may adsorb three H 2 molecules comfortably, whereas additional H 2 are only physisorbed. Basically, the adsorption of H 2 on 2 is rather similar to that on 1. Up to five H 2 molecules may be adsorbed on the Li site in 2 (Fig. 5b). Again, the first three H 2 are located closely to Li, with the fourth H 2 being situated symmetrically on top of Li at a substantially larger distance. For the 2-5 H 2 case, there is a structural rearrangement for the H 2 molecules, suggesting that this Li site can potentially adsorb up to four H 2 molecules at a reasonable strength. Nonetheless, the fifth H 2 only interact with Li loosely. On the basis of the consecutive AEs (0.22-0.17 eV for the first three H 2 ; Table S1), we conclude that the Li site in 2 is capable of adsorbing at least three H 2 molecules with further possibility for a fourth, whereas additional H 2 molecules should be considered physisorbed. Based on the above results of H 2 adsorption on single Li-decorated B 40 , we constructed and optimized the H 2 adsorption configurations on the Li 6 &B 40 (3) complex, which aims at exploring the hydrogen storage capacity. The starting configurations were constructed by attaching the corresponding H 2 molecules around Li atoms above the 4 heptagonal and 2 hexagonal holes on the B 40 cage. Successively, 6 H 2 , 12 H 2 , 18 H 2 24 H 2 , and up to 34 H 2 are adsorbed on Li 6 &B 40 (3), whose optimized structures are shown in Fig. S2 and 6. The former four cases correspond to the adsorption of one to four H 2 on each Li site. For the 34 H 2 case, that is, Li 6 &B 40 -34 H 2 (4), 6 H 2 are adsorbed on each heptagonal Li site and 5 H 2 are on each hexagonal Li site, as depicted in Fig. 6. The total interaction energy of 34 H 2 in 4 is 3.43 eV, yielding an average AE of 0.10 eV/H 2 . The calculated consecutive AEs are collected in Table 2, which reflects the adsorption nature more faithfully. Similar to 1 and 2, the first three 16 B-H bonds for the tetracoordinate B sites, which can also be decorated with six Li atoms, resulting in a D 2d Li 6 &B 40 H 16 (5) complex as depicted in Fig. 7. The Li-B distance in 5 remains to be 2.33 , which is very close to that in 3. In complex 5, each Li atom carries a charge of 0.88 |e|. Interestingly, the B-H bonds markedly alter the Li-decoration properties in 5 and the average BE of Li atom now increases to 4.17 eV per Li, compared to 3.07 eV in Li 6 &B 40 (3). Li 6 &B 40 H 16 (5) can also adsorb from 6 H 2 , 12 H 2 , 18 H 2 , 24 H 2 , and up to 34 H 2 molecules, resulting in a series of 5-nH 2 complexes (Fig. S3). The optimized structure for Li 6 &B 40 H 16 -34 H 2 ( 6) is shown in Fig. 7. Note that hydrogen remains in the molecular state with a uniform H-H distance of 0.75 in all 5-nH 2 species. For the first 6 H 2 molecules in 5-6 H 2 , the average AE amounts to 0.22 eV/H 2 . The average Li-B and Li-H distances, 2.33 and 1.97 , respectively, are almost the same as those in Li 6 &B 40 -6 H 2 (that is, 3-6 H 2 ). With further H 2 adsorption, the average AEs for the first 18 H 2 in 5-nH 2 decrease slightly down to 0.17 eV/H 2 , which are in the ideal thermodynamic range for reversible hydrogen storage 10,11 . The Li 6 &B 40 H 16 (5) complex thus behaves rather similar to Li 6 &B 40 (3) in terms of hydrogen storage properties, except for the B-H passivation in 5. The 18 "core" H 2 in Li 6 &B 40 H 16 -34 H 2 (6) represents a gravimetric density of 6.5 wt%, where an additional 8.6 wt% of dissociated H atoms and loosely physisorbed 16 H 2 are not counted. On the Possibility of Doubling the H 2 Adsorption Sites: Li-Decorated Triangular B 6 Corners. To further improve the hydrogen storage capacity of Li-decorated B 40 , we also attempted to place Li atoms on top of the close-packing, triangular B 6 corner sites of the cube-like B 40 cage. As a test case, adsorption of a single H 2 molecule on a corner Li site is optimized (Fig. S4). The BE of Li is 1.87 eV, which is lower than those in Li&B 40 (1) and Li&B 40 (2), but the value still represents a reasonable strength. In fact, it is comparable to the corresponding value for C 60 (1.80 eV) 13 . Moreover, the AE for the first H 2 amounts to 0.28 eV, which is comparable to and even slightly (5) to fourteen in Li 14 &B 40 (7), owing to the eight triangular B 6 corners (versus six hexagonal/heptagonal holes). The optimized structure of Li 14 &B 40 (7) is shown in Fig. 8. Here, upon Li decoration, the boron structure distorts considerably from the free-standing B 40 borospherene, but the cage motif maintains. The average BE is 2.57 eV/Li. Following the strategy for Li 6 &B 40 (3) and Li 6 &B 40 H 16 (5), we build a series of model complexes: 7-14 H 2 , 7-28 H 2 , and Li 14 &B 40 -42 H 2 (8), whose optimized structures are depicted in Fig. 8. The calculated average AE for the first 14 H 2 in 7-14 H 2 is 0.32 eV/H 2 , for the second 14 H 2 in 7-28 H 2 is 0.22 eV/H 2 , and for the third 14 H 2 in Li 14 &B 40 -42 H 2 (8) is 0.12 eV/H 2 , suggesting that all these H 2 molecules are thermodynamically favorable for a hydrogen storage material 10,11 . For the extreme case of 8, a maximum gravimetric density of 13.8 wt% is obtained. We do not exclude the possibility of further H 2 adsorption onto the 8 complex, albeit those additional H 2 are anticipated to interact rather loosely with the Li sites. ## Concluding Remarks In conclusion, we have carried out a comprehensive density-functional study on the lithium-decoration of B 40 borospherene and the potential utilization of Li-B 40 complexes as a novel nanomaterial for hydrogen storage. We showed that all six heptagonal and hexagonal holes on B 40 surface can be decorated with Li atoms and each Li site is capable of adsorbing up to six or five H 2 molecules. This results in an ultimate Li 6 &B 40 -34 H 2 complex, in which 18 H 2 are bound to Li sites with ideal adsorption energies of 0.11-0.18 eV per H 2 , corresponding to a gravimetric density of 7.1 wt%. The additional 16 H 2 are physisorbed in nature. We further showed that the eight close-packing, triangular B 6 corner sites on the B 40 cage are also readily decorated with Li, which more than double the number of sites for hydrogen storage. The corresponding Li 14 &B 40 -42 H 2 complex can store all 42 H 2 molecules at adsorption energies of 0.12-0.32 eV per H 2 , suggesting a maximum gravimetric density of 13.8 wt%. The Li-B 40 -H 2 complexes as a hydrogen storage material differ markedly from the prior Li-C 60 -H 2 and Ti-B 40 -H 2 systems. The Li-C 60 -H 2 complex 13 adsorbs H 2 rather loosely and is thus not efficient for hydrogen storage, whereas the Ti-B 40 -H 2 complex 40 bounds H 2 too strongly, for which a substantial portion of H 2 stored are not reversible for release. In fact, preliminary attempts also suggest that the structural integrity of B 40 unit is maintained when they are allowed to interact with each other. Considering the presence of chemical bondings between them, we forecast it is possible to construct boron-based nanomaterials for hydrogen storage using lithium-decorated B 40 unit as a building block or connecting the exohedral metalloborospherene with organic linkers. And the hydrogen storage capacity of the boron-based nanomaterials could be better than previously reported carbon-based counterparts. ## Methods All calculations were based on DFT, using a plane-wave basis set with the Projector Augmented Wave (PAW) 48,49 pseudopotential method as implemented in the Vienna ab initio Simulation Package (VASP) 50,51 . Generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) 52 functional was adopted to treat the electron exchange correlation. The GGA-PBE method has been previously utilized to treat Li-decorated fullerenes and heterofullerenes for hydrogen storage 19,53 which is thus a suitable choice for our current system. The dispersion corrected DFT (DFT-D) scheme was used to describe the van der Waals (vdW) interaction. The supercell approach was used, where the B 40 -based systems were placed at the center of a 25 × 25 × 25 3 vacuum space. And only the Γ point was used to sample the Brillouin zone. The energy cutoff for the plane-wave basis set was set to 500 eV. All structures were fully relaxed until the force acting on each atom was less than 10 −2 eV/ and a tolerance in total energy was at least 10 of Li-decorated B 40 , E H2 is the total energy of isolated H 2 molecule, and n stands for the number of adsorbed H 2 molecules. We note that for comparison with D 2d B 40 in our previous work (ref. 36), the HOMO-LUMO energy gaps of 1, 2, and 3 were calculated using the Gaussian 09 package 57 , which is usually used for calculations on the isolated molecules. And the corresponding structures were optimized at the PBE0 levels with the 6-311 + G* basis set 58,59 , which has been benchmarked in prior works as a reliable method for boron clusters.

chemsum

{"title": "Lithium-Decorated Borospherene B40: A Promising Hydrogen Storage Medium", "journal": "Scientific Reports - Nature"}

hierarchical_two-dimensional_molecular_assembly_through_dynamic_combination_of_conformational_states

## Abstract: Self-sorting of multiple building blocks for correctly positioning molecules through orthogonal recognition is a promising strategy for construction of a hierarchical self-assembled molecular network (SAMN) on a surface. Herein we report that a trigonal molecule, dehydrobenzo [12]annulene (DBA) derivative having three tetradecyloxy chains and three hydroxy groups in an alternating manner, forms hierarchical triangular clusters of different sizes ranging from 2.4 to 16.4 nm, consisting of 3 to 78 molecules, respectively, at the liquid/graphite interface. The key is the dynamic combination of three different conformational states, which is solvent and concentration dependent. The present knowledge extends design strategies for production of sophisticated hierarchical SAMNs using a single component at the liquid/solid interface. ## Introduction Self-assembled molecular networks (SAMNs) 1 spontaneously formed by organic molecules through self-assembly on solid surfaces are a subject of keen interest because of their prospect of application in the felds of nanoscience and nanotechnology. Sophisticated structural control of SAMNs based on new design principles may lead to complexity as seen in biological systems, thereby enriching their potential for applications. One of the major challenges in this feld of research is the construction of hierarchical structures that extend several tens of nanometers or even sub-micrometers in size. However, the structural periodicities in the known SAMNs using small molecular building blocks are typically limited to several nanometers. 9,10 SAMNs with a periodicity of more than 100 nm could be fabricated only using large DNA molecules. 11 Therefore, there is a need for formulating clear design principles for the formation of hierarchical SAMNs with a long-range periodicity using small molecular building blocks. 12,13 The most common strategy to construct large hierarchical structures is the use of building blocks of C 3 or D 3h symmetry capable of assembling through strong non-covalent interactions such as hydrogen bonding and metal coordination, and they are typically prepared under ultrahigh vacuum (UHV) conditions. Under these conditions, the size of unit cells can be modifed by surface coverage. By fully exploiting this approach, a hierarchical structure with a large unit cell of 45 nm was reported recently. 18 Another strategy for the construction of hierarchical SAMNs with long-range periodicity is to optimize multiple intermolecular interactions using elaborated building blocks or multi-component building block(s). Such experiments are typically undertaken at the liquid/solid interfaces or occasionally in air. 25 Compared to UHV conditions, the unit cell size and the number of building blocks for constructing such hierarchical structures are limited, because the presence of a supernatant solvent renders the self-assembly and system complicated. Solvation and surface wetting compete with intermolecular and molecule-substrate interactions, and solvent molecules are often co-adsorbed in SAMNs. In hierarchical superstructures, the constituent clusters consist of three different parts, i.e., vertices, edges and internal core, each bearing different coordination numbers (Fig. 1). In most cases, a single building block forms these different parts by adapting the modes of intermolecular interactions as enforced by an external factor, i.e., surface density. Such selfassembly behavior may be regarded as a kind of self-sorting, even though this term is defned as "mutual recognition of complementary components in artifcial self-assembly". On a surface, there exists only one example of such self-sorting through dynamic combination of conformational states of a single molecular building block, yet the structural controlling factors were not discussed in detail. 32 Here, we extend the above concept to produce hierarchical SAMNs using a C 3h -symmetric building block with orthogonal coordination sites for van der Waals interaction and hydrogen bonding. By changing its conformation on the surface depending on the solvent polarity, the building block selfassembles to form hierarchical structures made of triangular clusters with size ranging from 2.4 to 16.4 nm, and consisting of 3 to 78 molecules, respectively. The molecules form the vertices, edges and core positions of the clusters as well, thanks to balanced orthogonal intermolecular interactions. Each triangular cluster comprises the feature of Pascal's triangle which is a triangular array of binomial coefficients in mathematics. Scanning tunneling microscopy (STM) is our method of choice, since it provides high quality structural information in real space of SAMNs on atomically flat conductive surfaces such as graphite, even at the liquid/solid interface. 33,34 Over the past decade, we studied the self-assembly of dehydrobenzo annulene derivatives DBA-OCns having six long alkoxy groups at the liquid/graphite interfaces (Fig. 2a). 35 DBA-OCns form low density porous structures and high density nonporous structures. The length of the alkoxy groups, solute concentration and temperature determine the relative abundance of the polymorphs. 36 The main driving forces for the formation of these SAMNs are the intermolecular interaction through van der Waals interactions between the alkoxy groups and molecule-substrate interactions between the alkoxy groups and graphite. Of further relevance is the fact that in the nonporous structures, four alkoxy groups are physisorbed on the surface while the remaining two are probably solvated, whereas in the porous SAMNs all six alkoxy groups are adsorbed. To further extend the self-assembling ability of DBA derivatives for the formation of various 2D patterns, we reduced the symmetry of the building blocks from D 3h to C 3h , thereby rendering them more flexible. Indeed, we found that DBA-OC14-OC1 having three long alkoxy groups and three methoxy groups in an alternating manner exhibits a rich structural polymorphism due to the increased mobility of the alkoxy groups and the variable number of physisorbed alkoxy groups (hereby denoted as m: m ¼ 3-1). This structural variety depended on the type of solvent and solute concentration. 37 This led us to hypothesize that by dynamic self-sorting, based on the ability of the molecule to adsorb with a different number of alkoxy chains (m) in contact with the surface, in combination with extending the type and number of intermolecular interactions, it may become possible to assemble a C 3h -symmetric DBA molecule in a hierarchical manner. Therefore, we designed DBA-OC14-OH bearing three OC14 and three hydroxy groups in alternating positions on the DBA core (Fig. 2a and b). Indeed, we found that (i) this DBA derivative forms hierarchical patterns, consisting of triangular clusters ranging from 2.4 (3 molecules) to 16.4 nm (78 molecules) in size, (ii) each cluster is formed by hydrogen bonding through self-sorting of DBA conformers, (iii) ## Results and discussion The synthesis of DBA-OC14-OH was reported previously. 38 As solvents, we chose 1,2,4-trichlorobenzene (TCB) and 1-hexanoic acid (HA) representing nonpolar and polar solvents, 2,39 respectively, to modify the solvation of DBA-OC14-OH, and thereby also the adsorption probability on graphite. We also expected hydrogen bond formation between the carboxy group of HA with the hydroxy groups of the DBA, 40 and potential coadsorption of both solvents. 41 Firstly, the self-assembly behavior of DBA-OC14-OH in a pure solvent was examined. To investigate the concentration dependent structural polymorphism, DBA solutions were prepared ranging from 3.0 10 6 to 1.0 10 4 M. The solution (40 mL) was poured into a liquid cell placed on a freshly cleaved surface of highly oriented pyrolytic graphite (HOPG). Reaching equilibrium was facilitated by annealing the liquid/graphite interface at 80 C for 3 h. 38,42 After the annealing treatment, the cell was allowed to cool to room temperature, prior to STM imaging at the liquid/ graphite interface. At the TCB/graphite interface, DBA-OC14-OH exclusively forms a hexagonal porous structure with four bright triangles at each vertex of the hexagonal pore at all concentrations examined (Fig. 3a and S2 †). Since p-conjugated cores are typically resolved as bright features in the STM images because of their relatively higher tunneling efficiency, 43 the bright triangle is assigned to a tetramer of DBA-OC14-OH with three physisorbed alkoxy groups (m ¼ 3) forming the triangular shape by both van der Waals and hydrogen bonding interactions. The neighboring tetramers are connected by seven alkyl chains (Fig. S3 †). 44 A molecular model optimized by molecular mechanics (MM) simulation (COMPASS force feld) is shown in Fig. 3b. The mean O/H distance of the hydroxy groups between the central and surrounding DBA-OC14-OH molecules is 2.2 AE 0.2 , supporting the hypothesis that the molecules are clustered through hydrogen bonding interactions (Fig. 3b). 45,46 The presence of enantiomorphous domains was identifed by the orientations of the alkyl chains located at the rims of the pore (Fig. S4-S9 †). The details of the chirality aspects both at single and supramolecular levels are described in the ESI. † In contrast to the results in TCB, the self-assembly of DBA-OC14-OH at the HA/graphite interface exhibits a sharp concentration dependence. At 6.0 10 6 M, a hexagonal assembly of triangular clusters consisting of three molecules of DBA (n ¼ 2: n refers to the number of DBA-OC14-OH molecules forming each edge of the triangular cluster) with two physisorbed alkyl groups (m ¼ 2) was exclusively observed (Fig. 3c and S10 †). The total number of DBA-OC14-OH molecules (N) per cluster, represented by N ¼ n(n + 1)/2, is three in this case. Based on the unit cell parameters and the STM image, it is safe to conclude that the triangular cluster is formed by hydrogen bonding mediated by one HA molecule per adjacent DBA pair. The triangular clusters are bridged by four interdigitated alkyl chains of DBA-OC14-OH. Note that two HA molecules stick in between the interdigitated alkoxy groups (Fig. 3d). In the model, the mean O/H atomic distances between the hydroxy group of DBA-OC14-OH and the carboxy group of HA measure 1.70 AE 0.02 and 1.75 AE 0.01 , respectively, suggesting the presence of hydrogen bonding interactions (Fig. S11 †). Similar to the hexagonal porous structure, there are both antipodal domains differentiated by the alkyl chain orientations located at the rims of the hexagonal pore (Fig S12 and S13 †). Again, the details are discussed in the ESI. † By increasing the concentration to 1.0 10 5 M, in addition to small domains of the triangular cluster (n ¼ 2), irregular areas consisting of triangular clusters of various sizes (n ¼ 2 up to 6) and a densely packed phase cover the surface (Fig. S14 †). Further increase of the concentration to 1.0 10 4 M led to the formation of a dense structure consisting of DBA-OC14-OH without physisorbed alkoxy groups (m ¼ 0); all alkoxy groups orient to the solution phase (Fig. 3e, S15 and S16 †). There exist two chiral domains of the dense structure, which could be differentiated by the molecular orientations with respect to the substrate axes. In the molecular model, all intermolecular O/H distances between DBA-OC14-OH are 2.0 , indicating hydrogen bonding (Fig. 3f). Thus, at the HA/graphite interface, though several hierarchical structures of n ¼ 2 to 6 emerged, concentration control over the formation of clusters of specifc size was not achieved. Next, we examined the effect of solvent polarity by changing the ratio of TCB and HA in the solvent mixture on the size control of the hierarchical triangular clusters. 47 To compare the affinity of the solvents to DBA-OC14-OH, its solvation energies in TCB and HA are estimated by molecular dynamics (MD) simulations to be 54.22 and 0.26 kcal mol 1 , respectively (see the ESI †). The larger negative value in TCB indicates smaller adsorption probability than in HA, which is consistent with the favorable formation of the low density hexagonal porous network by the DBA with three physisorbed alkoxy groups (m ¼ 3). In contrast, the less favorable solvation in HA is consistent with the formation of the densely packed structure by the DBA with none of the alkoxy groups (m ¼ 0) physisorbed at high concentration. By varying the ratio of these two extreme solvents, we hypothesized that the distribution of the DBA molecules with a different number of physisorbed alkoxy groups m may be controlled to form triangular clusters of a specifc size. Based on this hypothesis, solutions of mixtures of TCB and HA at different molar fractions (X HA ; 0.020-0.50 and X TCB ; 0.98-0.50) were prepared keeping the total DBA concentration constant (1.0 10 4 M), and the resulting SAMNs were analyzed. The overall results are summarized in Table 1. At X HA ¼ 0.020, both a hexagonal porous structure and a triangular cluster co-exist (n ¼ 2, Fig. S17a †). At the mixing ratio X HA ranging from 0.049 to 0.066, DBA-OC14-OH mainly forms the triangular cluster of n ¼ 2, similar to the conditions in the pure HA solution of low concentration (Fig. S17b †). Upon increasing X HA to 0.070-0.082, the n ¼ 4 cluster appears and its surface coverage increases up to about 50% with increasing X HA (Fig. 4a, S17c, d and S19 †), reaching a maximum at X HA ¼ 0.082, forming large domains over 80 80 nm 2 (Fig. S18d †). The cluster of n ¼ 4 consists of DBAs in three different conformational states (three of m ¼ 2, six of m ¼ 1, and one of m ¼ 0) and 9 molecules of HA attached to the edges linked by hydrogen bonding interactions (Fig. 4b and S20 †). The surrounding six DBA molecules are bound to the central core DBA of m ¼ 0 via hydrogen bonds. The adjacent clusters are connected by van der Waals interactions between interdigitated alkoxy groups of DBA-OC14-OH with HA molecules stuck in between. We were unable to fnd clusters of n ¼ 3. a The surface coverages and standard deviations were determined from more than 20 images in three independent experimental sessions. b Collapsed triangular clusters. c The relative ratios of different clusters at an X HA of 0.20 are listed in Table 2. d The error was not determined due to its small area ratio. Hierarchical structures consisting of larger triangular clusters emerge upon further increasing X HA to 0.090 (Fig. S21 †). At an X HA of 0.20, DBA-OC14-OH produces a mixed phase of hierarchical clusters containing the n ¼ 12 cluster as the major component (ca. 27%) together with other clusters of similar sizes (n ¼ 10, 11, 13 and 14, Fig. 5 and Table 2). Note that the length of the edge of the cluster of n ¼ 12 reaches 16.4 nm and the cluster is composed of 78 DBA molecules (three DBAs of m ¼ 2, 30 DBAs of m ¼ 1, and 45 DBAs of m ¼ 0). In the hexagonal assembly of the large clusters, even if the sizes of the triangular clusters are not all uniform, the hexagonal packing is sustained by small gaps (Dn) in between adjacent clusters (typically Dn is smaller than or equal to 2) by distorting the central hexagonal pore. When the size difference (Dn) becomes larger than 2, the triangular clusters are often chipped out from the vertex to maintain the overall hexagonal packing (triangles with red numbers in Fig. 5). These domains are stable against STM tip scanning. Moreover, a longer annealing period from 3 to 6 h led to no notable structural change and the hierarchical structure remains even after 1 day at room temperature, indicating that the hierarchical structure is thermodynamically stable. The formation of the triangular clusters rather than other clusters with different shapes, i.e. hexagonal shape clusters would be related to the favored intermolecular interactions (Fig. S25 †). 48 We emphasize that each n ¼ 12 cluster is formed by 78 molecules through a dynamic combination of its three different conformational states, each being a part of vertices, edges and core, even though the control is not perfect probably due to thermal fluctuations. This process can be regarded as a dynamic version at the solid/liquid interface of integrative self-sorting, which is a new aspect of self-assembly mainly observed in solution and in crystals. Jester and Höger reported a related preliminary case in which a six-alkyl-bearing building block capable of adopting several conformations with a different coordination number at the solvent/graphite interface is a part of vertices, edges and core, forming a hierarchical structure of variable sizes. 32 In our case, the hierarchical structures are formed through orthogonal van der Waals and hydrogen bond interactions not only between the DBA molecules but also between the DBA and solvent molecules. Moreover, the cluster size, namely the number of different dynamic conformers can be modulated to some extent by changing the degree of solvation. The present work improves the design strategy for the construction of hierarchical SAMNs. Further increase of X HA to 0.50 leads to the dense structure exclusively covering the whole surface (Fig. S23 †). At all mixing ratios, domains of n with small odd numbers of 3 or 5 were never or scarcely observed. Though this may be related to the epitaxy with the substrate lattice or formation mechanism of the clusters (vide infra), it is not understood yet. In contrast, both odd and even clusters coexist for the larger clusters (n ¼ 10-14), which is attributed to the small differences in size and molecular density among these large clusters. To shed light on the formation mechanism of the clusters, monolayers formed without annealing treatments were observed. At X HA ¼ 0.076, the surface is covered with the hexagonal porous structure and the smallest triangular cluster (n ¼ 2, Fig. S24a †), implying that 2 or larger) obtained in three independent experimental sessions were used. Some imperfect triangular clusters in which the number of missing DBA molecules is less than 10% are included in the statistics. b The error was not determined due to its small occurrence. the triangular cluster of n ¼ 4 is formed by the lateral replacement of the smallest triangular clusters through the annealing treatment. This may also account for the absence of the triangular cluster of n ¼ 3. At X HA ¼ 0.20, the DBA molecules form a disordered (non-uniform) structure consisting of incomplete large triangular clusters and smaller clusters of n ¼ 2 and 4 (Fig. S24b †), implying that the larger triangular clusters are grown from the smaller clusters by the ripening process. ## Conclusions In conclusion, we demonstrated the formation of hierarchical hexagonal assemblies of DBA-OC14-OH consisting of triangular clusters of different sizes ranging from 2.4 to 16.4 nm at the solution/graphite interface. By modulating the polarity of the solvent by changing the ratio of nonpolar solvent (TCB) and polar solvent (HA), the size of the clusters was controlled to some extent. The n ¼ 2 cluster containing two DBA molecules on the triangular edge was exclusively formed and the clusters of n ¼ 4 and n ¼ 12 were produced as the major components among clusters of similar sizes. Each cluster consisted of a discrete number of the DBA molecules with a different number of physisorbed alkyl groups m ¼ 2-0, and its formation involved hydrogen bonding and co-adsorption of solvent molecules. A key element for the hierarchical pattern formation is the dynamic combination of three different conformational states of the building block. This result conceptually expands the scope of molecular self-assembly and may be useful for constructing self-assembled patterns of enhanced complexity. ## STM experiments All the experiments were performed at 20-26 C using a Nanoscope IIID or V (Bruker AXS) with an external pulse/function generator (Agilent 33220A or TEXIO FGX-295) with a negative sample bias. STM tips were mechanically cut from Pt/Ir wire (80%/ 20%, diameter 0.25 mm). All STM images were taken in a quasiconstant current mode. For preparation of the sample solution, commercially available 1,2,4-trichlorobenzene (TCB, purchased from Nacalai Tesque) and 1-hexanoic acid (HA, purchased from Wako) were used as the solvent after distillation. A drop of this solution (40 mL) was poured into a homemade liquid cell placed on a freshly cleaved basal plane of a 1 cm 2 piece of highly oriented pyrolytic graphite (HOPG, grade ZYB, Momentive Performance Material Quartz Inc., Strongsville, OH). The liquid cell was employed to minimize the effect of solvent evaporation by using a large amount of the sample solution (40 mL). Moreover, this liquid cell was covered with a stainless lid in an oven. The proportion of the solvent loss was estimated by weighing the liquid cell system to be 7% after annealing at 80 C for 3 hours. After annealing treatment, the substrate was naturally cooled to room temperature. All STM images were taken within 3 hours after annealing treatment to minimize the concentration changes due to solvent evaporation during the observation. By changing the bias voltage applied to the substrate, the SAMNs of DBA-OC14-OH and the graphite substrate surface could be observed at low and high voltages, respectively. The distortion of the image due to the thermal drift effects was corrected by using the STM image of the underlying graphite surface. Such image correction was performed using SPIP software (Scanning Probe Image Processor, SPIP, version 4.0.6 or 6.0.13, Image Metrology A/S, Hørsholm, Denmark). In the STM images, the white arrows and the line indicate the directions of the main symmetric axes of graphite and the scale bar, respectively. For each concentration, more than 10 large area STM images (image sizes: 50 nm 50 nm or larger) were acquired per session. Three independent sessions were performed at each concentration and mixing ratio to confrm reproducibility. The unit cell and its parameters were determined using over 50 experimental data points from at least fve calibrated STM images (image sizes: 30 30 nm 2 or smaller). The surface coverages of SAMNs in the weighted average values with the standard deviations were determined from more than 20 images in three independent experimental sessions. The average height profles (11 lines, width of each line is one pixel) of the alkyl chains and DBA cores were measured from more than three different calibrated images (typically 30 30 nm 2 , 512 pixels) by the use of SPIP software. In each cross section of the hexagonal porous structure, two structurally inequivalent DBA cores in the tetramer were included because the apparent height of the DBA core is slightly different depending on its position. The apparent heights of the central DBA core parts and the surrounding DBA core parts were averaged. The triangular cluster sizes and its distribution were statistically analyzed at a HA molar fraction (X HA ) of 0.20. The size of triangular clusters is classifed by the number of DBA-OC14-OH molecules (n) at each triangular side. To determine size distribution statistically, 22 large area images (80 80 nm 2 or larger) were used. There were also imperfect triangular clusters as defects. Some imperfect triangular clusters in which the number of missing DBA molecules was less than 10% were included in the statistics. In the case where the side lengths of the triangular clusters were different, the number of DBA-OC14-OH molecules at the longest side was used as n. ## Molecular mechanics simulation The initial geometry of DBA-OC14-OH was built from the respective molecular model optimized by the semiempirical PM3 method. 49 Then the orientation of the alkyl chains relative to the p system was adjusted based on that observed in the STM images. All MM/MD simulations were performed with Materials Studio 2017 R2 using the Forcite module with the COMPASS force feld. The molecules were placed 0.350 nm above the frst layer of a two-layer sheet of graphene (interlayer distance is 0.355 nm) which represents graphite. Experimentally derived unit cell parameters are used as periodic boundary conditions (PBCs). This double layer graphene flake was frozen during the simulations, and a cutoff of 2.0 nm was applied for the van der Waals interactions (Lennard-Jones type). ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Hierarchical two-dimensional molecular assembly through dynamic combination of conformational states at the liquid/solid interface", "journal": "Royal Society of Chemistry (RSC)"}

copper_carbenes_alkylate_guanine_chemoselectively_through_a_substrate_directed_reaction

## Abstract: Cu(I) carbenes derived from a-diazocarbonyl compounds lead to selective alkylation of the O 6 position in guanine (O 6 -G) in mono-and oligonucleotides. Only purine-type lactam oxygens are targetedother types of amides or lactams are poorly reactive under conditions that give smooth alkylation of guanine.Mechanistic studies point to N7G as a directing group that controls selectivity. Given the importance of O 6 -G adducts in biology and biotechnology we expect that Cu(I)-catalyzed O 6 -G alkylation will be a broadly used synthetic tool. While the propensity for transition metals to increase redox damage is well-appreciated, our results suggest that transition metals might also increase the vulnerability of nucleic acids to alkylation damage. ## Introduction Catalytic reactions with metal carbenes are broadly applied in synthetic chemistry. 1,2 Although metal carbenes are reactive intermediates, their chemoselectivity can be steered through the right combination of metal, ligand, and carbene precursor. 3 Even in water metal carbenes can still perform an impressive array of transformations. In our studies on the post-synthetic tailoring of large nucleic acids through metal catalysis we have found that donor-acceptor substituted carbenes 8 derived from Rh(II) and Cu(I) give primarily N-H insertion of exocyclic amine groups in unpaired regions of DNAs and RNAs. Here we show that unstabilized Cu(I) carbenes derived from a-diazo acetate and amide derivatives have a completely different chemoselectivity: they quickly and cleanly alkylate the O 6 position in guanine and inosine (see Fig. 1B for a representative reaction with GMP and ethyl a-diazoacetate (EDA)). Cu(I) is known to bind amides (especially amidates), 12 but there is no precedence which would have predicted such a pronounced selectivity for the lactam-like oxygen of guanine. While most instances of copper 13 direction by heterocycles in synthetic chemistry and biology 17,18 employ histidine or pyridine (see Fig. 1A), our results show that the purine ring system of guanine is also a powerful directing group in copper catalysed alkylation (see Fig. 1B) reactions. DNA is constantly exposed to electrophiles, both natural and man-made. 19 In response, all organisms have evolved DNA surveillance and repair pathways to maintain genomic integrity. 20 Alkylation at the O 6 of guanine 21 is particularly dangerous because polymerases tend to misincorporate thymidines in response to O 6 -(alkyl)-G during replication, leading to potentially carcinogenic G-to-A transition mutations. The colon cells of people on high red meat diets have increased carboxymethyl damage at O 6 -G. 26,27 Epidemiological studies have shown that the high-red meat diet demographic in Western cultures suffer as well from a higher prevalence of gastrointestinal cancers. 28,29 Whether there is a causative connection between these two occurrences is currently unclear. Studying the biology of O 6 -(carboxymethyl)-G damage (O 6 -cmG), however, is hampered by synthetic challenges. Multi-step procedures, protecting group manipulations, 30 and the need for strong electrophilic phosphorous reagents, 31 combine to make both the synthesis and purifcation of O 6 -(alkyl)-G derivatives laborious. With copper catalysed alkylation the carboxymethyl motif is introduced in a single step from naturally occurring precursors. ## Results and discussion Reaction discovery and substrate scope Cu(I) carbenes derived from a-diazocarbonyl compounds are highly selective for O 6 alkylation of unpaired guanines and inosines in ssDNA, monophosphates (MP) and triphosphates (TP) (see Fig. 1B and Table 1). The surprising selectivity prompted us to synthesize an O 6 -alkyl-inosine adduct by traditional routes to check the site-selectivity: identical characterization data confrm the proposed structures (see the ESI, † page S5-6). The method is the most concise and selective approach ever reported to the O 6 -(alkyl)-G motif and it is postsynthetic, obviating the need for solid phase DNA synthesis with modifed phosphoramidites. The convergent assembly opens the door to a range of O 6 -(alkyl)-G derivatives from simple starting materials. We explored the efficiency and selectivity of O 6 -G targeting by characterizing reactions with nucleotide monophosphates. Treatment of dGMP or GMP with 20 mol% Cu(II)SO 4 (in situ reduced by ascorbate) and ethyl a-diazoacetate (EDA) in aqueous buffer resulted in high conversion and isolated yields of ethylcarboxymethyl product (see Table 1, entry 1 and 2) within 30 min. Under the same conditions, the structural analogues IMP, GTP, and dGTP were efficiently alkylated at their O 6 positions (see Table 1, entry 6, 11 and 12). In contrast, reactions without catalyst (entry 5) or with different MPs like UMP, AMP, CMP, or TMP (entries 7-10) resulted in little or no detectable alkylation. Rather, in cases where substrates lacked a guanine, the starting material was partially converted through unselective pathways that we could not completely characterize in each case some alkylation was usually observed in the MS but these did not give substantial HPLC peaks that we could isolate. The same was also seen in short DNA fragments such as d(TAT) (see entry 14); here again starting material was consumed over time, but no single alkylation product could be isolated. In contrast, when guanine was present post-synthetic modifcations of a DNA 3-mer d(TGT) and 9-mer d(TTTTGTTTT) with the EDA-derived Cu(I) carbene gave a single large alkylation product (only O 6 -G alkylation observed, see Table 1 entries 13 and 17). In the substrates that lack guanine we cannot account for the mass balance: we believe in these cases that oxidation is a side-reaction that consumes starting material when no guanine is available for the carbene-based reaction. Cosolvent, oligonucleotide size, and diazo substrate all modulate reactivity While oligonucleotides also reacted selectively at O 6 -G, the reactions were slower. For example, d(ATGC) was alkylated at O 6 -G (determined by MS-MS sequencing, see ESI page S18 †) but required longer reaction times and conversion was lower than for the mononucleotides (cf. Table 1 entries 11 & 15). This reduced reactivity can likely be attributed to the unproductive sequestration of copper to other binding sites (phosphate backbone, other nucleobases) in the nucleic acid. The lower reactivity with longer oligonucleotides was mitigated by using 1,4-dioxane as a cosolvent. For example, changing the stock solution of EDA from 10% tBuOH in water to dioxane nearly doubled the conversion of d(ATGC) (cf. Table 1 entries 15 & 16). Nevertheless, these observations point to a limitation of the method: the larger the oligonucleotide the slower it is likely to react. Diazo substrate variation is another approach to improve the reaction with challenging nucleic acid substrates. Diazoacetamides 33 (DAAs) led to higher conversions than EDA and high O 6 selectivity was maintained (see Table 2). For example, 2), whereas the same substrate with EDA led to only 38% conversion (see entry 15, Table 1). While DAAs are better substrates in nucleic acid alkylation, they are also less stable; it was, for example, impossible to prepare a stock solution in 10% tBuOH and water due to rapid decomposition. DMSO as co-solvent stabilizes the DAAs, with the drawback that their reactivity in nucleic acid alkylation is reduced. Changing to a less coordinating cosolvent like dioxane seems to be the best compromise between these competing effects. For example comparing entries 3 & 4 in Table 2 we see that in 20% dioxane higher conversion was obtained in a ffth of the time (62% after 150 minutes versus 74% after 30 minutes). With the more reactive DAAs, even longer DNAs such as 16-mers were alkylated efficiently (see entries 8 & 9, Table 2). The amide substrates offer the additional advantages of stability and modularity. Customization of substrates for pull-downs or with fluorescent tags would simply require the addition of the appropriate amine in the DAA synthesis. 33 As an example, modifcation of guanine at the O 6 position of d(ATGC) with a terminal alkyne-bearing diazo substrate (see entry 10, Table 2) followed by a click reaction with azide-conjugated biotin (see the ESI page S31 †), installed a complex pull-down tag in only two steps. ## Mechanistic hypothesis The selectivity observed with copper carbenes does not reflect the intrinsic nucleophilicity of DNA nucleophiles. 34 In studies with powerful alkylating agents, it has been established that N7G is the most potent nucleophile in both dsDNA and ssDNA. O 6 -G alkylation is found fve to tenfold less, while the remaining heteroatoms account for approximately fve percent of alkylation products. 21 Since unstabilized Cu(I) carbenes should be highly reactive with water, we expected that the selectivity might stem from pre-association of copper with the guanine nucleobase. Guanine is the most electron rich of the nucleobases and readily chelates with divalent metals, including copper(II). 35,36 DFT calculations of molecular dipoles in DNA bases indicate that guanine has the largest molecular dipole (6.9 D) and its negative pole is located between O 6 and N7. On the other hand, little is known about the interaction of Cu(I) with DNA, 37 likely because the Cu(I) oxidation state has limited stability in aqueous buffer at neutral pH (it tends to oxidize or disproportionate) 38 and is therefore difficult to study. Nevertheless, based on analogy to Cu(II), which readily binds guanine N7s, we hypothesize that the selectivity derives from carbene formation with an N7 bound copper atom. Crystal structure data for Cu(II) 39 and other divalent metals 36 show a strong preference for N7G coordination, with a consequent positioning of the metal approximately 3.5 away from O 6 . The bound carbene would be thus poised to form a metal ylide with the O 6 position (see lower mechanistic scheme in Fig. 2), and an irreversible protonation would complete the alkylation. The selectivity that we observe mirrors the preference in sites of Cu-catalysed oxidative (peroxide or superoxide) damage in DNA, which show increased oxidation at and around guanine residues. 40 The research on copper-mediated redox damage supports two main pathways. In the frst an unbound copper(I) ion releases reactive oxygen species upon oxygen reduction or hydrogen peroxide cleavage, the resulting hydroxyl radical diffuses to the nearest site on DNA and typically leads to DNA fragmentation. This mechanism is unselective and presents a constant danger to DNA and RNAit is, however, inefficient. In the second mechanism a guanine-bound copper(I) forms a copper(II) oxo complex by reacting with peroxide, leading to oxidative damage of the guanine itself as well as neighbouring bases. These reports on oxidative damage, however, provided no insight on why guanine was the target or how the copper bound. We propose that both the specifcity of oxidative damage observed in earlier reports and the alkylation specifcity we observe here is a result of the same thing: guanine's superior ability to coordinate copper. The hypothesis that guanine pre-coordination controls chemoselectivity would predict the following: (1) Other amide or lactone substrates that lack the strong coordination ability of guanine should be poorly reactive. (2) Strongly chelating multidentate ligands should be deleterious to reactivity, since the combination of the guanine and the multidentate ligand would not leave space in copper's coordination environment for carbene formation. Over the next two sections we will examine experimental data for each of these scenarios and demonstrate that the hypothesis holds. 1. Normal amide or lactone substrates are poorly reactive. The most compelling example for the importance of N7G coordination is the 7-deaza-2 0 -deoxyguanosinetriphosphate (deaza-dGTP), which shows substantially reduced reactivity and selectivity in comparison to normal guanine (see the HPLC traces in Fig. 3). The simple thymidine substrate as well as the cyclic lactam piperidin-2-one are similarly unreactive, adding further weight to the idea that the N7 coordination is the decisive factor in O 6 alkylation specifcity. Although no literature precedence exists for this Cu(I)-based mechanism, a recent report has shown that histidine residues in proteins can control selectivity in the Cu(II)-catalysed alkenylation reaction of the amide backbone. 16 2. Multidentate ligands inhibit and N-heterocyclic carbene (NHC) ligands promote guanine alkylation. The copper-catalysed azide alkyne cycloaddition (CuAAC) is a widely used ligation reaction in chemical biology. Ligands such as bathocuproine and Fig. 2 Mechanistic proposal accounting for the high O 6 -G chemoselectivity. Electrostatic potential map and the molecular dipole moment (6.9 D) were calculated in Spartan 0 14 with DFT calculations at the EDF2 6-31G* level of theory using 9-methylguanine. tristriazoles are often used to stabilize Cu(I) and accelerate the reaction. In our case, however, the addition of strongly coordinating ligands such tris(3-hydroxypropyltriazolylmethyl) amine (THPTA) or bathocuproinedisulfonic acid disodium salt (BCA) lead to a suppression of guanine alkylation. Even after several hours little or no alkylation product could be detected (see Fig. 4A). In the case of THPTA the copper carbene was still formed (see disappearance of EDA in HPLC trace) but no guanine alkylation was observed. Instead the diazo compound is consumed in unproductive decomposition reactions. A copper catalyst bearing an N-heterocyclic carbene (NHC) ligand behaved differently: mesitylimidazolinium copper chloride (MesCuCl) gave the same alkylation profle as the standard 'ligandless' conditions, with O 6 -G alkylation being the dominant product (see Fig. 4B). It is revealing that the NHC ligand, which is not bidentate and hence would still allow coordination with N7G, is the only type tested that maintained alkylation activity. The NHC ligands have other favourable properties for the alkylation reaction (such as reducing oxidative damage) and we are currently exploring the full potential of NHC ligands in DNA and RNA alkylation. ## Biochemical studies with O 6 -(alkyl)-G adducts With ready access to O 6 -G derivatives we next used the products to study two open questions in the biochemistry of O 6 -G adducts: frst, their repair by alkylguanine transferases (AGTs) and second, their incorporation during DNA replication. Human AGT repairs peptide-like carboxamide adducts AGTs are found in all domains of life and are responsible for repairing O 6 -G adducts in the genome. Human AGT (hAGT) repairs O 6 -(alkyl)-G adducts from single stranded and duplex DNAs. The repair pathway is important since polymerases often mis-incorporate T when bypassing this lesion during replication. 48,49 Evolution has found no effective catalytic solution to the direct reversal of O 6 -G damage; instead AGT acts as a suicide protein. It bears a nucleophilic cysteine that attacks the alkyl group of the adduct and restores the native guanine. 44 Previous reports conflict as to whether carboxymethyl damage is repaired by hAGT. 50,51 In addition, since diazotized peptides are the likely source of these alkylation adducts, 26,50 it is important to understand whether adducts bearing amides are also effectively repaired. Bacteria also produce diazo peptide natural products that may alkylate DNA, 52,53 adding additional impetus to understand the repair of amide adducts. Thus, we tested whether a DNA 9-mer modifed with a cyclohexyl carboxymethylamide (see Table 2, entry 6), was a substrate for hAGT (see Fig. 5). Although removal of the adduct was slow (reduced rates are expected for ssDNAs), 54 analysis of both the deoxyoligonucleotide (see panel B in Fig. 5) and the protein (see panel C in Fig. 5) indicate repair occurred. These results support the most recent claim that carboxymethyl adducts are indeed repaired, 51 and it adds the carboxymethyl amide motif to the list of deoxyoligonucleotides repaired by hAGT. 6A). The direct modifcation approach offers unprecedented efficiency in the formation of O 6 -G adducts (in this case from a commercial triphosphate without any protecting group or phosphorylation steps). Previous syntheses of O 6 -G modifed nucleosides and triphosphates involved four 31 to seven steps including triphosphate introduction using strong electrophilic phosphorous reagents. 59 To investigate the incorporation of O 6 -(cm)-dGTP and O 6 -(ethylcm)-dGTP during DNA synthesis, we performed single nucleotide primer extension experiments with KTqM747K, a mutant of KlenTaq polymerase that bypasses several DNA modifcations. Unmodifed 28-mer templates (5 0 -ATT ATG CTG AGT GAT ATC CCT CTX CTC A) in which X ¼ G, A, T or C were annealed to a 5 0 -end [g-32 P] radiolabeled 23-mer primer (5 0 -AGA GGG ATA TCA CTC AGC ATA AT) and then subjected to primer extension conditions (see Fig. 6B). Reaction products were separated by denaturing polyacrylamide gel electrophoresis and visualized with phosphorimaging. Nucleotides O 6 -(cm)-dGTP as well as O 6 -(ethylcm)-dGTP were incorporated opposite both C and T, but C was especially well tolerated. Furthermore, whereas with T incorporation the polymerase stalls after the frst extension, the O 6 -(alkyl)-G:C mismatch seems to pose no problem for the polymerase's procession since the second extension also occurs quickly (see the C lane in Fig. 6C). Consistent with steric matching as a key component in polymerase fdelity, purine mismatches were poorly tolerated (see G and A bands in the gel in Fig. 6C). These experiments serve as the proof-of-concept that O 6 -(cm)-G damaged nucleotides can be incorporated during DNA replication. In light of these results a full study of how different polymerases (particularly human) deal with O 6 -(cm)-dGTP in the dNTP pool during replication is warranted. ## Conclusions We have discovered a new way to create O 6 -G adducts postsynthetically in nucleic acids of varying size, from nucleotide mono-or triphosphates to long oligonucleotides. The ability to obtain high yields of precisely defned O 6 -G adducts eliminates a major roadblock in studying their biochemistry, a point we have illustrated through the frst synthesis of O 6 -(cm)-dGTP and the frst demonstration of its incorporation during DNA replication. Copper-catalysed O 6 -G alkylation is a dump-and-stir procedure that employs readily available materials. a-Diazo ester substrates are stable enough that several are sold commercially and a simple 33 two-step (one-pot) protocol for a-diazo ester synthesis makes nearly any diazo compound in this class synthetically accessible. Research on O 6 -G adducts in biology 67 and biotechnology 68 demands ready synthetic access to these adducts and the method we have described is the simplest to date. Nucleic acids are found in diverse contexts and can fold in myriad ways, complicating attempts at broad generalizations of reactivity. Nevertheless, the weight of evidence suggests N7G is the most powerful nucleophile 34 and ligand 36 in DNA and RNA. It is a primary target of organic electrophiles, 34 as well as organometallic anticancer drugs. 69 To the list of N7G's functions we add the role of directing group in copper catalysed reactions. We have focussed on alkylation reactions, but the same effect is likely responsible for the well-known propensity of Fenton-type oxidative damage to be localized around guanines. research commission (ETH-43 14-1) for fnancial support. Prof. Nathan Leudtke and Mr Aaron Johnson are gratefully acknowledged for helpful discussions.

chemsum

{"title": "Copper carbenes alkylate guanine chemoselectively through a substrate directed reaction", "journal": "Royal Society of Chemistry (RSC)"}

chiral_anion_recognition_using_calix[4]arene-based_ureido_receptors_in_a_<i>1,3-alternate</i>_confor

## Abstract: The introduction of chiral alkyl substituents into the lower rim of calix [4]arene immobilised in the 1,3-alternate conformation led to a system possessing a preorganised ureido cavity hemmed with chiral alkyl units in the near proximity. As shown by the 1 H NMR titration experiments, these compounds can be used as receptors for chiral anions in DMSO-d 6 . The chiral recognition ability can be further strengthened by the introduction of another chiral moiety directly onto the urea N atoms. The systems with double chiral units being located around the binding ureido cavity showed better stereodiscrimination, with the highest selectivity factor being 3.33 (K L /K D ) achieved for N-acetyl-ʟ-phenylalaninate. The structures of some receptors were confirmed by single crystal X-ray analysis. ## Introduction The recognition and complexation of anions has become undoubtedly one of the most important branches of modern supramolecular chemistry, as can easily be demonstrated by an immense number of recent reviews and books devoted to this topic. Due to the omnipresence of anions in bio-logical systems, their irreplaceable roles in cell functioning have gradually been revealed and are well recognised to date. Consequently, given the importance of anions in many areas of everyday life, including, e.g., biology, medicine, environmental pollution issues, or industrial processes, the design and develop- ment of novel artificial receptors/sensors for anions is becoming more and more significant . There are many strategies aiming at anion recognition in the literature. Most of the receptors, however, rely on electrostatic interactions. These systems are represented by positively charged molecules, such as quaternary N-, S-, and P-containing onium salts, protonated or alkylated aza-crown ethers and azacryptands, amidinium and guanidinium cations, etc. . Due to the low directionality of the Coulomb force, the successful application of purely ionic interactions in the design of selective anion receptors is rather limited. The shapes and geometries of anions are widely different, and therefore the design of corresponding tailor-made receptors is based mostly on more directional interactions, such as hydrogen bonds. Indeed, an incredible number of neutral receptors bearing amide, sulfonamide, urea, thiourea, pyrrole, or triazole functional groups (to name at least some of them) has appeared during the last two decades . Due to well-established functionalisation approaches, calix arenes are frequently used as molecular plat-form in the design of more complex receptor systems. The existence of four basic conformations (cone, partial cone, 1,3-alternate, and 1,2-alternate) offers the combination of a precisely defined 3D structure, with functional groups being introduced at exactly defined mutual positions. This makes calix arenes an ideal molecular scaffold for the construction of highly sophisticated molecules, including anionic receptors . During our ongoing research on anion complexation, we have reported various calix arene receptors based mainly on amide, urea, or thiourea groups , some of which are available in different conformations. Although the overwhelming majority of calixarene-based receptors makes use of the cone conformer A (Figure 1), the corresponding diureidocalix arenes in the 1,3-alternate conformation B showed surprisingly good complexation abilities towards selected anions. Especially for chiral anion recognition, contrary to the cone receptor, a design based on the 1,3-alternate conformer enables the introduction of chiral units into the phenolic functions of the inverted aromatic moieties nearby the ureido cavity responsible for the binding, as in C. This design can be exemplified by our previously published receptors C1 based on a calix arene moiety Scheme 1: Synthesis of the calix arene-based chiral anionic receptors 7 and 8. or by C2 using thiacalix arene as the core scaffold . Moreover, the introduction of the tert-butyl groups into the 1,3-alternate conformer should lead to the overall increase rigidity of the molecule, possibly enhancing the interactions within the binding cavity. In this context, we realised that further strengthening of the chiral induction can be reached via the synchronous application of chiral units on the ureido moieties as well, as in D. In this paper, we report the preparation and complexation study of the latter type of receptor, bearing double chiral units in the immediate proximity to the preorganised ureido cavity. ## Results and Discussion The introduction of the chiral alkyl moiety based on (S)-2methylbutan-1-ol into the starting calix arene 1 was carried out using recently described Mitsunobu reaction conditions . Refluxing the reaction mixture of PPh 3 , DIAD, and tolu-ene for two days provided the distally dialkylated calixarene 2 in 64% yield (Scheme 1). Compound 2 was regioselectively ipso-nitrated with 30 equiv of 65% aq HNO 3 in an AcOH and CH 2 Cl 2 mixture, making use of the higher reactivity of the nonalkylated phenolic moieties . The fast reaction afforded the corresponding dinitro derivative 3 within a few minutes with 70% yield. As we found in our previous attempts , the alkylation of dinitrocalixarenes to form the 1,3-alternate conformation is a synthetic challenge. In fact, irrespective of the base or solvent used, the alkylation with appropriate n-propyl or n-hexyl halides always led to the partial cone conformation as the main product. To overcome this problematic step, we used the conditions described by Böhmer et al. for a similar system bearing propyl groups on the lower rim. Indeed, one week of stirring 3 with allyl bromide in the presence of Cs 2 CO 3 provided the expected 1,3-alternate conformer 4a in 40% yield accompanied by a small amount of the partial cone conformer 4b (10%). The unequivocal proof of the structure of the isomer 4a was provided by single crystal X-ray analysis. The compound crystallised in a tetragonal system, space group P4 1 2 1 2 as a 1:1 complex with methanol used as a solvent for crystallisation. As shown in Figure 2, the calixarene clearly adopts the 1,3-alternate conformation with an almost ideal tetragonal shape of the cavity. The lengths of the two main diagonals (the distances between opposite bridging CH 2 moieties) are essentially identical (7.183 and 7.206 ). If the main plane of the molecule is defined by the four bridging C atoms, all phenolic subunits are almost perpendicular to this plane, with the aromatic parts being slightly tilted out of the cavity. The corresponding interplanar angles Φ with the aromatic subunits are 81.55°, 80.78°, 77.08°, and 80.57°, respectively, starting counterclockwise from the upper subunit bearing a nitro group (Figure 2a). The 1,3-alternate isomer 4a represents quite an interesting synthetic intermediate as the presence of the allyl groups immobilizes the required conformation and at the same time enables the potential incorporation of the macrocycle into a polymeric matrix. Consequently, the subsequent reduction step was carried out in two alternative ways: (i) exclusive reduction of the nitro groups or (ii) concomitant reduction of the nitro groups and the allyl moieties. Thus, the reaction of 4a with SnCl 2 •2H 2 O in ethanol gave the corresponding amine 5 in 57% yield after column chromatography on alumina. On the other hand, the four-day stirring of 4a with Pd/C under a H 2 atmosphere (5 atm) in an autoclave at room temperature provided compound 6 in 91% yield. The starting compound 5 was then reacted with commercially available isocyanates comprising p-nitrophenyl isocyanate, p-nbutylphenyl isocyanate, (S)-α-methylbenzyl isocyanate, and (R)-α-methylbenzyl isocyanate. The reactions were carried out at room temperature in anhydrous dichloromethane, and the products 7a-d were isolated in 40-60% yields. Similarly, the propyl-substituted analogues 8a and 8b were obtained from the reaction of 6 with the corresponding isocyanates in 38% and 45% yield, respectively. The structures of the selected receptors 7a and 7d were further proven by single crystal X-ray studies. The calixarene 7a crystallised in a monoclinic system, space group C 2 , and the unit cell contained two receptors with four molecules of DMSO used as the crystallisation solvent (7•2DMSO complex). Both calixarene molecules exhibited an almost ideal square shape of the cavity, with the lengths of the main diagonals being 7.322 × 7.122 and 7.323 × 7.137 , respectively. Every ureido group held one molecule of DMSO via synchronous hydrogen bonding interactions between the two NH protons and a sulfoxide oxygen atom (Figure 3a). The S=O•••H-N distances were 1.995, 2.285, 2.033, and 2.328 , indicating strong interactions in the solid state. At the same time, the carbonyl groups from the neighbouring receptor urea moieties interacted with the C-H bonds of the DMSO methyl group (the C=O•••H-C distances were 2.486 and 2.452 ), thus forming the calixarene dimer with a head-to-tail mutual orientation. The overall supramolecular binding motif was completed by the close contacts between carbonyl oxygen atoms (of the urea group) and S atoms (of DMSO), indicating possible chalcogen interactions , and the C=O•••S=O distances were 3.269 and 3.308 (Figure 3a). The molecular packing was further strengthened by the π-π interactions of the p-nitrophenyl moieties, exhibiting several close C Ar •••C Ar contacts at a 3.373 distance (Figure 3b). The receptor 7d crystallised in a triclinic system, space group P 1 , as a 1:3 complex with acetone (used as solvent for crystallisation). The main packing motive (see Figure 4) was represented by an infinite chain of calixarene molecules joined together by intermolecular hydrogen bonds between the ureido groups (the C=O•••H-N distances were 2.293 and 2.048 ). The complexation ability of the novel receptors towards selected chiral anions was studied using standard 1 H NMR titra-tion experiments. There are several reasons why DMSO-d 6 was selected as the solvent for the complexation studies: (i) it dissolves all anionic species tested, (ii) prevents the receptor molecule from self-association, and (iii) reduces the complexation constants to values easily measurable by 1 The results summarised in Table 1 and Table 2 revealed that the complexation properties of the receptors do not depend significantly on the lower-rim substitution of the calixarene. Comparing the complexation constants for the otherwise identical receptors (7a vs 8a or 7b vs 8b), it is obvious that the presence of allyl vs propyl substituents does not impose much difference in terms of absolute K values and selectivity. As expected, the nitro-substituted receptors 7a and 8a exhibited complexation constants higher than the butyl-substituted analogues 7b and 8b for all anions -compare 7a (Table 1, run 5, K = 660 M −1 ) vs 7b (Table 2, run 15, K = 90 M −1 ). On the other hand, despite the differences in the K values, the enantioselec- tivity remained almost the same in both receptor series. Thus, the selectivity factor s = K L /K D for phenylalaninate was 1.06 for 7a and 1.11 for 7b, and similar values were also obtained for 8a (1.05) and 8b (1.04), using the ʟ-enantiomer in each case. The highest chiral discrimination (for an ʟ-enantiomer) was achieved for N-acetyl-ʟ-leucinate, with a selectivity factor of s = 2.0 and 1.87 for 7b and 8b, respectively. The above-mentioned results indicate that the binding cavity composed of two preorganised ureido groups and chiral alkyl substituents in the near proximity possesses some ability of enantioselective recognition. In fact, this assumption was already established by our previous receptors C1 and C2 (Figure 1) , although the direct comparison with our novel results is rather difficult due to the different types of derivatives (C1 has a cavity without tert-butyl groups, and C2 is based on thiacalix arene). Nevertheless, the simultaneous introduction of chiral isocyanate to form chiral urea moieties has not been accomplished yet. The formation of another chiral centre within the derivatives 7c and 7d led to the expected decrease of the complexation constant values (by approximately one order of magnitude) due to the presence of alkyl instead of aryl urea receptors (Table 3). Thus, going from 7a to 7c, the complexation constant K for ᴅ-leucinate decreased from 480 to 50 (run 7, Table 1 vs run 27, Table 3), and the appropriate values for ᴅ-phenylalaninate are 660 vs 40 (run 5, Table 1 vs run 25, Table 3). On the other hand, the chiral recognition properties of the receptors 7c and 7d are accentuated compared to 7a and 8a or 7b and 8b. The introduction of another chiral centre leads to higher selectivity factors in almost all the cases. The stereodiscrimination of 7c (bearing an (S)-α-methylbenzyl moiety on the urea group) for N-acetylphenylalaninates represents the maximum value achieved (s = 3.33 for ʟ). Interestingly, the diastereomeric isomer 7d, possessing an (R)-chiral centre, prefers the ᴅ-isomer, with a selectivity factor of s = 1.75, and the same holds for N-acetyl-ᴅ-leucinate (s = 1.34). These results indicate that both chiral moieties (the alkyl group on the calixarene and the chiral centre on the urea moiety) function synergistically, and a proper choice of both substituents can lead to an even better stereoselectivity. ## Conclusion In conclusion, the introduction of chiral alkyl groups into the lower rim of calix arene immobilised in the 1,3-alternate conformation resulted in a macrocycle with a preorganised ureido cavity bearing chiral alkyl substituents in the near proximity. As shown by 1 H NMR titration experiments, these compounds function as receptors for chiral anions in DMSO-d 6 . The chiral recognition ability can further be strengthened by the introduction of another chiral moiety directly to the urea nitrogen atoms. The systems with double chiral units located around the binding ureido cavity showed a better stereodiscrimination, with the highest selectivity factor being 3.33 (for ʟ) achieved for N-acetylphenylalaninate.

chemsum

{"title": "Chiral anion recognition using calix[4]arene-based ureido receptors in a <i>1,3-alternate</i> conformation", "journal": "Beilstein"}

<i>β</i>-arylation_of_oxime_ethers_using_diaryliodonium_salts_through_activation_of_inert_c(sp)–h_bo

## Abstract: An efficient method of selective b-arylation of oxime ethers was realized by using a palladium catalyst with diaryliodonium salts as the key arylation reagents. The reaction proceeded smoothly through the activation of inert C(sp 3 )-H bonds to give corresponding ketones and aldehydes. This convenient procedure can be successfully applied to construct new C(sp 3 )-C(sp 2 ) bonds on a number of complex molecules derived from natural products and thus serves as a practical synthetic tool for direct late-stage C(sp 3 )-H functionalization.Scheme 1 Pd-catalyzed b-arylation reaction of oxime ethers via C(sp 3 )-H bond activation. ## Introduction Arylation via direct activation of inert C-H bonds has emerged as a fascinating feld, which could provide useful aromatic compounds with high atom-and step-economy. 1 In the past decade, signifcant progress has been made in the development of transition-metal catalyzed arylation on C(sp 2 )-H which enables coupling a large range of substrates to various aromatic reagents, 2 including more challenging work on enantioselective construction of stereo-centers published most recently. 2j-l In comparison, arylation on an inert C(sp 3 )-H bond (simple alkyl C-H bond) is much less explored, with a scope of limited substrates reported. 3 One of the most successful strategies is transition-metal catalyzed-arylation of carboxylate derivatives, including carboxylic acids, esters and amides. 4 Assisted by big auxiliary groups, alkyl amines could also be selectively arylated on the chain. 5 In light of these advances, we were encouraged to develop new arylation reactions through direct activation on alkyl C-H bonds with a wider scope of substrates which offers unprecedented opportunities to efficiently synthesize valuable aromatic compounds. 6 With regard to chelation-assisted C-H bond functionalization, facile introduction and removal of the directing group could enlarge the scope of substrates from predesigned molecules and consequently enable the protocol to be applicable to various natural products, generating new and attractive sources for bioactive compounds with high diversity and functionality. 7 With this ultimate goal in mind, we successfully developed a Pd-catalyzed b-arylation reaction on the inert C(sp 3 )-H bond of oxime ethers to give useful products, which could be easily converted to important b-aryl ketones, amines and so on. 8 The reaction proceeds smoothly via C(sp 3 )-H activation with diaryliodonium salts as the key coupling partner reagents (Scheme 1). ## Results and discussion As revealed by known work, the main issues of transition-metal catalyzed functionalization on C(sp 3 )-H bonds not only resulted from the inertness and abundance of C(sp 3 )-H bonds in organic compounds but, more essentially, from the inherent instability of in situ generated catalytic metal species, which may easily undergo b-hydrogen elimination or side reactions. 9 To some extent, a good solution is adjusting electronic and coordination effects of the catalytic species, but this generally required additional modifcation of the substrates. 10 Hence, a more straightforward solution might be to accelerate the transformation of the catalytic metal species by accomodating the proper coupling reagents with an appropriate chelating group. 11 With this strategy, we successfully realized the b-arylation reaction on the C(sp 3 )-H bond of oxime ethers with Pd(OAc) 2 as the catalyst. During the study of C-C bond formation on inert C-H bonds, we initiated the investigation with arylating 2methylcyclohexanone O-methyl oxime, which is quantitatively synthesized from a-methylcyclohexanone. To begin with, when PhBr or PhI was used as the coupling reagent under various conditions, it always failed to give the desired arylation product, 3aa, and dehydrogenation side products 1a 0 and 1a 00 were observed with the formation of Pd-black. This implied that the putative Pd-species 5 was generated from 1a with Pd(OAc) 2 , but due to the instability, PhBr or PhI could not be coupled to species 5 before it underwent b-hydrogen-elimination. In order to facilitate the desired arylation reaction, Ph 2 I + PF 6 , 2a$PF 6 (1 equiv.) was chosen as the coupling reagent and 3aa was formed, albeit in 17% yield (Table 1). 12 The addition of a base slightly increased the yield of 3aa, but the best yield was only 27% when Ag 2 CO 3 (2 equiv.) was employed. By adding pivalic acid, the starting materials were fully converted to produce 3aa with 50% yield. 13 In order to further accelerate the transfer step of the phenyl group, we added the polar solvent, hexa-fluoroisopropanol (HFIP) to the reaction, resulting in an increase to 82% yield of 3aa. The use of 2a$OTf gave an even better result (87%, entry 17, isolated in 83%), while 2a$BF 4 failed to generate 3aa. For further modifcation, when the reaction was completed under the optimized condition, entry 17, the mixture was treated with formaldehyde in an acidic system, 14 and 2-benzyl cyclohexanone, 4aa, was isolated in 80% yield (for experimental details, see ESI †). In accordance with the initial proposal, ketoximes and ketones can be easily interconverted, making this method an efficient approach to b-aryl ketones. With the optimized conditions established, the scope of diaryliodonium salts was examined for b-arylation of 1a. As shown in Scheme 2, diaryliodonium salts with a range of substituents were efficiently coupled using these conditions. Most of the products were isolated in the ketone form (4) rather than oxime ethers (3), since ketones were considered to be more synthetically useful. The diaryliodonium triflates with F , Cl and Br substituents at the para position (2b-2d) reacted smoothly with 1a, generating corresponding products (4ab-4ad) in good yields. The use of methyl and t butyl substituted diphenyliodonium triflates (2e-2f) and 1a provided products (4ae-4af) in slightly lower yields. Diaryliodonium triflates bearing a strong electron-withdrawing group (-CO 2 Me and -CF 3 ) or strong electron-donating group (-OMe) also yielded desired products (4ag-4ai) while 3ah was obtained under a lower temperature of 70 C. The reaction of 1a with orthosubstituted diaryliodonium triflates (2j-2l) also afforded expected products (4aj-4al) in satisfactory yields. However, the use of asymmetric diaryliodonium salts Ar-I + -MesX failed to give products. A number of selected oximes (3aa, 3ad, 3 ag, 3ah, 3ai, 3aj, and 3ak) were isolated to demonstrate the original efficiency. To further investigate, a series of substituted acylic oxime substrates were examined to explore the regioselectivity of the arylation with di(4-bromophenyl)iodonium 2d. As shown in Scheme 3, ketoximes 1b-1d all reacted with 2d to give monoarylated products at the a-methyl group in good yields, while the generation of bi-arylated compounds remained at trace amounts. The use of aldoxime 1e only produced 4ed in moderate yield while the bulky ketoxime 1f with four a-methyl groups afforded the mono-arylated product 4fd in good yield. In addition, we examined the effect of substituents on cyclohexyl oxime ethers and 4nd was isolated in 83% yield from 1n. It is known that the activation of methylene C-H bonds is harder than of methyl C-H bonds in transition-metal catalyzed reactions because C-H bonds in the methylene position are more hindered. 15 As shown herein, some of the ketoximes with the appropriate confguration, for instance 1g, could successfully be arylated with 1d to give 4gd in 42% yield. Similarly, it also worked with an oxime ether containing a 1-adamantanyl group, 1h, to produce 3hg in 47% yield (Scheme 4). Since the direct modifcation of natural products at a normally inert position has attracted much attention due to the potential to access new bioactive compounds from "starmolecules", 1a,16 some complex substrates with natural product backbones were examined in our new arylation reactions for reactivity, selectivity and tolerance of functional groups (Scheme 5). Oxime 1i, derived from naturally-occurred fenchone which contains three methyl groups, could be selectively arylated on an exo-methyl group with 2g under standard conditions, giving 3ig in 65% yield. Naturally abundant in many essential oils, (+)-carvone can be easily converted to a,b-unsaturated ketone oxime with a-methyl group 1j. This can then can be arylated with 2g to give 3jg under the synthetic method established, albeit in 38% yield. However, with 1 equivalent of diaryliodonium salt 2g, the mono-arylated product of 1k derived from lanosterol was observed as two inseparable isomers in a relevantly low yield. Alternatively, when 2 equivalents of 2g were employed in the transformation, the bis-arylated product 3kg was successfully obtained in 73% isolated yield. The structure of 3kg was confrmed by XRD analysis, shown in Fig. 1A. b-Glycyrrhetinic acid is a major metabolite of glycyrrhizin, one of the main constituents of licorice, and has been shown to exhibit anti-ulcerative, anti-inflammatory, and immunomodulatory properties. Substrate 1l, derived from glycyrrhetinic acid, could also be bis-arylated on both methyl C-H bonds to form 3lg with 69% isolated yield. Besides natural product-like compounds, the synthetic reagent, methasterone was converted to oxime 1m, which was then successfully arylated with 2d to give 4md in high yield, with the hydroxyl group unchanged during the reaction. In addition, oxime ethers can be easily converted to corresponding amines which are useful building blocks with a range of potential applications (Scheme 6). 17 In terms of the mechanism study, we proposed that the process was initiated by a cyclometalated complex. 18 Using 1a as the starting backbone, the isolation of the palladation intermediate always failed due to the strong tendency of b-H elimination of 5a. When treating 1d with Pd(OAc) 2 , the existence of palladation intermediate 5d was proved. By converting 5d to 6 (Scheme 7), 19 the crystal structure of 6 was identifed, indicating that palladium was bound with CH 2 and oxime-nitrogen atom as a cyclometalation species (Fig. 1B). In order to confrm the catalytic competency, complex 6 was treated with 1 equivalent of 2d and 2 equivalents of Ag 2 CO 3 , analogously to the standard reaction conditions, and 3dd was observed in 80% yield by in situ NMR (Scheme 7). Based on the above results and literature reports, a plausible mechanism is proposed in Scheme 8. First, the reaction of oxime ether 1 with Pd II species (Pd(OAc) 2 or other palladium salts generated in situ) would produce cyclopalladation species 5. And the oxidative addition of diaryliodonium salt 2 to 5 would afford Pd IV intermediate 7. 11 The reductive elimination of 7 would give product 3, and release the Pd II species into the catalytic cycle. ## Conclusions In summary, we have developed a novel b-arylation reaction on inert C(sp 3 )-H bonds of oxime ethers. The reaction offers new opportunities to prepare useful b-arylated oximes from/to ketones and aldehydes via simple transformation with good efficiency and step-economy. The easy manipulation and good tolerance of functional groups enable the method to be used to modify many complex compounds derived from natural backbones. Further investigations on the scope, mechanism, and synthetic application of this new reaction are under way in our laboratory.

chemsum

{"title": "<i>\u03b2</i>-Arylation of oxime ethers using diaryliodonium salts through activation of inert C(sp)\u2013H bonds using a palladium catalyst", "journal": "Royal Society of Chemistry (RSC)"}

aqueous_pka_prediction_for_tautomerizable_compounds_using_equilibrium_bond_lengths

## Abstract: The accurate prediction of aqueous pK a values for tautomerizable compounds is a formidable task, even for the most established in silico tools. Empirical approaches often fall short due to a lack of pre-existing knowledge of dominant tautomeric forms. In a rigorous first-principles approach, calculations for low-energy tautomers must be performed in protonated and deprotonated forms, often both in gas and solvent phases, thus representing a significant computational task. Here we report an alternative approach, predicting pK a values for herbicide/therapeutic derivatives of 1,3-cyclohexanedione and 1,3-cyclopentanedione to within just 0.24 units. A model, using a single ab initio bond length from one protonation state, is as accurate as other more complex regression approaches using more input features, and outperforms the program Marvin. Our approach can be used for other tautomerizable species, to predict trends across congeneric series and to correct experimental pK a values. A pproximately 21% of the compounds that make up pharmaceutical databases are said to exist in two or more tautomeric forms 1 . Tautomerism is a form of structural isomerism that is characterized by a species having two or more structural representations, between which interconversion can be achieved by "proton hopping" from one atom to another. Issues surrounding pK a prediction for species exhibiting this feature have been noted a number of times in the literature. Recently 2 , Connolly suggested that a lack of experimental information on both relative tautomer stability and the properties of distinct tautomeric forms are the likely causes of such issues. Tautomeric species present a challenge, not just to empirical-based approaches, but also to those that attempt to solve the pK a prediction problem using first-principles . For tools implementing the latter approach (e.g. Jaguar, Schrödinger 4,6,7 ), the most rigorous protocol includes quantum chemical calculations for conformations of each, or a select few low lying tautomer(s), in both gas-and solvent phase, and in both protonated and deprotonated forms. Therefore, without some element of empiricism, first-principles approaches often incur significant computational expense. For methods of pK a estimation that generate descriptors starting from 2D fingerprints, each tautomeric form of a species will correspond to a unique representation. Therefore, the user must either (i) possess prior knowledge of tautomeric stability in order to maximize prediction accuracy, or (ii) tautomer enumeration must be performed by the program based on an arbitrary user input, followed by selection of the optimal tautomer for calculation of chemical descriptors . A comparative study 11 of 5 empirical pK a prediction tools (ACD/pK a DB (http://www. acdlabs.com/home), Epik (http://www.schrodinger.com), VCC (http://vcclab.org), Marvin (http://www.chemaxon.com) and Pallas (www.compudrug.com)) on 248 compounds of the Gold Standard Dataset compiled by Avdeef 12 , demonstrated a tendancy for prediction errors to be higher for compounds with a larger number of possible tautomeric states. For the tool they tested, the guanidine group of the drug Amiloride and the enolic hydroxyl groups of herbicides Sethoxydim and Tralkoxydim were also identified as common outliers. Compounds containing a 1,3-diketo group exhibit tautomerism (shown in Fig. 1a(i), (ii)). For cyclic 1,3-diketones, the diketo state (Fig. 1a(i)) can be transformed into two keto-enol forms (Fig. 1a(ii)). Tautomeric states of the same molecule may be nondegenerate, with the ratio being influenced by the solvent environment and temperature 13 . The compounds 1,3-cyclohexanedione (1,3-CHD) and 1,3-cyclopentanedione (1,3-CPD) are known to possess significant keto-enol character in solution, a phenomenon attributed to the formation of hydrogen bonded solute dimers, and additional stabilization from solute-solvent interactions 14 . 1,3-CHD is a fragment prevalent to both agrochemically and pharmaceutically active compounds in use today. Alloxydim (Fig. 1b(i)) is currently used as a selective systemic herbicide for post-emergence control of grass weeds in sugar beet, vegetables and broad-leaved crops. Adding a derivatized benzoyl group at the 2-position in place of Alloxydim's 2-oxime forms what is known as triketone herbicide (e.g. Mesotrione, Fig. 1b(ii)). Pharmaceutically relevant compounds containing the 1,3-CHD group include the antibiotic Tetracycline and its analogues. Previous work from our group, as well as the earlier work of others, has highlighted the utility of bond lengths and Fig. 1 Structures of 1,3-diketone derivatives and schematic of our workflow. a (i) The diketo form of a 1,3-dione, (ii) the resonance canonicals for the keto-enol form of 1,3-diones, and (iii) the resonance canonicals for the anionic state, where n = 0 or 1 if the ring is fiveor six-membered, respectively. K T denotes the equilibrium constant between tautomeric states, K a(DK) denotes the dissociation equilibrium from the diketo state and K a(KE) the dissociation equilibrium from the keto-enol state. b (i) The global minimum geometry of Alloxydim, a 2-oxime herbicide and Mesotrione in the keto-enol anti state, (ii) a triketone herbicide. c The AIBL-pK a workflow implemented here for cyclic β-diketones. other quantum chemically derived descriptors in the context of Quantitative Structure Property Relationship studies 26 . Most recently, our approach to pK a prediction, which uses only internuclear distances as descriptors, called AIBL-pK a (ab initio bond lengths), showed remarkably accurate prediction of acidity variation across congeneric series of guanidine-containing species 19 and sulfonamides 20 . The current work brings attention to the issue of pK a prediction for tautomerizable compounds and delivers an intuitive solution to this problem for 1,3-CHD and 1,3-CPD derivatives, which remain important scaffolds in pharmaceutical and agrochemical research. ## Results and discussion Scheme for model construction. Our proposed method of predicting pK a values (Fig. 1c and Methods) makes use of equilibrium bond lengths from density functional theory calculations (B3LYP/6-311G(d,p) with the conductor-like polarizable continuum model or CPCM) as input features for regression models. The full dataset of 71 compounds used in this work represent a wide variety of substituent types and patterns (generic structures and examples of dataset compounds are shown in Fig. 2a). After an initial analysis of the linear fit of each individual bond length, we investigate whether the use of multiple bond lengths as input features could provide an advantage in prediction accuracy and model applicability radius. For this task, we considered all subset combinations of the bonding distances of the fragment common to each species. We also compared a number of machine learning methods for their regression onto pK a values, namely, random forest regression (RFR), support vector regression (SVR), Gaussian process regression (GPR) as well as partial least squares (PLS). PLS 27 and SVR have been implemented in the context of pK a prediction many times, using many different types of descriptors. A brief overview of the theory and method used for these approaches can be found in the Supplementary Methods section of the Supplementary Information (SI). Further details and formalism for the validation metrics used in this work (r 2 , RMSEE, MAE) can also be found in Supplementary Methods. Through our analysis, we demonstrate that a powerful model may be constructed from simple linear regression of a single ab initio bond length, thereby potentially negating the need for the more complex approaches. Current approaches. To exemplify the issues surrounding prediction for cyclic 1,3-diketones using existing empirical approaches, the commercial program by ChemAxon known as Marvin was used to estimate values for a series of 1,3-CHD and 1,3-CPD derivatives (o1-o8, tk1-tk15 and dk1-dk12 shown in Supplementary Table 1 of the SI). The Marvin program uses Gasteiger partial charges 30 , polarizabilities and structure specific increments to predict pK a values using ionizable group specific regression equations 11 . The results are shown in Fig. 2b, where the orange diamonds denote experimental values, blue squares represent Marvin predictions without the option to "consider tautomers/ resonance", while the magenta triangles are predictions made with this option. For the compounds in Fig. 2b where the blue and red points overlap, the program predicts the keto-enol state to be dominant, and delivers predictions that lie 0.8 units away from experimental values on average. However, for 60% of the compounds, the program predicts the diketo state to be dominant. For the series o1-o8, Marvin gives values of ~16 log units for 5 out of 8 species. For the remaining three compounds, o1, o3 and o7, the program identifies the acidic proton (pK a ~17) at the 4 or 6 position on the 1,3-CHD ring. The above results suggest that if accurate predictions are to be made (i.e. residual errors <1 pK a unit), then the user must have prior knowledge of the dominant keto-enol tautomeric form (blue squares in Fig. 2b). In the following sections we show that our method, which uses quantum chemically derived geometric descriptors, avoids such problems intrinsically. Despite the increased computation time compared with empirical Fig. 2 Exemplar cyclic diketone compounds studied in this work and the performance of Marvin versus Experiment. a pK a data for compounds of the dataset used were procured from both Syngenta's database and literature sources. The pK a values of 17 compounds were also measured for the purpose of this work. Each compound either contains a 1,3-cyclohexanedione (1,3-CHD) or 1,3-cyclopentanedione group (1,3-CPD), examples of which are shown in blue and green, respectively. Substituent variation occurs at 2, 4, 5 and/or 6 position on 1,3-CHD, and 2, 4 and/or 5 for 1,3-CPD. The full set of structures and experimental pK a values can be found in Supplementary Table 1 of the SI. b Experimental (orange) pK a values across the series o1-o8, tk1-tk15 and dk1-dk12, are compared with Marvin predictions with the "consider tautomers/resonance" option (magenta) and without this option (blue). Values are in excess of 14 log units for the acidic proton at C2 (labelled for 1,3-cyclohexanedione) for o2, o4, o5, o6 and o8 and values for o1, o3 and o7 correspond to C4/C6. approaches, AIBL avoids the need to compute pK a values for both protonation states. Moreover, descriptor calculations may be carried out only in the solvent phase using an implicit approach (CPCM). Identifying AIBL-pK a relationships for triketones. The relationship between the structure and herbicidal activity of triketones (Fig. 3a) was first reported 31 by Lee and co-workers. One of the primary conclusions of that early work was that the orthosubstituent on the phenyl ring is a requirement for the compound's herbicidal activity. The authors also noted that compounds with more electron-withdrawing para-substituents required a lower dose to obtain a 50% weed-control rating across 7 variants of broad-leaf plants (the metric known as lethal dose 50, or LD 50 ). It was thereby deduced that a linear relationship exists between Hammett constants of para-substituents, log (LD 50 ) and pK a . Therefore, a more electron-deficient benzene is associated with enhanced acidity and herbicidal activity 31 . As there is already evidence of a structure-property/activity relationship for these species, we took the set of 10 compounds from the work of Lee et al. as a starting point to assess the prevalence of AIBL-pK a relationships across available tautomeric states. The identities, pK a values, equilibrium bond lengths and log (LD 50 ) values of the compounds studied by Lee et al. are shown in Supplementary Tables 2-5, labelled as tkn1-tkn4 and tkc1-tkc6. All tkn species possess one 2-NO 2 group whereas each tkc species has a 2-Cl substituent (Fig. 3b). Across each subset the parasubstituent varies. We find that the order of stability of each compound in their four lowest energy tautomer/conformations (Fig. 3a) The triketo form a is ~9 kJ mol −1 less stable than the (endo) keto-enol anti form b, which in turn is ~29 kJ mol −1 less stable than the (exo) keto-enol syn form d. Although both d and c possess a stabilising intramolecular hydrogen bond, the most stable form is c by around 7 kJ mol −1 . Experimental pK a values were regressed onto bond lengths i-viii (Fig. 3b) of the triketo or keto-enol fragment of tautomers a-d and the fit was assessed using r 2 . For all tautomers a-d, there is a significant improvement in r 2 when the set is split into two subsets (r 2 generally 0.9 or above), with one group containing tkn derivatives and the other containing tkc substituted compounds. The slope for the tkn series is consistently 22% larger (i.e. steeper) than that of the tkc derivatives. We can interpret this steeper gradient as the resonance electron-withdrawing effect of the 2-NO 2 substituent heightening the para-substituent's electronic effect on dissociation propensity. The heightened acidity of the tkn compounds is also likely to be linked to the marked difference in geometry between the two subsets. For the tkc series, the exocarbonyl group is almost co-planar with the phenyl ring, whereas for the tkn series, the exo carbonyl is co-planar with the keto-enol moiety. In the latter orientation (of the tkn series), the orbital overlap allowing hydroxyl oxygen lone pair delocalisation across the keto-enol and exo-keto group is possible. It may be asserted that this increased conjugative effect would result in less delocalization between O and H atoms, a longer, weaker O-H bond and greater propensity for dissociation. The bond lengths of the enol anti-conformer b exhibit the most strongly correlated relationships with pK a values (see Supplementary Tables 2-5). With the exception of O-H(i) and the exocyclic C=O(vii) bond lengths, all pairs of subsets tkn and tkc exhibit r 2 values above 0.90 (q 2 > 0.9 and RMSEE ~0.2). This is an interesting result, considering that b is not the most stable tautomer according to the ranking at B3LYP/6-311 G(d,p)/ CPCM. It may be asserted that the emergence of stronger relationships between geometric features (bond lengths) and pK a using the anti keto-enol tautomer is indicative of its prevalence in solution. A thorough analysis using explicit solvation to explore this hypothesis is beyond the scope of this work. However, preference for this conformation could be linked to its increased propensity for dimerization and H-bonding to solvent molecules. For both subsets, the trend in the bond variation of O-H (i), C-O (ii) and C=C (iii) with pK o is such that more acidic compounds have longer O-H and C=C bonds but shorter C-O distances. These observations therefore fit with the intuition that a longer, weaker O-H bond should exhibit an increased propensity for cleavage. Conversely, bonds C-C (iv) and C=O (v) are found to show opposing trends between each series (Fig. 3b). The aim of this work is to derive a generally applicable model for compounds containing the diketone fragment. Therefore, we deemed it important to understand this disparity in C-C (iv) and C=O (v) bond length variation. To this end, we performed an interacting quantum atoms (IQA) analysis to partition the interaction energy between pairwise atoms A and B into V xc (A, B) (exchange-correlation) and V cl (A,B) (electrostatics). For further methodological and theoretical details of this approach see the Methods section. By taking V xc (A,B) as our dependent variable in place of bond distances, we can look at how the extent of delocalization of electrons between two topological atoms A and B changes with pK a . In doing so, we find analogous relationships between V xc (A,B) of bonds i-v and pK a values. Longer bonds exhibit less negative V xc (A,B) values (i.e. there is less delocalization), and vice 6 of the SI. b The trend in bond length variation and exchange-correlation (V xc ) energy of bonding interactions for tkn1-tkn4 is consistent with delocalization of electrons across the whole endocyclic keto-enol fragment. Conversely, the variation in bond lengths for tkc1-tkc6, as well as the increased co-planarity of the keto-enol group, is indicative that there is more conjugation with the exo-carbonyl. Supplementary Table 7 of the SI lists bond lengths i-v and pK a values for the b tautomer. versa (Fig. 3b). The trend in V xc (A,B) for bonds i-v across the keto-enol fragment of the tkn series is consistent with hydroxyl oxygen lone pair delocalization across the whole keto-enol fragment, akin to the resonance forms shown in Fig. 1a(ii). Conversely, for the tkc series this delocalization effect is not reflected in the distance variation of iv and v. Further discussion pertaining to the origin of the difference in bond and V xc variation with pK a between subsets can be found in Supplementary Methods. Overall, the discrepancy in AIBL-pK a trends with substituent type (Supplementary Note 2) suggests that, in the search for a bond that has a relationship with pK a over a wide variety of substituent patterns/types, it is logical to look to the enolic hydroxyl group, i.e. O-H (i), C-O (ii) and C=C (iii). Due to the prevalence of well-correlated relationships between bonding distances and pK a for the keto-enol anti-conformation for tkn1-tkn4 and tkc1-tkc6, this tautomeric form was used for all subsequent analysis on the remaining dataset. The bonds that are under investigation are those of the keto-enol fragment (i-v in Fig. 3b), which are common to all 1,3-CHD and 1,3-CPD compounds of the dataset. Selection of these specific bond lengths therefore allows us to construct one generally applicable model, rather than assembling many models for more specific subregions of chemical space. Single bond length models. Our dataset of 71 compounds (Supplementary Table 1) consists of 46 triketones and diketones from Syngenta, plus an additional 9 diketones and 2 triketones measured for the purpose of this work (experimental details can be found in Supplementary Methods in the SI). A further 8 pK a values for Alloxydim analogues were also obtained from the literature (Supplementary Table 1). Due to a discrepancy between predicted and literature values, samples were procured and pK a values were re-measured for 7 of these 8 compounds. Literature values for 6 Tetracycline derivatives were also included. The full set was split into 70% training and 30% test set, i.e. 49:22 training to test set. Table 1 lists internal, cross-validation and external validation statistics of each single bond length regression model (i.e. the typical AIBL approach). The values listed in Table 1 are found using a reduced training set, due to the removal of two outliers, dk29 and tk3. The reason for the removal of these compounds will be discussed in the next section. The most active bond, i.e. the model exhibiting the highest r 2 and lowest RMSEE is the C-O (ii) bond (0.72 and 0.57, respectively). We note that these values are somewhat less impressive than the threshold values used to mark the presence of an active bond in our other case studies (~0.90 for r 2 and ~0.3 for RMSEE). This decrease in goodness of fit can be attributed to the higher structural diversity of the set: the model covers 5-and 6-membered rings, compounds with substitution at the 2, 4 and 6 position of the 1,3-CHD fragment and compounds containing more than one ionizable group. Nonetheless Outliers. Two species were found to have residual errors exceeding 1.5 log units for 4 out of 5 bonds. One outlier is dk29, a 1,3-CPD derivative with a CH 2 -2-pyridyl group at the 4-position. The pK a value of 5.78 listed for this species was identified as the pK a for dissociation of the 2-pyridyl group, rather than the ketoenol fragment (pyridine itself has a pK a of 5.23). The other incongruous data point corresponds to tk3, which has a fourth keto group at the 5-position of the 1,3-CHD ring, a feature that is also present in compounds tk1 and tk4. The C-O bond distances of these three compounds sit below the trend line for the rest of the set, with an r 2 value of 1 for a linear fit, i.e., compounds with the 5-C=O structural motif in common form their own highcorrelation subset. More accurate predictions for compounds such as tk1 (error = +0.92) could therefore be made using the equation of this line as a new model, rather than the original C-O model. Both compounds were removed from subsequent analysis. Other regression approaches. Table 2 shows the 7-fold CV and external validation statistics for optimal models. These were derived using PLS (4 bonds), RFR ( 8, the full list of statistics for each model is shown in Supplementary Tables 9-13 and predictions are shown in Table 3). The optimal model for each method was then used to predict test set pK a values. Overall, all optimal models for each method include C-O as an input feature. The lowest 7-fold CV MAE and RMSEE correspond to the GPR model using a radial basis function kernel, which uses C-O, C-C and C=O as input features (MAE = 0.30, RMSEE = 0.39). However, this same GPR model also delivers the least accurate predictions for the 22 compounds of the external test set with an RMSEP of 0.59 and a MAE of 0.43, possibly indicative of overfitting to the training set data. Overall, SVR[RBF] using C-O, C-C and C=O returns the lowest MAE and RMSEP for the test set (0.29 and 0.36, respectively) and is consistent in its accuracy (s.d. = 0.22). However, PLS using C-O, C=C, C-C and C=O also performs similarly well (MAE = 0.31, RMSEP = 0.36) and exhibits the lowest standard deviation of absolute errors (s.d. = 0.19). There is one consistently large error across every model, corresponding to the predicted value for tk1. This compound shows an average error across all models of −1.21, with the lowest error exhibited by the PLS model (−0.72) and the largest for GPR[RBF] (−1.60). This compound was previously identified as belonging to a new subset of 5-C=O (Upper) Statistics for the single bond length models obtained via ordinary least squares regression. The row labelled "slope" features a "+" sign for a positive slope (i.e. pK a increases with increasing bond distance), and a "−" sign to denote a negative slope (i.e. pK a decreases with increasing bond distance). The squared correlation coefficient was not significant enough ("×") to assign a slope direction for iiii, iv and v. containing compounds, along with tk3 and tk4 for the C-O model, and may therefore be considered to be on the edge region of the domain of applicability for the model. The comparable accuracy of the single bond length C-O model for the test set, with respect to more complex regression methods using more input features is a remarkable result, given the simplicity of the approach. This result also validates our previous work, in which models using multiple input features were deemed unnecessary given the strength of the correlation for individual bond distances. ## Marvin. A comparison between error metrics for all models shows significant improvement compared with Marvin (Figs. 4b, c), either with or without consideration of tautomer/resonance. Furthermore, AIBL provides predicted values that correctly suggest the dominant microstate at pH 7 is the enolate, i.e. the ionized form. After tautomer enumeration and selection, Marvin's pK a values predict that 15 out of 22 compounds would be >50% unionized at this pH. However, this result is reduced to only two incorrectly assigned microstates when the keto-enol form is used explicitly. All experimental and predicted values can be found in Table 3. Correction of experimental value for Profoxydim. Experimental pK a data were initially procured from literature sources for the series of "dim" herbicides used in this work (Supplementary Note 1). Upon performing the fits for the single bond length models, the residual error for Profoxydim (Fig. 4d) using the literature pK a value of 5.91 was found to be anomalously high, at +1.30 units. Marvin predicts the pK a of the enolic hydroxy group to be 5.44, i.e. very close to this experimental value. Due to the excellent accuracy observed for species o1-o7 (residuals < 0.50), we decided to re-measure all pK a values. Seven of the eight compounds (all except Clethodim) were procured and re-measured using the UV-metric method (see Supplementary Methods for details). Excellent agreement was found between old and new values for all compounds but Profoxydim, for which a value of 4.82 was found. This new value lies only 0.22 units from our original prediction (4.61), yet it lies ~1.10 log units from the literature value. Therefore, we demonstrate the power of the AIBL approach to check internal consistency of pK a values for a given congeneric series. Structures and predictions for all dim herbicides can be found in Supplementary Fig. 2 of the SI. Tetracyclines. Aside from tautomerism, one of the more complex issues in the field of pK a prediction is the estimation of values for multiprotic compounds. Two of the species of our dataset contain a secondary ionizable group (dk26 and dk29, 2-pyridyl, pK a = ~5). In recent work we have demonstrated that prediction for a specific ionizable group may be performed by using the relevant microstate to the dissociation of interest. Therefore, in the case of dk26 and dk29, we performed all calculations on the cationic form of the 2-pyridyl group. To showcase the applicability of the AIBL model derived here in the context of larger multiprotic compounds, 6 tetracycline derivatives were included. For the correct microstate (the neutral state) of each species the most stable form is analogous to the keto-enol syn c conformation. The anti-conformation was constructed by manual rotation of the C 2 -C 1 -O 9 -H 10 (Fig. 3b) torsional angle from this form. For tet1, tet3, tet5 and tet6 of the training set, residual errors from the C-O model are below 0.1 log unit in all cases. For the test set compounds, predictions for tet2 and tet4 also lie within 0.1 log units. Use of Marvin with consideration of tautomers on this occasion identifies the keto-enol state as the relevant tautomeric form, delivering predictions of 2.83, 2.63, 2.55, 2.92, 2.84 and 2.51, for tet1-tet6, respectively, whereas experimental values are 3.35, 3.48, 3.25, 3.50, 3.53 and 3.30, respectively. Therefore, despite making the prediction using the correct tautomer, there is a distinct prediction bias towards higher acidity for the enolic hydroxy group for these compounds. Structures and predictions for tetracyclines can be found in Supplementary Fig. 3 of the SI. Future application of AIBL. The poorer performance of Marvin, as illustrated by Figs. 4b, c, can most likely be partly attributed to a lack of coverage of this type of compound (cyclic 1,3-diketones) in their training dataset. The predicted preference of the diketo state of many test compounds can also likely be attributed to the lack of knowledge on relative tautomeric stability, as previously pointed out by Connolly. The results in Fig. 4 illustrate the excellent performance of the C-O AIBL-pK a model in predicting The "Marvin" column corresponds to statistics for predictions made without considering tautomers/resonance (without parentheses), and the values in parentheses correspond to the predictions made with consideration of tautomers/resonance. The "features used" row lists the combination of features that minimized the RMSEE of the training set for each method. These features were subsequently used in the model used to predict for test set compounds. The row labelled "hyperparameters" lists the values obtained through minimization of RMSEE of the training set during 7-fold cross-validation (RFR and SVR). For PLS the number of latent variables (LV) was varied up to the number of features and the final number chosen on the basis of minimizing the RMSEE of the training set, which is also shown. For the GPR model, feature selection as carried out using 7-fold validation of each combination/subset of features using the training set and 100 restarts were used to locate the global maximum log likelihood of the y-values. The MAE, RMSEP, standard deviation of absolute errors (s.d.) and r 2 of observed vs predicted values are shown for the test set. the pK a variation across the series. Furthermore, we show that the accuracy is such that we can correct experimental values. We assert that a powerful future application of the AIBL approach is a method of fleshing out areas of chemical space that are sparse in the experimental pK a databases of empirical predictors, such as Marvin. Once a model has been set up with existing experimental data, hypothetical compounds with a variety of substituents can be assembled and their pK a values predicted and added to the training set. Therefore, the empirical approach is calibrated using the highly accurate AIBL approach, whilst still maintaining userfriendly computational speed. We have shown bonding distances to be an intuitive and powerful descriptor of ionization propensity for much of 1,3-CHD and 1,3-CPD space. Due to the use of quantum chemically derived descriptors, the dominant tautomeric state is easily identified as the keto-enol form, from which chemically meaningful relationships are derived; a longer O-H and a shorter C-O bond are generally indicative of a species with heightened acidity compared with the parent compound. A simple but accurate AIBL-pK a method is proposed and validated; good results are derived using only simple linear regression of pK a onto C-O bond distances, which is shown to be applicable to a diverse array of analogues. For the test set, this simple model is found to outperform regression using various approaches and multiple bond lengths relevant to the dissociation at the keto-enol ionizable group. Furthermore, the method is applicable to multiprotic compounds, which along with tautomerizable species, represent one of the most challenging areas of pK a prediction. All of the models developed showed superior accuracy compared with the industry standard, represented by the program Marvin, for which the user must have prior knowledge of the dominant tautomeric form. At present, there is still a time/cost barrier to feasible use of quantum chemical QSPR methods in large scale screening studies. However, this work suggests that the inclusion of some description of electrons and their distribution (via a highly populated geometric representation of molecules), provides a significant advantage in terms of prediction accuracy over an approach (Marvin) that does not describe a compound quantum mechanically. Thanks to AIBL predictions, we also amend the literature experimental value for Profoxydim, which is corrected from a previous value 5.91 to a new value of 4.82. Based on the Fig. 4 Performance of each regression method tested in this work using bond lengths as input features compared with results obtained using Marvin. a The 7-fold RMSEE for each model tested, for each method, where "Model ID" corresponds to one of 31 combinations of features out of the 5 bonds i-v chosen for consideration (see Supplementary Table 8 for the full list). The C-O, ii bond is used as a feature for the Model ID numbers shaded in blue. b Experimental pK a variation across the test set (dark blue), along with Marvin predictions using the diketo state with tautomer consideration turned on (blue), and using the keto-enol state with tautomer consideration turned off (magenta), as well as the AIBL-pK a C-O bond model (green). c Root-mean squared error of prediction for the test set (RMSEP, blue) and mean absolute error for the test set (MAE, green) for each method of prediction. Marvin predictions are removed for the plot shown in the inlay, so that AIBL models can be compared. d The structure of Profoxydim, for which the literature experimental pK a value (5.91) and Marvin's prediction (5.44, tautomer/resonance not considered, keto-enol form used) deviated significantly from our prediction. The new experimental value of 4.82, measured in this work matches our initial prediction more closely. work shown here, and on previous results, we propose that AIBL-pK a is applicable to any tautomerizable congener series, given that pK a data exist for model calibration. ## Methods Data. Structures and pK a values with references are given in Supplementary Table 1 for all compounds studied in this work. Equilibrium bond lengths for the most stable geometries identified are listed in Supplementary Table 7. The pK a data for the compounds investigated in this work have been procured from various sources. Sixteen triketones, labelled tk-1 to tk-15, tk18 and tk19 were procured from the Syngenta and are analogues of the herbicide Mesotrione. A further 20 diketone compounds were procured from Syngenta, which are labelled as dk-1 to dk-12 and dk22 to dk29. These values were obtained using the UV-vis metric approach with a Sirius T3 instrument at standard conditions (see Supplementary Methods in the SI for more details). A set of 10 compounds of triketone (tk) type labelled in as tkn1-tkn4 and tkc1-tkc6 were taken from the work 32 of Lee et al. Samples of 11 diketones (dk), labelled dk-13 to dk-21, tk16 and tk17 have been procured and measured for the purpose of this work, using the potentiometric method with a Sirius T3 instrument at standard conditions. Finally, literature values were procured for 8 "dim" herbicides Alloxydim, Cycloxydim, Butroxydim, Clethodim, Sethoxydim, Tepraloxydim, Tralkoxydim and Profoxydim were procured, samples were purchased for all except Clethodim (due to unavailability) and pK a measurements were taken using the same apparatus and experimental procedure as described above and in Supplementary Methods. Literature values for 6 tetracycline derivatives (tet1-tet6) were obtained from literature sources. Quantum chemical calculations. An ensemble of 15 conformers were generated for each tautomeric form of each compound tkn1-tkn4 and tkc1-tkc6 using the conformer generator plug-in within the Marvin program. Geometry optimization and frequency calculations were then performed using B3LYP/6-311G(d,p) with CPCM implicit solvation for each conformer of every ensemble using GAUS-SIAN09 33 . Conformers were ranked according to internal energy and the most stable species was taken as the global minimum. For the anti and syn conformers of the keto-enol state, an input geometry for the higher energy anti-conformation was manually generated by rotating the orientation of the O-H bond of the syn conformer by 180°. This process of generating the keto-enol anti state 15,16, was repeated for the remaining 61 species. IQA calculations. The extent of electronic delocalization between two atoms can be calculated within the context of a topological energy decomposition framework called interacting quantum atoms (IQA). Originating from the quantum theory of atoms in molecules 33 (QTAIM), IQA has been used to analyze a variety of chemical phenomena . By decomposing the total energy of a system into intra-and interatomic terms, we derive the exchange-correlation potential energy V xc , which is the sum of the exchange energy V x , and the correlation energy V c . The former term usually dominates and denotes the Fock-Dirac exchange, which describes the ever-reducing probability of finding two electrons of the same spin close to one another (i.e. the Fermi hole). The latter term is associated with the Coulomb hole and the electrostatic repulsion between electrons. The absolute value of V xc evaluated between two atoms can be taken as the extent delocalization of electrons between them and so can be interpreted as a measure of covalency. These values were obtained by the AIMAll program 39 (version 14), using DFT-compatible IQA partitioning, and using default parameters on wavefunctions obtained at the B3LYP/6-311G(d,p) level using CPCM. Models. For more details of regression methods implemented in this work see Supplementary Methods in the SI. Model training and error evaluation were performed using scikit-learn 40 . Initially, ordinary least squares (OLS) regression of single bond distances and pK a , and validation was performed using r 2 and 7-fold CV RMSEE and MAE to assess the linear relationships between bond lengths and pK a . A random 70:30 split of training set to external test set was then performed (i.e. training set = 49, test set = 22). We compared the results of using more than one bond length of the keto-enol fragment using support vector regression (SVR) with a linear and radial basis function (RBF) kernel, random forest regression (RFR), partial least squares (PLS) and Gaussian process regression (GPR) with an RBF kernel. We also compared our test set prediction errors results with those obtained using the program Marvin. Each model was evaluated using error-based metrics, mean absolute error (MAE), standard deviation of absolute errors (s.d.), root-mean-squared error (RMSEP) and the r 2 of observed vs predicted values. An overview of the AIBL workflow used in the context of cyclic β-diketones is shown in Fig. 1c. The optimal hyperparameters for the SVR models, C, ε (and γ for the RBF kernel) and RFR (number of estimators n est , maximum depth) were found in each case by applying a grid search (GridSearchCV in scikit-learn). The final hyperparameter values were chosen to minimize a 7-fold cross-validation RMSEE. The GPR model was implemented in python using the GPR package called George. The squared exponential (SE) kernel, or RBF, was used to set up the GPR models with a unique length scale (hyperparameter) for each dimension, also known as the automatic relevance determination kernel of the SE-ARD, SE ARD x; x 0 ð Þ¼exp 1 2 The hyperparameters for this kernel were found by maximizing the loglikelihood function using the training set. The implementation for this used the gradient descent BFGS algorithm (implemented by scipy) on the negative gradient of the log-likelihood function (therefore finding the maximum of the function). As there can be many local maxima, the optimizer was restarted with random weights 100 times in an attempt to find the global maximum.

chemsum

{"title": "Aqueous pKa prediction for tautomerizable compounds using equilibrium bond lengths", "journal": "Nature Communications Chemistry"}

deciphering_the_role_of_anions_and_secondary_coordination_sphere_in_tuning_anisotropy_in_dy(iii)_air

## Abstract: Precise control of the crystal field and symmetry around the paramagnetic spin centre has recently facilitated the engineering of high-temperature single-ion magnets (SIMs), the smallest possible units for future spin-based devices. In the present work, we report a series of air-stable seven coordinate Dy(III) SIMs {[L2Dy(H2O)5][X]3•L2•n(H2O), n = 0, X = Cl (1), n = 1, X = Br (2), I (3)} possessing pseudo-D5h symmetry or pentagonal bipyramidal coordination geometry with high anisotropy energy barrier (Ueff) and blocking temperature (TB). While the strong axial coordination from the sterically encumbered phosphonamide, t BuPO(NH i Pr)2 (L), increases the overall anisotropy of the system, the presence of high symmetry significantly quenches quantum tunnelling of magnetization, which is the prominent deactivating factor encountered in SIMs. Although the local coordination geometry and the symmetry around the Dy(III) in all the three complexes are similar and display only slight deviations, the variation of halide anions in the secondary coordination sphere which is hydrogen-bonded to the coordinated equatorial water molecules, show subtle alteration in the magnetic properties. The energy barrier (Ueff) and the blocking temperature (TB) decrease in the order 3 > 2 > 1 with the change of anions from larger iodide to smaller strongly hydrogen-bonded chloride in the secondary coordination sphere. Ab initio CASSCF/RASSI-SO/SINGLE_ANISO calculations further provide deeper insights into the dynamics of magnetic relaxation in addition to the role of the secondary coordination sphere in modulating the anisotropy of the D5h systems, using diverse models. Thus, in addition to the importance of the crystal field and the symmetry to obtain high-temperature SIMs, this study also probes the significance of the secondary coordination sphere that can be tailored to accomplish novel SIMs. ## Introduction Single-molecule magnets (SMMs) are superparamagnetic molecules that behave as molecular-level classical magnets at low temperatures. This scripts them as potential candidates for fabrication of next-generation high-density data storage devices. Besides, they are also recognized as prospective candidates for application in future molecular spintronics and quantum computing due to the observance of phenomena such as quantum tunneling of magnetization (QTM) and quantum phase interference. However, these properties are witnessed only at very low temperatures in most SMMs, thus rendering them unfit for technological applications. While the first slow relaxation of magnetization in a molecular complex was observed in a 'Mn12' cluster, the report on high energy barrier (Ueff) in the double-decker complexes, [Pc2Tb] -[TBA] + , by Ishikawa et al. in 2003 shifted the attention to lanthanide ions for designing SMMs with high Ueff. In the last decade, several groups have made remarkable efforts to decrypt the factors that can aid the synthesis of SMMs with higher blocking energy barriers (Ueff) and blocking temperatures (TB). Particularly lanthanide ions having electronic configurations greater than 4f 7 such as Dy(III), Er(III), and Tb(III) are more attractive due to large spin-orbit (SO) coupling and larger magnetic moments (as the f-orbitals are deeply buried and shielded, they do not significantly interact with the crystal field (CF)). [1d, 6] However, the presence of significant QTM in the case of 4f complexes between the ground state doublets significantly inhibits the slow relaxation of magnetization. QTM has been quenched by either incorporating a radical that induces a strong exchange coupling or a 3d metal ion in the complex. QTM has also been quenched either by maintaining a strong axiality and/or a higher-order symmetry. [1d, 6c, 6f, 8] In 2011, Reinhart and Long put forward an idea based on the electrostatic model that proposed that a specific CF could be designed to enhance the anisotropic charge distribution of 4f ions. [6a] While strong axial coordination makes oblate ions such as Dy(III), Tb(III), etc., more anisotropic, the reverse is true for ions with prolate electronic charge distribution. Thus, a better understanding of the elements that play a decisive role in realizing the slow relaxation dynamics such as the effect of crystal field (CF) and the symmetry in addition to the synthetic efforts, has resulted in SMMs with Ueff and TB values as high as 1541 cm -1 and 80 K, respectively in a dysprosocenium complex. [8d] While maintaining a strict axial symmetry seems to readily impart a very high anisotropic barrier in Ln(III) SIMs as recently reported in some interesting D4h, D5h, D6h symmetric systems, the nature of axial ligand and geometry around the central Ln(III) ion seems to be very important to obtain higher blocking temperatures. [6d, 6f, 8a, 9] While D4h and D6h systems appear to impart very high Ueff, they suffer from significant QTM at zero-field as observed in the magnetic hysteresis loop measurements. [9d-f] Thus, the high Ueff does not always translated in high TB. Moreover, apart from possessing high Ueff and TB, these molecules also need to retain additional properties such as stability, solution processability and sublimability for fabrication. Therefore, it becomes necessary to outline design strategies to synthesize air stable SIMs/SMMs. In the present work, we report a series of air-stable pseudo-D5h symmetric Dy(III) SIMs, {[L2Dy(H2O)5][X]3•L2•n(H2O), n = 0, X = Cl (1), n = 1, X = Br (2), I (3)}, that exhibit high anisotropy energy barrier (Ueff) and blocking temperature (TB). These SIMs have been rationally designed from a sterically encumbered phosphonamide t BuPO(NH i Pr)2. This study particularly unravels the role of halide ions in the secondary coordination sphere in fine-tuning the magnetic properties. The effect of coordinated halide anions on the SIM properties of 3d and 4f ions has been recently reported. The magnetic properties of complex 3 has been reported by our group in a previous communication. [6f] These SIMs possesses some of the highest Ueff and TB values for any air-stable 3d or 4f systems. Additional ab initio calculations performed disclose the role of the phosphonamide ligand, secondary coordination sphere, and the higher-order symmetry in the realization of unique properties exhibited by these complexes. ## Synthetic aspects and molecular structures The sterically bulky phosphonic diamide, [ t BuPO(NH i Pr)2] (L), was derived from the reaction of tert-butylphosphonic dichloride, t BuPOCl2, and excess isopropyl amine employing a literature procedure. The pentagonal-bipyramidal dysprosium complexes, {[L2Dy(H2O around 1100 cm -1 for all the complexes due to the presence of two types of P=O bonds, one in the lattice and the other coordinated to the metal ion. Single crystal X-ray diffraction analyses reveal that all the three complexes have a similar core structure around the central Dy(III) ion, whilst some differences are found in the arrangement of the anions and the neutral lattice ligands in the secondary coordination sphere. Complex 1, [L2Dy(H2O)5][Cl]3•L2 , crystallizes in the orthorhombic space group Pbca. The asymmetric part of the unit cell contains one seven coordinate dysprosium ion coordinated to two phosphonic amide ligands and five water molecules (Figure 1). In addition, there are two phosphonic amide ligands present in the lattice along with three chloride anions, which balance the overall charge. The coordination sites in the equatorial plane of the dysprosium ion are occupied by the water molecules and the axial sites are coordinated by the phosphoryl oxygen atom of the amide ligand. Analysis of the {DyO7} core ion using SHAPE 2.1 shows the least deviation (0.492) from the D5h symmetry suggesting that Dy(III) ion occupies a distorted pentagonal bipyramidal coordination environment (Figure 2 and Table S2). The axial Dy-O(P) distances (2.203(1) and 2.213(1) ) are shorter than the equatorial Dy-O(aqua) distances (2.335(2) -2.407(2) ). This, in addition to the near-linear trans O(P)-Dy-O(P) angle (172.19(6) o ), renders a virtual quasi-two coordinate coordination environment to the Dy(III) ion, a highly sought after geometry in the case of 4f-SMMs. The two Dy-O-P angles are 169.53(1) o and 166.89(1) o . The hydrogen atoms of the water molecules coordinated to the Dy(III) ion are hydrogen-bonded to three chloride anions and two neutral phosphonic diamide ligands giving a star-like H-bonded architecture (Figure 1). The closest Dy(III)•••Dy(III) distance in the lattice of 9.815 is largely aided by the presence of two uncoordinated phosphonic diamide ligands and the three chloride ions present in the lattice per formula unit. The chloride ions are further involved in weak Hbonding with the amide protons giving rise to a two-dimensional network of mononuclear dysprosium complexes (Figure S2). Complexes 2 and 3 crystallize in the triclinic space group P1 ̅ and are isomorphous. While the primary coordination sphere around the Dy(III) ion in 2 and 3 have similar core structural features as in complex 1 (Figure 1, 2, and Table 1), they differ in the arrangement of the non-coordinated three lattice anions and two phosphonic amide ligands in the second coordination sphere. Besides, one molecule of water is present in the lattice. The trans O(P)-Dy-O(P) angle is more linear in 2 (177.68(8) o ) and 3 (175.14(9) o ) compared to 1. However, the major difference appears in the bent P-O-Dy angle, where the bent angle is more linear in the case of 1. This also leads to the decrease of the distance of phosphorous atom from the mean {DyO5} equatorial plane in 2 and 3 as compared to 1. With increasing anion size, the average Dy(III)•••X and X•••O(aqua) distances increase (X = halide). This also results in a considerable increase of nearest Dy(III)•••Dy(III) distance in the crystal lattices of 2 (10.459 ) and 3 (10.819 ). Further, the increase in the ionic size of the lattice anions leads to the weakening of X•••H-O hydrogen bonds in 2 and 3 (see supporting information). Analysis of the {DyO7} core with the standard symmetry using SHAPE 2.1 suggests an almost ideal D5h symmetry with a deviation of 0.18 and 0.223 for 2 and 3, respectively (Table S2). The corresponding diamagnetic Y(III) complexes 4-6 also reveal similar core structural features. While complex 4 crystallizes in the monoclinic space group P21/c, complexes 5 and 6 crystallize in the triclinic space group P1 ̅ as in the case of their dysprosium analogues. ## Magnetic Studies The static and dynamic magnetic susceptibility measurements of 1-3 have been carried out using an MPMS-XL SQUID magnetometer. The direct current (dc) susceptibility measurements carried out on a polycrystalline sample under an applied magnetic field of 1000 Oe shows MT values of 14.09, 13.90, and 14.15 cm 3 K mol -1 at 300 K for 1-3, respectively, which is close to the estimated value of 14.18 cm 3 K mol −1 for an isolated Dy(III) ion (ground state = 6 H15/2) (Figure S9). The MT values of 1-3 remain almost constant with lowering of the temperature, but fall sharply near 10 K indicating a large energy separation among the low-lying Kramers doublets (KDs), indicating magnetic blocking. The field (H) dependent magnetization (M) curve for 1-3 shows a sinusoidal behavior (Figure S10), a signature of large anisotropy, with a steep increase in magnetization at the lower field before reaching ~ 5.0 μB at 7.0 T as seen in several hightemperature SMMs. [6f] Alternating current (ac) susceptibility measurements were carried out to unravel the slow relaxation dynamics of 1-3 at zero applied dc field between 0.1 and 1500 Hz at an oscillating ac field of 3.5 Oe. Clear frequency and temperature-dependent maxima in the out-of-phase signals were observed up to 36 ~ 40 K for 1-3 indicative of a very high thermal energy barrier. Maxima in the outof-phase component of the frequency-dependent ac susceptibility (M′′) signals were observed up to 36 K for 1 indicative of a very high thermal energy barrier (Figure 3). To extract the relaxation times, the ac susceptibilities were fitted with a generalized Debye model which shows a temperature-dependent regime at higher temperatures. A linear fit of the temperature-dependent relaxation times () at high temperatures to the Arrhenius law yields Ueff = 582 K and 0 = 1.43 x 10 -11 . However, complex 1 shows a linear behavior only until 30 K and deviates from linearity at lower temperatures indicating the presence of other competing relaxation processes. Thus the relaxation times extracted over the entire temperature range were treated considering the QTM, direct, Raman, and Orbach processes with the following expression: 𝜏 The best fit to eq. ( 1) for 1 yields an anisotropy barrier Ueff of 609 K with τ0 = 6.6 × 10 −12 s, relative to the Orbach process with further contribution from the Raman relaxation mechanism (C = 2.04 x 10 -4 s −1 K −n , n = 3.6). This indicates that the QTM is effectively quenched due to the high symmetry around the Dy(III) ion and strong axial CF. The application of dc fields shows only a slight effect on the energy barrier (Figure S11). Complex 2 also displays similar relaxation dynamics like complex 1, nevertheless, the maxima in the out-of-phase are shifted towards higher temperature (~ 39.0 K) while deviating from linearity below 30 K (Figure S15). The best fit to eq. ( 1) for 2 yields an anisotropy barrier Ueff of 640 K with 0 = 1.06 × 10 −11 s, relative to Orbach process with further contribution from the Raman relaxation pathway (C = 7.24 x 10 -7 s −1 K −n , n = 4.74) as observed in the case of complex 1. As in the previous case, the application of dc fields has only negligible effects on the energy barrier (Figure S16). A detailed comparative magnetic study of complex 3 along with the isomorphous Er(III) complex has been reported already in a previous communication. [6f] Thus, among the series, Ueff decreases in the order 3 > 2 > 1 with the change of anions from larger iodide to smaller chloride in the secondary coordination sphere. These results are interesting as we find that although the local coordination environment in 1-3 is same, the replacement of halide anions in the secondary coordination sphere results in higher Ueff in the order 3 > 2 > 1. This points out that the presence of higher negative charges in the equatorial position of oblate Ln(III) such as Dy(III) are detrimental to the effective Ueff. To further ascertain the effect of halide ions on the SIM properties of these Dy(III) complexes, additional magnetic measurements were carried out. For example, zero-field cooled (ZFC)field cooled (FC) variable temperature magnetization measurements were carried out to determine the blocking temperature of 1-3 (Figure 4 a-d). TB is defined as the maxima in the ZFC curve. [1a] As can be seen from Figure 4, 3 possess the highest blocking temperature among the series. While TB for 3 is 12.9 K (measured at a sweep rate of 2 K min -1 ), 2 and 1 possess TB of 11.6 K and 8.2 K, respectively. It is to be noted that the maxima in ZFC curves are highly dependent on the sweep rate (Figure 4d). Tirr for 1, 2, and 3 are 14.7 K, 13.2 K, and 9.2 K, respectively. To further compare the magnetic blocking in 1-3, magnetic hysteresis measurements were carried out. Similar to the effect of sweep rate for ZFC-FC experiments, the opening of the hysteresis loop and coercivity are highly dependent on the magnetic field sweep rate (Figure 4h). Among the series, 1 possesses significant tunneling around zero-field as against 2 and 3. Further, the coercivity in the case of 1 is the least. 1, 2, and 3 display the opening of the hysteresis loop until 16.0 K, 14.0 K, 9.0 K, respectively (at a field sweep rate of 20 Oe s -1 ). A comparative summary of the magnetic data is presented in Figure 5. These combined properties place these SIMs amongst the best performing SIMs that are stable to air and moisture. To further understand the relaxation dynamics, we prepared a 10 % diluted sample (2@Y and 3@Y) with the isomorphous Y(III) analogues. However, 2@Y and 3@Y display similar magnetic properties as 2 and 3, thanks to the bulky phosphonic diamide ligands that keep the magnetic centers far apart in the crystal lattice. No significant improvement in blocking temperature and hysteresis were observed for 2@Y and 3@Y (Figures S23-S28). ## Electronic structure calculations To comprehend the effect of counter anion on the magnetic properties, we have performed ab initio CASSCF/RASSI-SO/SINGLE_ANISO calculations on complexes 1, 2, and 3 using MOLCAS 8.2 programme package (see computational details). The chosen ab initio methodology was found to provide better guidance to obtain insights and understand the single-ion magnetism shown by lanthanide complexes. [9d, 16] The magnetic anisotropy axis of 1-3 is found to lie along the axial O-Dy-O bond, perpendicular to the ground state beta electron density [16a, 16b, 17] to minimize the electrostatic repulsions (Figure 6 S12- 14) and first excited KD. This enables negligible QTM/TA-QTM (TA = thermally activated) in the mechanism of magnetization relaxation (Figure 7). Furthermore, the angle of the anisotropy axis between ground and first excited states is found to be < 6. This indicates magnetization relaxation via other higher excited states (Table S12 -14). The eight ground KDs generated from the 6 H15/2 state for the three complexes span up to 902.0 (1), 957.7 (2), and 1028.4 K (3). The significant TA-QTM value at the second excited state reinforces the magnetization relaxation (|-15/2>→|-13/2>→|-1/2>→|+1/2>→|+13/2>→|15/2>, Figure 7). This results in large Ucal values of 606.3, 645.7, and 668.9 K for 1, 2, and 3, respectively, consistent with the estimated Ueff values (Table S12-14, Figure 7). , (here 𝐵 𝑘 𝑞 is the crystal field parameter and 𝑂 ̂𝑘 𝑞 is the Stevens operator respectively). The value of larger axial crystal field (CF) parameter (k = 2, 4, 6; q = 0) compared to non-axial (k = 2, 4, 6; q  0) suggests substantial axiality in all the three complexes (Table S15). The higher Ucal value of 3 compared to 1 and 2 correlates with the larger axial 𝐵 2 0 CF parameter. We have also estimated the (E2 and E1 is the spin free energy of the first and second excited state, respectively) which is considered as a figure of merit of axiality in Dy(III) SIM. The larger value of 2 and 3 compared to 1 suggests a larger axiality of the former compared to the latter (Table S16). The computed Loprop charges also explain the increasing axiality from 1→2→3. The Loprop charges on the equatorial water oxygens vary as 1 < 2 < 3, but the charge of the surrounding halide counter anions decreases in the order 1 > 2 > 3 with a more pronounced effect, rationalizing the trend in the Ucal/Ueff values (Table S17). To unravel whether the larger distance between Dy and iodide ions (av. Dy-halide distance is 5.04, 4.80 and 4.56 in 3, 2 and 1, respectively) offers a larger Ucal value in 3, we have performed a model calculation on 3-Cl where the three iodide ions in 3 have been replaced by chloride ions. The calculation on 3-Cl reveals enhancement of Ucal value to 693.4 K (Table S18) compared to 3, although a significant increase in the Loprop charges of chloride ions is observed (Table S17). This unveils the metal-halide distance rather than the nature of the halide ions controlling the magnetic anisotropy. On the other hand, in our earlier studies, we have established a structural 𝑅 parameter ) to correlate it with the estimated Ucal values in pseudo-D5h Dy(III) complexes. [16a] The 𝑅 value is found to be in the order of 3 > 2~1, which also explains the trend in the Ucal values (Table S19). Further, to analyze the effect of counter anions in magnetic anisotropy, we have removed the halide ions from the secondary coordination sphere of 1, 2, and 3 (model 1a, 2a, and 3a, respectively, see Figure 5 and S30-31). Calculation on these models yields enhancement in Ucal value to 1032.1, 1115.5, and 1079.5 K for 1a, 2a, and 3a, respectively (Table S20 -22). This is due to a reduction in the computed charge for the equatorial oxygen atoms (Table S23 -25), compared to the original structure. Quite interestingly, the computed charge of the axial oxygen atoms increases in 1a, 2a, and 3a compared to 1, 2, and 3 (Table S23 -25). Both these effects lead to the increase in the axial 𝐵 2 0 CF parameter (Table S26). For models, the mJ = |11/2> stabilized as the second excited state, contrary to 1, 2, and 3 (Table S20- 22) where mJ = |1/2> was stabilized as second excited state, leading to magnetization relaxation via 3 rd excited state (Table S20 -22). To investigate the effect of non-coordinating phosphonamide ligands and halide ions in Ucal value, the ligands have been removed from 1, 2, and 3, keeping the halide ions intact (model 1b, 2b, and 3b, respectively, see Figure S32 for beta electron density). The calculations yield a decrease in the Ucal value (Ucal = 793.9 (1b), 850.5 (2b) and 898.4 K (3b)) compared to 1a, 2a and 3a (the magnetization relaxation via second excited state, Table S27-29). In these models, equatorial oxygens were found to possess less charge as their hydrogen bonding solvates/halides were removed, placing their Ucal values larger than 1, 2, and 3 (Table S23 -25). More pronounced changes in the axial/equatorial oxygen charges and the computed crystal field parameters are observed in models 1a-3a compared to 1b-3b (Table S23 -25, S30). This suggests that the halide counter anions (equatorial non-bonding interaction with Dy(III)) play a dominant role in magnetic anisotropy compared to the phosphonamide ligands (H-bonding interaction with equatorial water molecules). In the next step, we have performed calculations on models 1c-3c where both halides and phosphonamide ligands were removed (see Figure 5 and S30-31). The ab initio calculations on these models reveal magnetization relaxation via third excited KD with the in the Ucal value to 1343.7 (1c), 1407.2 (2c), 1403.1 K (3c) (see Table S31-33) with respect to the original structures and other models constructed. Here the computed values are twice as large as compared to the original structure (2.22 times in 1, 2.18 times in 2, 2.09 times in 3), and this can be rationalized from the computed charges and crystal field parameters (Table S23-25, S34). The Ucal value of 2c is slightly higher compared to 3c due to the larger O(P)-Dy-O(P) angle found in 2 compared to 3 (Table 1). To obtain the absolute axial limit with these ligands, the equatorial water molecules from 1c-3c have been removed to build models 1d-3d (see Figure 5, S30-31 for beta electron density). For these models, calculations reveals magnetization relaxation via 6 th excited KD (mJ:|-15/2>→|-13/2>→|-11/2>→|-9/2>→|-7/2>→|-5/2>→|-3/2>→|+3/2>→|+5/2>→|+7/2>→|+9/2> →|+11/2>→|+13/2>→|+15/2>, Table S35) in 1d while 2d and 3d relaxes via 5 th excited KD (mJ:|-15/2>→|-13/2>→|-11/2>→|-9/2>→|-7/2>→|-5/2>→|+5/2>→|+7/2>→|+9/2>→|+11/2> →|+13/2>→|+15/2>, Table S36 -37).This leads to the Ucal value of 3108.7, 3172.8, and 3188.2 K for 1d, 2d, and 3d, respectively. The water molecules in 1-3 offer significant equatorial ligation which hinders them from achieving a very large blocking barrier. To decrease the equatorial ligand field in 3, we have carved out model 3b-acetone where the five equatorial water molecules in 3b were replaced by acetone (Figure S33 and Appendix S1 for optimized coordinates). The ab initio calculations on 3b-acetone reveal a decrease in the Ucal value to 491.4 K compared to 3 (Table S39). This is due to the loss of planarity in the optimized structure 3b-acetone (Figure S33). Again, the axial O(P)-Dy-O(P) angle also reduces to 171.48 compared to 3b (175.16). Furthermore, we have also performed calculations on 3b-THF (carved out from 3b with replacing the equatorial -OH2 group by THF, Figure S34, Appendix S2) to reduce the equatorial ligation. But the calculations on these models reveal a decrease in the Ucal value to 427.1 K due to a decrease of axial O(P)-Dy-O(P) angle (172.22) and loss of equatorial planarity in optimized 3b-THF compared to 3b (Table S40). The unsuccessful attempts to enhance the blocking barrier with model 3b-acetone and 3b-THF suggests that the nature of the equatorial donor atom is important compared to its ligand environment in dictating the magnetic anisotropy. Keeping this in mind, we have performed ab initio calculation on model 3-H2S by replacing the five water molecules of 3 with hydrogen sulfide molecules. The optimized structure of 3-H2S leads to a decrease in the axial O(P)-Dy-O(P) angle to 168.64 and at the same time, the equatorial planarity in 3-H2S is also lost compared to 3 (Figure S35 and Appendix S3). The calculations on 3-H2S unveil an increase in the Ucal value to 1325.9 K compared to 3 with the magnetic relaxation via second excited KD (Table S41). This is due to the reduction of the computed charge of the equatorial sulphur atoms compared to the oxygen atom of the water molecules (Table S42). ## Conclusions In summary, complexes 1-3 present a series of highperformance pseudo-D5h Dy(III) SIMs that are isostructural with the same set of ligand systems in the first coordination sphere. Whereas the high anisotropic barrier mostly results from the strong axial phosphonamide ligands and the higher symmetry around the central Dy(III) ions, the nature of halide ions in the secondary coordination sphere that is hydrogen-bonded to the equatorial coordinated aqua ligand have a profound effect on the relaxation dynamics of the SIMs. The substitution of a larger iodide anion to the smaller strongly hydrogen-bonded chloride anion in the secondary coordination leads to a decrease in the effective energy barrier (Ueff) and blocking temperature (TB) in the order 3 > 2 > 1. Thus, in addition to the first coordination sphere, the secondary coordination sphere can also generate a subsidiary difference in the magnetic properties of SIMs. Ab initio calculations aid the understanding of the effects and role of the secondary coordination sphere in modulating the anisotropy of the D5h systems. In brief, this study presents anion fine-tuning of a fascinating series of air-stable Dy(III) SIMs and highlights the significance of careful selection of secondary coordination sphere anions that can have a subtle effect on the overall performance of molecular magnets. Studies in order to further fine-tune and understand the effect of other factors are currently underway in our laboratory. ## Experimental Section Materials, instruments, and methods: All the new compounds reported in this study are stable towards air and moisture and hence all the operations were carried out under normal aerobic conditions. Solvents were distilled before use. The phosphonic diamide ligand, t BuPO(NH i Pr)2 (L) was synthesized using a previously reported procedure. Fourier transform infrared spectra were obtained on a Perkin Elmer Spectrum One FT-IR spectrometer as KBr diluted discs. Microanalyses were performed on a ThermoFinnigan (FLASH EA 1112) microanalyzer. The metal content in the samples was measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES). The samples were digested in nitric acid and diluted with distilled water. The magnetic measurements were carried out on a Quantum Design MPMS-XL SQUID magnetometer equipped with a 7 T magnet in the temperature range 2-300 K using polycrystalline powder samples. The data were corrected for any background diamagnetic contribution using Pascal's constants. Alternating current (ac) susceptibility measurements were carried out in an oscillating ac field of 3.5 Oe and frequencies ranging from 0.1 to 1500 Hz. Hydrated halides salts of Dy(III) and Y(III) were prepared from the corresponding oxides (Alfa Aesar) using suitable mineral acids. X-ray crystallography: Suitable single crystals of the complexes, obtained from slow evaporation of the solvent from reaction mixture, were mounted on a Rigaku Saturn 724+ ccd diffractometer for unit cell determination and three-dimensional intensity data collection. Data integration and indexing were carried out using CrystalClear and CrystalStructure. The structures were solved using direct methods (SIR-97). Structure refinement and geometrical calculations were carried out using programs in the WinGX module and Olex2 v1.2. The final structure refinement was carried out using full least square methods on F 2 using SHELXL-2014. Details of crystal data and structure refinement are reported in Table S1 Computational Details: All the ab initio CASSCF/RASSI-SO/SINGLE_ANISO multireference calculation has been performed on the X-ray crystal structures of 1, 2, and 3 with MOLCAS 8.2 programme package. The basis sets for our calculations were taken from the ANO-RCC library implemented in the package. The following basis set was used throughout the calculations: Dy: [Dy.ANO-RCC... Hamiltonian was used to account the scalar relativistic effect. The disk space was reduced by the Cholesky decomposition technique . The active space for our CASSCF calculation includes nine electrons in seven 4f orbitals, i.e., CAS(9,7). Using this active space, we have computed the energies of 21 sextets which are derived from 6 H15/2 ground state of Dy(III). In the next step, we have mixed 21 sextets to obtain spin-orbit states. In the final step, we have employed SINGLE_ANISO module of MOLCAS to estimate the g tensor, QTM, and blocking barrier. Geometry optimization of models 3-H2S, 3-acetone, and 3-THF has been carried out with UB3LY functional in Gaussian09 programme package. The Dy(III) ion has been replaced by Y(III) during optimization. We have used SDD ECP (electron core potential) for Y, and I, Ahlrichs TZVP basis set for S, N and O and SVP basis set C and H. General Procedure for the synthesis of {[L2Ln(H2O)5][X]3•L2•nH2O}. To a solution of t BuPO(NH i Pr)2 (330 mg, 1.5 mmol) in a solvent mixture of dichloromethane and benzene (4:1 v/v, 30 mL) was added LnX3.xH2O (0.25 mmol). The reaction mixture was stirred at 60 o C for 1 h and was cooled down to room temperature. The solution was allowed to stand for some time and was then filtered. The clear filtrate obtained was then kept for crystallization at ambient aerobic conditions. Single crystals were obtained by the slow evaporation of the solvent mixture within a week. The crystals were carefully washed with toluene. A few mL of methanol was necessary for the better dissolution of chloride and bromide salts of Dy(III) and Y(III). ## {[L2Dy(H2O)5][Cl]3•L2} (1): Yield: 0.160 g (52 %, based on ligand). Anal. Calcd. for C40H110Cl3DyN8O9P4: C, 38.74; H, 8.94; N, 9.04. Found: C, 38.63; H, 8.82; N, 8.74. FT-IR (KBr, cm -1 ): 3260 (br), 2971 (s), 2873 (w), 1646 (w), 1477 (m), 1424 (s), 1399 (w), 1385 (w), 1369 (m), 1313 (w), 1167 (s), 1131 (vs), 1098 (vs), 1051 (vs), 1028 (s), 943 (w), 910 (m), 886 (w), 830 (w), 659 (m), 511 (w). Anal. Calcd. for C41H112Cl5N8O9P4Y: C, 39.35; H, 9.02; N, 8.95. Found: C, 40.1; H, 9.32; N, 9.47. FT-IR (KBr, cm -1 ): 3262 (br), 2971 (s), 2871 (w), 1655 (w), 1467 (m), 1478 (s), 1427 (s), 1398 (w), 1385 (w), 1367 (m), 1320 (w), 1167 (s), 1132 (vs), 1106 (vs), 1055 (s), 1026 (s), 942 (w), 907 (m), 884 (m), 830 (m), 746(m), 654 (m), 513 (w). 1 H NMR (CD3CN, 400 MHz): δ 3.46 (m, 8H, -CH(CH3)2), 1. 15-1.21 (m, 84H, -CH3). 13 C NMR (CD3CN, 100 MHz): δ 42. 7, 32.6, 31.4, 25.8, 25.1, 25.0, 24.7. 31 P NMR (CD3CN, 162 MHz): δ 37.2ppm. 5): Yield: 0.160 g (49 %, based on ligand). Anal. Calcd. for C40H112Br3N8O10P4Y1: C, 36.46; H, 8.57; N, 8.50. Found: C, 36.21; H, 8.41; N, 8.82. FT-IR (KBr, cm -1 ): 3262 (br), 2968 (s), 2869 (w), 1477 (s), 1464 (s), 1421 (s), 1387 (w), 1365 (m), 1171 (s), 1132 (vs), 1108 (vs), 1051 (m), 1026 (s), 906 (w), 883 (m), 831 (m), 729 (w), 655 (w), 637 (w), 501 (w). 1 H NMR (CD3CN, 400 MHz): δ 3.43 (br, 8H, -CH(CH3)2), 3.22 (br, 10H, OH2), 1.20-1.12 (m, 84H, -CH3). 13 C NMR (CD3CN, 100 MHz): δ 42. 5, 32.5, 31.3, 25.4, 24.6. 31 P NMR (CD3CN, 162 MHz): δ 37.5 ppm. ## {[L2Dy0.10Y0.90(H2O)5] [Br]3•L2•H2O} (2@5): 2@5 was synthesized using a similar method using 1 : 9 molar ratios of the respective Dy(III) and Y(III) bromide salts. Yield: 0.140 g (42 %, based on ligand). Anal. Calcd. for C40H112Br3Dy0.10N8O10P4Y0.9: C, 36. 25; H, 8.52; N, 8.46. Found: C, 36.13; H, 8.64; N, 8.74. FT-IR (KBr, cm -1 ): 3271 (br), 3181 (br), 2972 (s), 2935 (w), 2873 (w), 1646 (w), 1477 (m), 1422 (s), 1400 (w), 1385 (w), 1369 (m), 1312 (w), 1169 (s), 1131 (vs), 1102 (vs), 1050 (s), 1027 (s), 909 (m), 885 (w), 830 (w), 659 (m), 509 (w). ## {[L2Dy0.10Y0.90(H2O)5][I]3•L2•H2O} (3@6) : 3@6 was synthesized using a similar method using 1 : 9 molar ratios of the respective Dy(III) and Y(III) iodide salts. Yield: 0.165 g (45 %, based on ligand). Anal. Calcd. for C40H112Dy0.10I3N8O10P4Y0.9: C, 32.77; H, 7.70; N, 7.64. Found: C, 32.56; H, 7.61; N, 7.78. FT-IR (KBr, cm -1 ): 3286 (br), 2969 (vs), 2909 (m), 2872 (w), 1469 (m), 1420 (s), 1399 (w), 1386 (w), 1368 (m), 1311 (w), 1168 (s), 1131 (vs), 1105 (vs), 1049 (s), 1024 (s), 942 (w), 906 (m), 885 (w), 829 (w), 727 (w), 655 (m), 544 (w), 512 (w).

chemsum

{"title": "Deciphering the role of anions and secondary coordination sphere in tuning anisotropy in Dy(III) air-stable D5h SIMs", "journal": "ChemRxiv"}

3d_chemical_imaging_in_the_laboratory_by_hyperspectral_x-ray_computed_tomography

## Abstract: We report the development of laboratory based hyperspectral X-ray computed tomography which allows the internal elemental chemistry of an object to be reconstructed and visualised in three dimensions. The method employs a spectroscopic X-ray imaging detector with sufficient energy resolution to distinguish individual elemental absorption edges. Elemental distributions can then be made by K-edge subtraction, or alternatively by voxel-wise spectral fitting to give relative atomic concentrations. We demonstrate its application to two material systems: studying the distribution of catalyst material on porous substrates for industrial scale chemical processing; and mapping of minerals and inclusion phases inside a mineralised ore sample. The method makes use of a standard laboratory X-ray source with measurement times similar to that required for conventional computed tomography. The penetrative nature of X-rays is exploited in X-ray computed tomography (XCT) to reveal the internal three-dimensional (3D) structure of an object 1 . XCT finds multiple applications across a range of scientific disciplines, including: medical imaging 2 , security scanning 3 , industrial inspection and metrology 4 , materials science 5 , geoscience 6 and archeology 7 , amongst others. The method consists of recording X-ray absorption images (known as radiographs or projections) of an object at multiple rotation angles. Contrast in each radiograph is provided by the differential absorption of the X-rays which is directly related to a materials local density. At a synchrotron x-ray source the image is often collected using monochromatic illumination, whereas for laboratory sources white beam illumination is normally used. Either way, the detector normally collects the transmitted photons with no discrimination in terms of photon energy. A mathematical algorithm is then applied to the data to reconstruct the 3D distribution of the X-ray attenuation (sample density). The gray level histogram is then used to segment the 3D image into phases or features which can then be quantified 5 . Most XCT systems operate in the hard X-ray range (approximately 10-100 keV) primarily using the bremsstrahlung and characteristic radiation resultant from high energy electrons impacting onto a metal target. In this energy range, for the majority of atomic elements in the periodic table, the main contribution to X-ray attenuation is the photoelectric effect, the magnitude of which is approximately proportional to the 4th power of the atomic number (Z) and inversely proportional to the 3rd power of the X-ray energy (E) 8 . Additionally, the atomic photoelectric absorption cross-section shows rapid changes at energies corresponding to core-electron states, the K, L, M-edges. Because these characteristic step-changes in absorption occur at well defined X-ray energies, it is feasible to use these edges to identify individual chemical elements inside the object, as well as reconstruct the sample's absorption contrast. All that is required to obtain this additional chemical sensitivity is the ability to precisely tune the X-ray photon energy to an absorption edge. Using the extremely bright X-rays generated at synchrotron radiation facilities, it is possible to use highly-selective optics (e.g. monochromators and focusing mirrors) that can produce radiation with a very narrow bandwidth (down to the 0.01% level) which can be tuned to specific absorption edges. 3D chemical imaging can then be performed by twice repeating the same tomographic scan at two different X-ray energies, one slightly above and one slightly below the desired absorption edge 9,10 . A basic subtraction of these datasets reveals the 3D distribution of the chemical element corresponding to the absorption edge. This technique has been exploited in areas such as electrochemistry 11 and biomedical research 12 . An alternative method for 3D chemical imaging is to use the X-ray fluorescence (XRF) signal emitted after absorption. In this case a small, intense beam of X-rays from a synchrotron source is focussed down to a point and scanned across the object recording weak XRF signals at each point. This is then repeated at distinct small rotation steps of the object, and a reconstruction algorithm is used to build a 3D map of the atomic elements in the sample 10,13 . This approach has additional sensitivity but is much slower due the beam scanning process. It has been exploited in areas such as environmental chemistry 14 , biological science 15 , earth and planetary science 16 , and materials science 10 . Unfortunately, both of these methods require the use of a synchrotron radiation facility, which have a major limitation from a throughput and access point of view inhibiting wide-scale adoption. Laboratory-based XCT instruments do not have the same limitations in terms of access. Due to the low intensity radiation emitted from a laboratory X-ray tube it is very time-consuming to undertake XRF imaging, while the low flux means CT is normally undertaken using a white beam ruling out imaging either side of an absorption edge. Over recent years, however, a range of pixelated spectroscopic X-ray detectors have been actively researched and developed which have the ability to measure X-ray energy or count the number of X-ray photons in specific energy bands with positional sensitivity . These devices can be roughly grouped into two categories. The first category comprises the so-called single photon counting or multispectral detectors which count every single photon separately by analysing the electrical pulse generated when the photon deposits its energy in a detector pixel. This pulse is compared to that from one or more references energies enabling images to be grouped into a series of spectral bands. These devices generally have a coarse energy resolution (typically 5-10 keV) and can normally have up to 5-10 energy bins 23 . An advantage of multispectral detectors is that they can operate at relatively high count rates. A second category comprises the so-called hyperspectral detectors which measure the energy deposited in each pixel during a certain exposure time (or frame) and consequently calculate the associated photon energy of that pixel in that frame. This repeated over many thousands of frames where events are binned into a spectrum per pixel. The advantage of these types of detectors is that they can achieve a finer energy resolution (typically less than 1 keV) and can image over hundreds of spectral bands. . However, due to the detector readout and associated processing, the measurement speed and maximum count rate is limited. Multispectral detectors have been used in several laboratory-based X-ray imaging studies illustrating the advantages of spectral imaging for differentiating materials 22,27,28 . An obvious application is in medical imaging and tomography for obtaining enhanced contrast between similar density tissues or in the identification of tracer markers like gadolinium or iodine 21,29,30 . For such applications, whereby a great deal of information is already known about the sample composition and/or tracer markers have been used which have widely spaced K-edges, a multispectral detector is more than adequate. However, for materials science applications where the sample composition is often unknown beforehand or the sample contains multiple atomic elements that have closely spaced K-edges, we require an X-ray detector with a finer energy resolution-a hyperspectral detector. Here, we demonstrate that it is possible to undertake chemically sensitive laboratory computed tomography by installing a hyperspectral pixelated detector (with a very high energy-resolution) inside a commercially available laboratory-based XCT system using the broadband bremsstrahlung radiation emitted from a microfocus X-ray tube. In this way we can identify multiple chemical elements simultaneously in 3D without prior knowledge of sample composition and with micron scale resolution. As we will demonstrate, the method has applications to a wide range of scientific disciplines, particularly materials science. Additionally, the ability to perform this kind of analysis in the laboratory (as opposed to a synchrotron) massively opens up the scope of the technique. In order to investigate the feasibility of the method and its practical utility we consider two examples: the first relevant to the chemical and catalytic sciences and the second relevant to geochemistry and environmental sciences. We start by describing the principle of the method as shown in Fig. 1; it is fundamentally analogous to that of regular laboratory-based XCT. X-rays generated from a microfocus X-ray tube illuminate a sample object and the projected radiograph is recorded on an area detector. The geometric magnification generated by moving the sample closer to the X-ray source facilitates micrometer scale spatial resolution. In this case however, we replace the conventional area detector with a HEXITEC hyperspectral detector which can record the absorption spectrum in each pixel with an energy resolution of less than 1 keV over the range of 10-100 keV 17 (for more information on the detector, please see the methods section). By taking the natural logarithm of the normalised intensity in every spectral band we obtain a linear measurement of the absorption cross-section as a function of X-ray energy. A step-change in the attenuation spectrum indicates the position of an absorption-edge, which can be used to identify and map chemical elements in the sample. By rotating the sample and recording energy-sensitive projections it is possible to reconstruct the 3D volume of the sample object, but crucially each voxel now contains an absorption spectrum, as opposed to a single grayscale value, creating a 4D dataset (3 spatial dimensions, 1 spectral dimension). The position of an absorption edge in each reconstructed voxel spectrum is used as a characteristic marker, enabling chemical elements to be mapped in 3D. Our first example application concerns the 3D distribution of catalytic metals supported on porous substrates, known as catalyst bodies. These are mm-sized gas porous support units (typically cylinders or spheres) onto which metals and metal oxides are deposited for the purpose of performing large-scale industrial catalysis. They are loaded in their thousands into industrial scale reactor units forming packed catalyst beds. The functioning of the bed and hence the efficiency of the entire chemical process is dependent on the performance of the individual catalytic body and as such it is important to understand their preparation and chemical form in 3D. The catalyst body provides a fairly inert and mechanically robust porous microstructure onto which a catalyst metal can be thinly dispersed and fixed in position by chemical bonding to the substrate. By choice of the preparation route, the catalyst can be directed to a desired distribution to maximize efficiency, both in terms of producing the desired industrial product and also minimising the amount of catalyst used. For example, expensive precious metals such as Pd are widely employed for both hydrogenation and oxidation reactions on an industrial scale, so any saving in terms of reactor efficiency or a reduction in the quantity of the catalyst metal required in the reaction, will correspond to large financial savings 31,32 . A cylinder shaped γ -Al 2 O 3 pellet with approximate diameter of 3 mm and height 3 mm was loaded with Pd metal ex-situ and scanned by hyperspectral XCT (see methods section for experimental details). Figure 2 shows the results from this scan-the data was reconstructed with a 53 μ m voxel size. Figure 2a shows a typical absorption spectrum for a single voxel taken from the 3D reconstructed data volume. The first thing to note is the y-axis scale and units, here we use voxel optical density (VOD): µ = VOD t voxel voxel where μ voxel is the voxel linear attenuation coefficient and t voxel is the voxel size, as such VOD is a dimensionless quantity. We use this construct, as opposed to the linear attenuation coefficient, because generally the absolute value of the measured attenuation coefficient cannot be relied upon. This is due to various non-linear effects in the data acquisition and handling process, such as inter-pixel charge sharing, inter-frame charge sharing and reconstruction artefacts (see methods section and Supplementary Materials). Relative changes in the absorption spectrum can be reliably reconstructed, for example the position of an absorption edge. Unsurprisingly, in this data we observe a large step in the spectrum at 24.3 keV, corresponding to the Pd K-edge (24.350 keV). The detector has sufficient energy resolution in order to discount the possibility of confusion with adjacent elements in the periodic table (for example confusion of Pd with Rh at 23.220 keV or Ag at 25.514 keV-see Supplementary Materials). Using this data, it is possible to perform semi-quantitative analysis by spectral profile fitting (Fig 2b). We took a region of interest along the spectral domain around the K-edge and performed least squares fitting of linear functions to the data before and after the edge (red lines). By evaluating both linear functions at the K-edge position (24.350 keV, vertical dotted line) and subtracting the result, we get a measure of the magnitude of the absorption step, Δ μ 0 . By doing this analysis we have removed the contribution of the voxel size on the optical density, but more importantly we have decoupled the individual contributions to the total absorption from the γ -Al 2 O 3 and Pd metal, giving us a measure of the absorption cross section from Pd alone. This is important since it cannot be assumed that the density of the alumina is homogeneous throughout the sample. If we scanned this sample using a regular (non-spectroscopic) XCT system, it might be assumed that the observed high density regions correspond to that of Pd. However, this assumes that the alumina pellet has a homogeneous density, and realistically, for a systematic study the pellet should be pre-scanned before metal deposition so a subtraction can be performed. This is not practically possible in many circumstances, and assumes that the alumina pellet is not damaged during processing. This in-situ capabilty really highlights the power of hyperspectral XCT. Additionally, Δ μ 0 is directly proportional to the concentration of the metal species. By performing voxel-by-voxel fitting in this manner, we can build a 3D dataset of Δ μ 0 which can be processed and visualised in 3D. In order to obtain full quantitative information (e.g. absolute concentrations in weight percentage) it would be necessary to run a series of standard samples, having different sizes and concentrations. This is a non-trivial and complex task with many parameters to understand and there is a clear need for future work in this area. Figure 2c,d show vertical and horizontal slices through the Δ μ 0 3D dataset, respectively. From these slices we can see that Pd has penetrated into almost all of the γ -Al2O3 pellet at varying concentrations. More importantly, regions of high concentrations can be seen, in particular around the periphery of the pellet, but also at certain 'hotspots' and at the surfaces of some of the large voids. Figure 2e shows a 3D exterior visualisation of the Pd distribution, where we can see the Pd appears to agglomerate around the circumference, but to a lesser extent on the top and bottom faces of the pellet; in good agreement with previous studies on these samples 33 . In our second case study, we investigate the 3D distribution of mineral phases and inclusions in a mineralised ore sample. The sample is taken from a hydrothermal vein from the Leopard Mine, Silobela, Zimbabwe, part of the ~2.6Ga, Greenstone-hosted auriferous mineralisation in the Zimbabwe craton 34 . The sample comprises pyrite (FeS 2 ), quartz (SiO 2 ), gold (Au) and minor amounts of galena (PbS), chalcopyrite (CuFeS 2 ) and bornite (Cu 5 FeS 4 ). Optical 2D examination of polished sections of this sample reveals quartz and euhedral pyrite (mm to cm sized) to have precipitated first in the vein and been subsequently brecciated with the angular pyrite fragments re-cemented by quartz derived from later fluids. Gold precipitation was coeval with this second quartz episode and occurs as irregular masses (mostly < 100 μ m) and in fractures replacing the pyrite. Gold also occurs as small masses in the quartz and at quartz-pyrite interfaces. The other sulfides forms as irregular (< 50 μ m) inclusions in the pyrite. The pyrite-gold association is common in many exploited ore deposits due to gold absorption to the negatively charged pyrite surface precipitated from hydrosulphidogold(I) complexes or by the reduction of Au-chloride species 35,36 ; the presence of arsenic in the pyrite can concentrate the gold. Understanding the distribution of gold in pyritic and other mineralogically complex ores is important in the development of extraction processing pathways. Economic gold deposits often involve Au in very low concentrations (< 10 gm/t) and the gold is as discrete grains. 3D chemical imaging has the potential to provide definitive information on the size, distribution and mineralogical association that will control liberation during processing. As gold mining has increasingly used heap leaching 37 to extract gold, chemical XCT imaging during processing could have an important role in explaining extraction rates and allowing pathway tuning. Furthermore, the search and exploitation of new mineral resources to replace exhausted deposits and satisfy increasing demand will employ chemical or bio-leaching 38 of large volumes of low grade ore and 3D chemical imaging has the potential to play a significant role in ensuring efficient exploitation. A 20 mm diameter core sample was scanned using the spectroscopic XCT system-the data was reconstructed with a 65 μ m voxel size. Figure 3a shows a grayscale tomographic slice through the sample produced by integrating over the full spectral range. Different mineral phases are observed by their variation in gray level contrast. Brecciation of the pyrite can be seen, particularly highlighted towards the bottom of the image. Figure 3b shows two voxel spectra from two separate inclusions. The first spectrum shows a step change in attenuation corresponding to the Au K-edge (80.725 keV) whilst the second spectrum shows a Pb K-edge (88.005 keV). The measured position of these edges positively differentiates the gold and galena (PbS) mineral phases. In order to segment and separate these two phases we used a K-edge subtraction method. By extracting images at energies above and below the desired absorption edge and subtracted them, we produced Au only and Pb only 3D datasets. A vertical slice through the grayscale data is shown in Fig. 3d with the segmented Au and PbS phases-coloured blue and red respectively-overlaid. Both particles are embedded in pyrite. An important feature of this data is that due to inter-particle variations in the attenuation, it was not possible to accurately segment the Au and PbS phases using grey level (energy indiscriminate) contrast. Some of the gold inclusions show a much higher density which can be easily segmented using a threshold, but many particles show either similar gray level contrast, or a much lower contrast to that of PbS. As such, on the basis of data produced by scanning this sample in a non-spectroscopic (standard) XCT system, it would not be possible to reliably differentiate or segment PbS from Au. Only the K-edge sensitivity of the spectroscopic XCT system can produce a positive identification and therefore successful elemental segmentation. Figure 3c shows voxel spectra from three other mineral phases in the sample: quartz, pyrite and chalcopyrite. These minerals show no observable absorption edges since they contain only low Z-atomic elements with edges below the sensitivity of the system. They can be easily segmented based upon their relative attenuation contrast. In this case however, we exploited the spectroscopic nature of the data and segmented based upon a least square fit of a linear function to the absorption spectra in the range 60-90 keV. The gradient of the linear function is used as a segmentation tool by thresholding. This approach gives better inter-phase distinguishability since it is less sensitive to image noise and was particularly useful in the quartz mineral due its low attenuation and high level of reconstruction artefacts. 3D visualisations of the distributions of each segmented mineral phase, including Au and Pb containing particles, are shown in Fig 3e. From this data we found that the Au inclusions are almost always embedded in pyrite, whereas Galena inclusions can be either embedded in pyrite or quartz. Many inclusions may be below the resolution of the system, either spatially-i.e. smaller than the voxel size of the scan (65 μ m), or spectrally-i.e. with concentrations below the sensitivity limit of the system. With regards spatial resolution, chemical XCT can ultimately produce voxel sizes on a par with current microtomography scanners (< 5 μ m) and is only really limited by the number of pixels on the detector. For this sample with a 20 mm diameter, it could be possible to achieve resolutions of around 10-15 μ m, given a large enough field-of-view. With a smaller sized sample it could be possible to achieve < 5 μ m resolution. With regards chemical sensitivity, we have yet to perform a systematic study, however previous experience suggests a limit of detection of around 1-2%wt per voxel. The above two case studies have highlighted the capabilities of laboratory-based 3D chemical imaging, however there are of course limitations. Most prevalent of these is sample self-absorption. Let's take a theoretical example to demonstrate this: Say we wanted to map the distribution of Mo in a sample of Ti 6246 alloy. This alloy is frequently used in applications where high strength and light weight are needed along with good corrosion resistance. Understanding the distribution of alloying elements in this material, particularly operating under real-world conditions (an attribute of non-destructive XCT) might be of interest 39 . The Mo K-edge is at 20.000 keV, so for absorption edge identification we require X-ray photons of this energy to be able to pass all the way through our sample. If the sample is too large or too dense, too few K-edge photons will make it to the detector and we will not be able to identify Mo in the sample. The attenuation length of 20.000 keV photons in Ti metal is about 140 μ m. An approximate upper limit on the maximum feasible sample size is roughly twice the attenuation length (i.e. 13.5% transmission), so in this case about 300 μ m. This represents a rather small sample, but not impractically so, for high resolution micro CT systems. A sample much larger than this and we will struggle to detect any characteristic photons and therefore be unable to identify this element. So in practice there are limitations on sample size, but also on sample density and as such the method is best suited to studying moderate-to-low density samples containing heavy elements (high atomic numbers-high energy K-edges). Having said that, it is feasibly possible to study lighter elements from the periodic table given the right experimental conditions, for example, the first row transition metals (e.g. Ti, V, Cr, Mn, Fe, Co, Ni, etc.). These elements have K-edges that are below the sensitivity of the detector used in this paper (< 10 keV), however using an appropriate low energy spectroscopic detector (e.g. a silicon pixelated detector with a very high energy resolution 26,40 ) it may be possible to map the distribution of such elements, given a suitably small and low density sample. A previous study demonstrated the ability to identify K-edges below 10 keV using such a detector 40 . In the same work, a CT scan was performed on a mouse artery injected with a gold-containing contrast agent (looking for the gold L-edge at approximately 13 keV). The authors managed to infer the presence of gold and produce a 3D render based on density contrast in energy subtracted images. However they failed to make a positive identification due to a poor signal-to-noise ratio in their data. They also failed to identify the presence of Zn, Ca and Cu in their sample which they had previously identified by X-ray fluorescence. These shortcomings may be more down to current detector technology (in particular the count rate limitations of silicon based detectors) as opposed to a fundamental inadequacy of absorption edge tomography at lower X-ray energies. As detector technology improves, sensitivity to these lighter elements will no doubt improve, as will spatial resolution. In summary, we have developed a laboratory-based XCT system that has the ability to produce 3D images with chemical sensitivity and micrometer scale resolution. By exchanging the regular area detector on an XCT scanner with a pixelated spectroscopic X-ray detector (with a very fine energy resolution < 1 keV) we can obtain hyperspectral 3D datasets with voxels containing absorption spectra. Step-changes in the absorption spectra signify the position of absorption-edges which are used as a fingerprinting tool to identify chemical elements inside each voxel. A particular attribute of this method is that it can be very simply retrofitted in to most currently available XCT systems, simply by adding in a spectroscopic detector, in turn enabling wide scale adoption and expanding the available range of materials characterisation techniques. This can be exploited further by utilising the non-destructive nature of X-ray tomography to study 3D chemical processes occurring inside materials and structures which are functioning under real-world conditions (e.g. temperature, pressure, corrosive environments etc.). We expect this method will find applications to a wide range of scientific disciplines covering materials science (e.g. elemental mapping in metal alloys), planetary, earth and environmental sciences (e.g. siting and association of toxic elements in highly complex materials such as soils), chemistry and catalysis (e.g. in-situ monitoring of catalytic and synthesis reactors), and archaeology (e.g. non destructive evaluation of metal antiquities). ## Methods X-ray detector. A HEXITEC spectroscopic detector was installed in a Nikon XTH 225 system. The HEXITEC detector consists of a 1 mm thick CdTe single crystal detector (20 × 20 mm 2 ) bump-bonded to a large area ASIC packaged with a high performance data acquisition system. The detector has 80 × 80 pixels on a 250 μ m pitch with an energy resolution of 800 eV at 59.5 keV and 1.5 keV at 141 keV 17 . During operation each photon event has its energy, pixel position and the frame in which it occurs recorded. Events are processed and histogrammed according to measured energy into 0.25 keV wide bins. We typical use between 400-800 bins, depending on the maximum X-ray energy. Normally, during this process, a correction is employed to deal with photons that may have shared its energy between two or more pixels which appear to be measured as multiple lower energy photon measurements on neighbouring pixels. When the flux is sufficiently low it is possible to identify these shared events and reconstruct the correct photon energy in the correct location. However due to a high flux of radiation and therefore a high percentage occupancy of events per frame (making it very difficult to identify shared events), we did not employ a charge sharing correction strategy in this case. Previous studies have shown that this does not significantly impact on the measured position of an absorption edge 24,41 . An inter pixel energy calibration was performed using a correlative optimised warping algorithm using data from a flat-field fluorescence image off a series of metals 42 .

chemsum

{"title": "3D chemical imaging in the laboratory by hyperspectral X-ray computed tomography", "journal": "Scientific Reports - Nature"}

erosion_protection_conferred_by_whole_human_saliva,_dialysed_saliva,_and_artificial_saliva

## Abstract: During dental erosion, tooth minerals are dissolved, leading to a softening of the surface and consequently to irreversible surface loss. Components from human saliva form a pellicle on the tooth surface, providing some protection against erosion. To assess the effect of different components and compositions of saliva on the protective potential of the pellicle against enamel erosion, we prepared four different kinds of saliva: human whole stimulated saliva (HS), artificial saliva containing only ions (AS), human saliva dialysed against artificial saliva, containing salivary proteins and ions (HS/AS), and human saliva dialysed against deionised water, containing only salivary proteins but no ions (HS/DW). Enamel specimens underwent four cycles of immersion in either HS, AS, HS/AS, HS/DW, or a humid chamber (Ctrl), followed by erosion with citric acid. During the cycling process, the surface hardness and the calcium released from the surface of the specimens were measured. The different kinds of saliva provided different levels of protection, HS/DW exhibiting significantly better protection than all the other groups (p < 0.0001). Different components of saliva, therefore, have different effects on the protective properties of the pellicle and the right proportions of these components in saliva are critical for the ability to form a protective pellicle. A wide range of prevalence data show that dental erosion is a common condition in developed societies. For adults (aged from 18 to 88 years), prevalences ranging from 4 to 100% have been reported 1 . Dental erosion is defined as a chemical dissolution of dental hard tissues that, in contrast to caries, does not involve bacteria 2 . It starts at enamel, the outermost layer of teeth, which consists mainly of calcium-deficient carbonated hydroxyapatite. In contrast to pure hydroxyapatite, part of the phosphate (PO 4 3− ) and hydroxide (OH − ) is replaced by carbonate (CO 3 2− ), and part of the calcium (Ca 2+ ) is replaced by sodium (Na + ) or magnesium (Mg 2+ ) in the crystal structure of this mineral 3 . These substitutions in the crystal lattice weaken the enamel structure, making it more susceptible to acidic dissolution than pure hydroxyapatite. Most acids relevant to erosion are weak acids, which dissociate in water into hydrogen ions (H + ) and their respective anions 3 , while some of the acid molecules remain undissociated 4 . While the H + will directly dissolve the apatite mineral surface, the undissociated form of the acid will also significantly contribute to enamel dissolution by penetrating enamel pores faster than the dissociated form because of its lack of charge. Once within the enamel, the molecule dissociates and the newly formed H + carries on the dissolution of mineral 5,6 . Enamel dissolution from dental erosion is, therefore, a dynamic process starting with a softening of the enamel surface followed by surface loss 7 . Once surface loss occurs, the mineral cannot be replaced, so erosion at this later stage is irreversible and can result in dentine exposure, and even pulp exposure has been reported in epidemiological studies in children 1 . Human saliva is the most important natural factor that is able to prevent acidic demineralisation and support remineralisation of the dental surface in different ways 8 . On the one hand, it can directly dilute, neutralise and buffer the acids or hinder them from reaching the dental surface. On the other hand, it can remineralise enamel because of its inorganic components. After erosion, ionic interactions between the demineralised tooth surface and saliva can occur, mostly with calcium, phosphate and fluoride ions 8 . A multitude of artificial saliva formulations and remineralizing solutions containing only inorganic components are able to remineralize erosive lesions to a certain extent , although a complete remineralization is unlikely 7 . Another way that saliva protects against erosion is by the formation of a salivary pellicle on the tooth surface. It is formed by adsorption of peptides and proteins onto the tooth surface. Pellicle formation starts immediately after toothbrushing because of ionic and hydrophobic interactions, as well as van der Waals forces, between the proteins and the enamel surface. It is a selective process, as only a specific subset of salivary proteins is present in the pellicle 12 . Calcium-and phosphate-binding peptides and proteins from saliva, especially statherins and acidic proline-rich proteins, can bind with high affinity to the calcium and phosphate on the tooth surface 13 . Additionally, the calcium-binding domains of these proteins can maintain calcium ions near the enamel surface and act as a reservoir 12 . Therefore, the salivary pellicle can simultaneously regulate the uptake and release of calcium and phosphate between the tooth surface and saliva. In this way, the pellicle acts as a semipermeable membrane and maintains the integrity and mineral homeostasis of the enamel surface 8,12 . Many studies have shown that the natural salivary pellicle can partially protect the enamel surface from changes owing to acidic attacks 12,14 . However, the preventive potential of the salivary pellicle is limited and varies with each individual; the composition of the pellicle playing an important role in this variability 15,16 . It has been reported that whole saliva with a high level of calcium and phosphate helps prevent erosive demineralisation compared to dialysed, ion-depleted saliva 12,17 . The importance of the right pellicle composition was demonstrated in a study analysing the pellicle composition of erosion patients, which showed that pellicles of such patients contain a lower level of calcium and proteins such as statherins 13 . However, other studies do not show any correlation between salivary ion content and susceptibility to erosive tooth wear, so there is still some controversy about the actual effect of salivary ions and proteins on erosive demineralisation 8,18,19 . The aim of the present in vitro study was to investigate the differences in erosion protection that is conferred by treating enamel with natural saliva, dialysed saliva containing only salivary proteins and artificial saliva containing only salivary ions, to gain further insight into the importance of the different components of saliva for the formation of a protective pellicle. ## Materials and Methods Preparation of the enamel specimens. A total of 75 enamel specimens were prepared from human molars. The teeth were selected from a pool of extracted teeth, which had been extracted by dental practitioners in Switzerland (no water fluoridation, 250 ppm F − in table salt) and were stored in 2% chloramine T trihydrate solution. Patients were informed about the possible use of their teeth and consent was obtained. The experiment was carried out in accordance with the approved guidelines and regulations of the local ethics committee (Kantonale Ethikkommission: KEK), which categorized the teeth as "irreversibly anonymized" because they had been pooled. The roots of the selected teeth were removed using an Isomet ® low speed saw, and the crowns were cut into buccal and lingual halves. Using two planar parallel moulds, the halves were embedded in acrylic resin. The thinner moulds (200 μ m thick) were removed and the teeth were ground and polished using a Knuth Rotor machine with abrasive silicon carbide paper discs of grain size 18.3 μ m, 10 μ m and 5-6 μ m. After removing the thicker moulds, they were further polished for 1 min with a 3 μ m grain diamond paste on a polishing cloth under constant cooling. The specimens were stored in a mineral solution (1.5 mmol/l CaCL 2 , 1.0 mmol/l KH 2 PO 4 , 50 mmol/l NaCl; pH 7) 20 until being used in the experiment. Immediately prior to the start of the experiment, they were additionally polished with a 1 μ m grain diamond paste for 1 min under constant cooling. Between the polishing steps and after the final polish, all slabs were ultrasonicated for 1 min in tap water and rinsed. The specimens were randomly distributed into 5 treatment groups (n = 15, Table 1): humid chamber (control, Ctrl group); human whole stimulated centrifuged saliva (HS group); artificial saliva (AS group); HS dialysed against artificial saliva (HS/AS group); HS dialysed against deionised water (HS/DW group). Collection of human whole stimulated saliva. Stimulated saliva was obtained from adults aged between 20 and 30 years with no active caries. They were instructed not to eat or drink anything except water for 2 h before saliva collection. Stimulated saliva was collected into chilled vials while chewing on paraffin wax for 10 min. Immediately after collection, the saliva was pooled and centrifuged for 20 min at 4 °C (4000 g). After centrifuging, the supernatant was divided into aliquots of 20 ml, which were stored at − 80 °C. The donors provided their informed oral consent to use the saliva for research purposes. Since the saliva was pooled, it was categorized as "irreversibly anonymized", and the experiment was carried out in accordance with the approved guidelines and regulations of the local ethics committee (Kantonale Ethikkommission: KEK). Dialysis of human whole stimulated saliva. Dialysis was performed with a Mega Pur-A-Lyzer TM Dialysis Kit with a membrane cut-off of 1 kDa (PURG10020, Sigma-Aldrich). Human saliva was dialysed against water or artificial saliva. Aliquots of 20 ml of human saliva were thawed and dialysed. Dialysis against artificial saliva (AS; 1.5 mM Ca(NO 3 ) 2 , 0.90 mM KH 2 PO 4 , 130 mM KCl and 60 mM Tris buffer; pH 7.4 21 ) was carried out in 2 l of AS for a total of 48 h at 4 °C. AS was exchanged after 2, 8, 16 and 24 h. Dialysis against deionised water was Although it is known that freezing and thawing of saliva can cause certain proteins to precipitate 22 , we did neither observe any precipitates nor turbidity after thawing the saliva. In any case, as all saliva prepared for the different groups were frozen and thawed, precipitation would have affected all groups equally, so differences between the groups caused by freezing and thawing can be excluded. Saliva collection and dialysis resulted in the four different kinds of saliva used in this study: human whole mouth stimulated centrifuged saliva (HS); artificial saliva (AS); HS dialysed against artificial saliva (HS/AS); HS dialysed against deionised water (HS/DW). Analysis of calcium and inorganic phosphate concentrations. Total calcium (Ca) concentration in all solutions was determined using an atomic absorption spectrometer. Lanthanum nitrate (0.5%, lanthanum nitrate hexahydrate: La(NO 3 ) 3 •6H 2 O) was added to the solution to eliminate the interference of other ions 23 . Total inorganic phosphate (P i ) was analysed using the method of Chen et al. 24 . A total of 2 ml of diluted solution was mixed with 2 ml of a phosphate reagent (2% ascorbic acid, 0.5% ammonium heptamolybdate, 0.6 M H 2 SO 4 ), stored for 90 min at 37 °C, allowed to cool to room temperature (24 °C), and analysed using a photometer. Analysis of total protein concentration. The total protein concentration was determined colourimetrically using a Pierce TM BCA Protein Assay Kit (23227, Thermo Scientific) with bovine serum albumin as a standard. The assay was performed in a 96-well micro plate, using triplicates of each sample and standard. The plates were read at 570 nm using an ELx808 Absorbance Reader. Incubation -erosion cycles. The experiment consisted of four incubation -erosion cycles. One cycle consisted of individually immersing the specimens for 1 h in 1.8 ml of the respective incubation solution, in a shaking water bath (70 rpm) at 37 °C (Gesellschaft für Labortechnik mbH, Burgwedel, Germany). In the case of the Ctrl group, the specimens were left in a humid chamber for 1 h at 37 °C. Afterwards, the specimens were rinsed with deionised water for 20 s, dried with oil-free air for 5 s, and surface hardness (SH) was measured. The specimens were then submitted to an erosive challenge, which consisted of individually immersing the specimens in 10 ml of citric acid (1%; pH 3.6) in a shaking water bath (70 rpm) for 1 min at 25 °C (P-D Industriegesellschaft mbH Prüfgerätewerk, Dresden, Germany). After the erosive challenge, the specimens were rinsed with 100 ml of tap water for 20 s and dried with oil-free air for 5 s. SH was measured again, while the citric acid solutions were labelled and stored for later calcium analyses. The examiner performing the SH measurements and calcium analyses was blind to the solutions. In total, the specimens underwent 4 cycles of incubation and a total of 4 min of erosion. Surface hardness measurement. Surface hardness (SH) measurements were carried out initially and after each incubation, as well as after each erosive challenge. The initial SH value was labelled SH 0 . Thereafter, odd numbered values (SH 1 , SH 3 , SH 5 , SH 7 ) represent values after incubation periods, while even numbered values (SH 2 , SH 4 , SH 6 , SH 8 ) represent values after erosions. SH measurements were carried out on a Fischerscope HM2000 XYp using a Vickers diamond under a pressure of 50 mN for 15 s and results were expressed in Vickers Hardness Numbers (VHN). For each SH measurement, a total of seven indentations were made parallel to each other, spanning an overall distance of 200 μ m. Subsequent measurements were performed 100 μ m from the measurements of the previous step. The mean value from the seven indentations at each step was considered for analysis. The changes in enamel hardness between the initial measurement (SH 0 ) and the following steps were calculated as percentage change (∆ SH) and used for statistical data analysis and interpretation. ∆ SH was calculated using the following formula: ∆ SH = (SH i /SH 0 ) × 100, where SH 0 is the initial measurement, and SH i is the hardness value after the i th measurement (after the i th incubation in a solution or after the i th erosion). ## Analysis of calcium release into the citric acid. After each erosive challenge, the amount of calcium released into the citric acid was analysed. Calcium was analysed as described before, and the measured calcium concentrations were normalized to the corresponding enamel surface areas of the specimens. To calculate the surface area, we used a light microscope (Leica, M420) at 16x magnification connected to a camera (Leica, DFC495). Using the software program IM500, the contour of the exposed enamel window was traced and the surface area of each specimen was calculated. ## Statistical analysis. Using an identical protocol, a preliminary study showed that specimens kept in a humid chamber presented a 55.3 ± 2.9% (mean ± sd) decrease in SH, whereas specimens incubated in human saliva presented a 52.3 ± 1.6% decrease in SH after 4 experimental cycles (effect size of d = 1.28). Considering an effect size of d = 1.3, an α error (type I) of 0.05, and a power (1 − β ) of 0.80, we calculated a sample size of 11 samples per group (actual power 0.81; t = 2.09). Therefore, we used a sample size of n = 15 specimens per group. The normality of the data sets was checked using graphical methods as well as Shapiro Wilk's test. Both outcomes SH and Ca were not normally distributed (p < 0.0001). The main effects of the whole-plot factor group and time and their interaction were thus tested using non-parametric time-related ANOVA 25 . The resulting p-values were corrected for multiple testing with Holm's method. Initial absolute values of SH were analysed by Kruskal-Wallis test, revealing no differences between the groups. It was therefore excluded as variable in the further analyses. For SH, the percentage change (∆ SH) was then used as the variable in the statistical analysis. In the case of significance in the global test, post-hoc analysis was performed with Kruskal-Wallis tests for simultaneously comparing all different groups and then with Wilcoxon-Mann-Whitney tests for pair-wise comparisons. P-values in this section were not corrected for multiple testing owing to the explorative nature of this part of the analysis. The level of significance was set to 0.05. ## Results Surface hardness. SH was measured nine times: initially (at baseline -SH 0 ) and after each subsequent incubation and erosion (SH 1-8 ). The overall SH 0 mean ± standard deviation was 504.2 ± 40.2 VHN, ranging between 444.8 and 612.7 VHN. Figure 1 shows the relative SH values for all five groups throughout the entire experiment. In general, SH values decreased in all groups after every erosive challenge. The incubation periods of 1 h generally led to an increase in SH values in all groups, except for the Ctrl group. This increase in SH was most evident in the HS group. However, the preceding hardness could not be restored, as the decrease in SH values after the erosive attacks was greater than the increase during the next incubation period. Both the degree to which SH values decreased (after the erosive attacks) as well as the degree to which SH values increased (after incubation) appeared to escalate with each experimental cycle. The SH values of the HS group showed the greatest increase in SH after each incubation, increasing by 4%, 4%, 6%, and 8%, for SH 1 , SH 3 , SH 5 , and SH 7 , respectively, compared to the values before incubations. However, SH also decreased considerably in the HS group after each erosion step, with drops of 9%, 13%, 15%, and 17%, for SH 2 , SH 4 , SH 6 , and SH 8 , respectively, compared to the values before the erosion step. The latter decreases in SH values were comparable to those in the HS/AS group. The HS/DW group exhibited the mildest decrease in SH values, leading to the highest SH values at the end of the experiment, and the HS/AS group exhibited the lowest SH values, significantly lower than the other groups from the SH 4 measurement onwards. Figure 2 shows the relative SH values at the end of the experiment. The SH in the HS/DW group decreased from 100% to an average of 85.5 ± 7.6%, which was a significantly lower decrease than in the other groups (p < 0.001). The HS and AS groups had SH values decreasing from 100% to 68.0 ± 8.3% and 66.3 ± 9.8%, respectively, and they were not significantly different from each other (p = 0.513). They were, however, significantly different to the Ctrl group (62.1 ± 6.5%) and to the HS/AS group (p < 0.05). The latter presented the greatest decrease in SH, with values decreasing from 100% to 55.3 ± 7.0% (p < 0.05 compared to all other groups). Calcium release during erosion. The cumulative calcium release for all erosion cycles is presented in Fig. 3. From the first cycle on, the HS/DW group presented the lowest calcium release, while no differences were observed between the other groups. Figure 4 shows the total amount of calcium released after all cycles. Significantly different amounts of calcium were released in some of the groups. The HS/DW group clearly released the least amount of calcium (6.28 ± 1.60 nmol/mm 2 ), followed by the Ctrl group (11.73 ± 1.84 nmol/mm 2 ), and they both released significantly less calcium (p < 0.001) than the HS group (15.86 ± 3.14 nmol/mm 2 ), the AS group (15.69 ± 1.80 nmol/mm 2 ), and the HS/AS group (15.33 ± 2.98 nmol/mm 2 ). However, no significant differences were observed in the pattern of calcium release for HS, AS and HS/AS groups throughout the entire experiment. ## Discussion The saliva used in this experiment was collected from many donors and pooled to avoid any bias that could arise from using saliva from only one donor 26 . This pooled saliva was then used to form two additional groups: one group was dialysed against artificial saliva, resulting in an equal ion content between artificial saliva and human saliva, and another group was dialysed against deionised water, resulting in a solution with very low ion concentrations and mainly salivary proteins. Since salivary proteins and peptides bind calcium and phosphate ions, it was not possible to completely remove these ions from saliva using dialysis (Table 1). Nevertheless, the concentrations of these ions were greatly reduced, and we considered the resulting HS/DW as deprived of free ions. For dialysis, we chose a membrane with a molecular weight cut off of 1 kDa. This would allow only the diffusion of ions and molecules smaller than 1 kDa, so the main protein content would remain in the dialysed solution. However, some salivary peptides are smaller than 1 kDa and may have diffused out of the saliva. Furthermore, some salivary proteins might stick to the dialysis membrane and be partly removed from the dialysed saliva. Since pellicle formation is a specific process, a possible minor partial removal of some proteins was tolerated as it should not affect the final pellicle composition. Hence, HS/AS and HS/DW had basically the same protein content since they were both dialysed, but different ion compositions (Table 1), and they both had lower protein content than the undialysed HS. Moreover, despite the small differences between the dialysed salivas and HS, comparisons can still be drawn. In addition to the dialysed salivas and HS, we also used artificial saliva (AS) to assess only the effect of the present ions. The ionic concentration of AS, however, was not identical to the concentration of ions in HS, so HS/AS was prepared to have (almost) the same protein content as HS, but with the same mineral ion concentrations as AS. This would allow a direct comparison of the effect of mineral ions alone (AS) to that of mineral ions together with proteins (HS/AS). We observed high initial SH values (SH 0 ), which decreased in all groups as the experiment progressed. As expected, the greatest decreases in SH were observed immediately after erosive challenges (SH 2 , SH 4 , SH 6 , and SH 8 ). After the saliva incubation periods following erosions, slight increases in SH were observed in all groups, except for the Ctrl group. The greatest increases occurred in the HS group, but the SH values obtained before the preceding erosion steps were never reached again. Moreover, this effect was not enduring, as the SH values of the HS group after erosion decreased to a similar level as those of the AS group, and at the end of the experiment we found no significant difference in SH values between the HS and AS groups. A possible explanation for the comparably high increase in SH values after incubation in the HS group is that ions formed mineral deposits on the eroded enamel surface. The reason for this deposition could be twofold: it could be due to the presence of the ions themselves in the solutions, and it could be due to the presence of some proteins in the solutions. HS contains a comparatively high phosphate concentration, as well as other ions that were removed in the other groups by dialysis. These different ion concentrations and compositions (Table 1) might explain the differences observed in the results. A comparison of HS/AS and AS groups supports this thesis. They have the same ion composition, and we observed no great differences in the increase in SH values after the incubation periods in those groups. In addition to calcium and phosphate, other ions also affect the dissolution of hydroxyapatite (HAp) 14 . It has been suggested that chloride and sodium are also associated with suppression of HAp dissolution, possibly as a result of competition for HAp surface protonation sites between Na + and H + ions, but not by the incorporation of Cl − ions into HAp 14 . Therefore, the presence of all these ions is probably an important factor in the increase in SH of enamel. Another factor that is involved in increasing the SH values is the presence of some proteins. Remarkably, the solution deprived of free ions (HS/DW group) also showed a slight increase in hardness after incubations after erosions, which must mainly be due to the proteins present in the solution. Different proteins have been shown to either inhibit or promote erosion or lesion remineralization . Also, faster adsorption of proteins to eroded enamel has been reported 30 . The influence of proteins on enamel demineralization varies significantly, depending on how much is adsorbed and how much free calcium is available near the surface 31 . The balance of enamel porosity, the degree of saturation of a solution with respect to enamel minerals, and protein concentrations in the solution seem to determine the exact effect that the solution has on the enamel surface 28 . This could explain the SH pattern we observed in the case of the HS/DW group, which seemed to have a slight rehardening effect on eroded enamel, but the opposite effect on polished enamel with a slight decrease in hardness after pellicle formation (SH 0 -SH 1 , Fig. 1). The exact mechanism underlying how SH can increase after an incubation with a solution containing primarily proteins but no meaningful amounts of ions is not clear and warrants further investigation. One possible explanation might be that salivary proteins penetrate into enamel pores opened by erosion, acting as bridging ligaments between adjacent enamel rods and thereby toughening the enamel, which has been reported for enamel proteins 32 . In addition to their effect on SH modulation, the proteins are most importantly responsible for the formation of the salivary pellicle, which also plays an important role in the protection against erosion. This was mainly observed in the HS/DW group, where there were very little ions present, and the incubation solution was mostly made up of proteins. The salivary pellicle formed with HS/DW presented the best protective effect in our study, with the least overall decrease in SH and the least calcium release (Figs 2 and 4). Our results, however, are not consistent with the results obtained in the study by Martins et al. 33 , which showed that dialysed saliva did not provide good protection against enamel demineralisation and that whole saliva provided the most effective erosion protection. These authors carried out only 16 h of dialysis, whereas we dialysed for 72 h, so our method might have removed the mineral ions more thoroughly. Furthermore, the authors measured the effect only with respect to calcium and phosphate release, not with respect to hardness. Owing to reasons explained below, our own calcium release measurements did not allow for a comparison of the undialysed HS with HS/DW, so a direct comparison of our results to the results of Martins et al. is not possible. Recently, it was suggested that the organic components (proteins) of saliva, either solely or by interacting with salivary chemical properties, could have a significant impact on the susceptibility of the tooth surface to demineralisation 8 . In case of HS/DW, most of the ions, especially calcium and phosphates, were removed from the saliva. When calcium and phosphates are present in saliva, they will compete for the binding sites on calciumand phosphate-binding proteins with the ions present on the enamel surface. Lacking calcium and phosphate ions, we speculate that the calcium-and phosphate-binding proteins in HS/DW could bind to the enamel surface better. Hence, the salivary pellicle basal layer formed by HS/DW was probably more strongly bound to the enamel surface, and therefore more protective against erosion. In other words, while being indispensable to allow remineralisation of dental hard tissues, calcium significantly affects the protective effect of the salivary pellicle, and its concentration has been shown to have a dramatic impact on pellicle thickness and density 34 . HS and HS/AS presented greater protein content, as well as greater calcium and phosphate concentrations, but they exhibited a significantly less protective effect than HS/DW. With regard to the calcium release results, we observed that the HS, AS and HS/AS groups released similar amounts of calcium throughout the experiment (Figs 3 and 4). Usually, this technique is used to measure the amount of calcium from dental hard tissues that dissolves into the citric acid during an erosive challenge 35 . The Ctrl group, being unprotected, was expected to exhibit the greatest calcium release. However, the HS, AS and HS/AS groups released larger amounts of calcium than the Ctrl group. Since the entire specimens were immersed in the incubation solutions, a pellicle and/or calcium deposits also likely formed on the resin surfaces and not only on the enamel surfaces. The calcium release values measured therefore originate not only from the dental hard tissue, but also from the pellicle itself or from calcium deposits formed during the incubation time. In this respect, other studies have shown a considerable amount of calcium released from the saliva film formed on porcelain discs 14 . Due to limitations in the amount of saliva available, we were not able to determine the amount of calcium released from deposits or pellicles formed on resin surfaces. Hence, we could not correct the calcium release values for this factor. Given an overall surface area of ~660 mm 2 for the specimens, with an average exposed enamel area of only 12.8 mm 2 , the amount of calcium released from deposits or pellicles formed on the resin surface can surely not be neglected. Therefore, the calcium release results for the three groups HS, AS and HS/AS did not allow us to draw any conclusions. However, we can still draw conclusions from the Ctrl and HS/DW groups. The Ctrl group calcium release values can originate only from the apatite crystals of the enamel. Considering that HS/DW contained only 0.15 mmol/l of calcium, the amount of calcium bound by the pellicle in this group was negligible, so the calcium released mainly derived from the tooth surface. However, the HS/DW group clearly showed the least amount of calcium loss, and this was observed from the first cycle (1 min) onward (Fig. 3). It released significantly less calcium than the Ctrl group (Fig. 4), which means that the pellicle formed in this group was able to significantly protect the enamel from erosion. Interestingly, the lower amount of calcium released is also consistent with the SH results, where samples incubated in HS/DW presented the least SH decrease (least erosion). Although in vitro results are difficult to apply to in vivo circumstances, it would be interesting to know to what extent we can extrapolate these study results to the clinical situation. We observed an increase in SH during the 1 h incubation in human saliva in our experiment. However, in vivo remineralisation is quite a slow process 8 . Additionally, contrary to subsurface caries lesions, any remineralisation of erosion lesions is restricted to the softened near-surface demineralised enamel layer 5,19 . Even if some enamel remineralisation occurred in our samples, it did not completely protect the enamel against further demineralisation. Furthermore, any remineralisation from saliva occurring in vivo during the short time between acid attack and toothbrushing would also not protect the enamel against substance loss from abrasive forces 19 . Therefore, any increase in SH observed in our study would not lead to clinically relevant protection, and the focus should be more on protection from erosion rather than on the seemingly unsustainable remineralization effects of saliva after erosion. ## Conclusion Salivary proteins as well as salivary ions both play a role in protecting enamel from erosion. Pellicles formed by salivary proteins depleted of salivary ions (HS/DW) provide the best erosion protection effect, as indicated by the least amount of decrease in SH and the least amount of calcium released throughout the entire experiment. This might be due to a stronger binding of calcium-and phosphate-binding proteins to the enamel surface in the absence of free ions. Incubation in AS, containing only salivary ions, and pellicles formed by HS provided slightly less protection, and no significant differences were observed between the respective treatment groups. If salivary proteins and ions are both present, the exact composition appears to be important for the erosion protection effect, as demonstrated by the differences between the HS and HS/AS groups.

chemsum

{"title": "Erosion protection conferred by whole human saliva, dialysed saliva, and artificial saliva", "journal": "Scientific Reports - Nature"}

pattern_of_altered_plasma_elemental_phosphorus,_calcium,_zinc,_and_iron_in_alzheimer’s_disease

## Abstract: for the AddNeuroMed consortium * & po-Wah so 1 Metal/mineral dyshomeostasis has been implicated in the development of Alzheimer's disease (AD).The aim of the study was to investigate the difference in absolute and percentage levels of plasma phosphorus, calcium, iron, zinc, copper, selenium in cognitively normal (CN) and AD subjects. total reflection X-ray fluorescence (TXRF) spectroscopy was used to detect plasma metals/minerals in CN and AD subjects (n = 44 per group). TXRF detected significantly increased plasma levels of phosphorus (p = 1.33 × 10 −12 ) and calcium (p = 0.025) in AD compared to CN subjects, with higher phosphorus/ calcium (p = 2.55 × 10 −14 ) ratio in the former. percentage concentrations calculated for phosphorus, calcium, iron, zinc, copper, selenium by dividing the concentration of each element by the total concentration of these elements and multiplying by 100%, demonstrated phosphorus was higher in AD compared to CN subjects, while calcium, iron, zinc, copper and selenium were lower in AD subjects, with area under the curves as high as 0.937 (p = 6 × 10 −5 ) computed from receiver operating curves. With exclusion of high levels of phosphorus and calcium from percentage calculations, iron levels remained low in AD whereas zinc was higher in AD, and copper and selenium levels were similar. We demonstrate altered distribution of elements in the plasma of AD subjects with high interdependencies between elemental levels and propose the potential of TXRF measurements for disease monitoring. Alzheimer's disease (AD) is the most common cause of dementia in the elderly population with a predicted incidence of 115.4 million cases by 2050. It is characterised by pathological inclusions of extracellular ß-amyloid (Aß) plaques and intracellular neurofibrillary tangles 1 . With the lack of a cure, and only mildly effective symptomatic treatments at best, there is a great impetus to understand this incurable dementia. Metal dyshomeostasis (particularly iron, zinc and copper) has recently been implicated in the development of AD 2 . These metals can accentuate plaque formation and affect tau hyperphosphorylation albeit to a lesser extent. Aberrant homeostasis of these transition elements and enhancement of protein aggregation and detrimental interactions with protein aggregates can generate oxidative stress contributing to neurodegeneration in AD 2 . Also, there has been considerable debate regarding the contribution of aluminium to AD. A common feature of neurodegenerative diseases such as AD are changes in ratios of various metals 3 . Dyshomeostasis of other micronutrients such as calcium is also evident in the AD brain 4 . Circulating calcium and phosphorus are regulated by parathyroid hormone (PTH). In secondary hyperthyroidism, PTH levels are high, alongside normal to low calcium and a concurrent decrease in vitamin D levels 5 . High serum phosphate has been associated with increased incidence of dementia 6 as well as changes in phosphorus/calcium ratios 7 . AD has a well-known prodromal stage 1 which provides a therapeutic window when pathological processes are still responsive to treatment. However, determining this prodromal stage is difficult and drives much of the current biomarker research. While cerebrospinal fluid (CSF) Aß, phosphorylated and total-tau are used as biomarkers of AD 1 , their use to monitor the prodromal phase is debatable and impractical. CSF collection is by lumbar puncture, an uncomfortable sampling method for the individual, especially the elderly and associated with morbidity, and while used to assist diagnosis, it is inappropriate for screening or repeat sampling for disease monitoring. Thus, there is the impetus towards peripheral-based biomarkers such as those in plasma which can be readily collected and repeatedly, if necessary. Investigators have tested whether alterations of brain metals and other elements are peripherally reflected to potentially aid in diagnosis and disease monitoring 5,6,8 . No consensus has been obtained with regards to whether the peripheral concentrations of various elements are modulated in AD owing to discrepant findings. This may be attributable to plasma levels, being not only affected by AD pathogenesis, but also affected by diet, nutritional status, sex, age, etc. Additionally, the analytical method used may impact the measurements, some methods requiring sample preparation, the complexity varying between analysis methods; whether free or bound elements or their isotopes are measured, etc. A few studies do report analysis of a panel of metals by inductively-coupled plasma mass spectroscopy (ICP-MS) showing differences between cognitively normal (CN), mild cognitive impairment (MCI) and AD 2,3 . In this study, we quantify the plasma elemental concentrations of phosphorus, calcium, iron, copper, selenium and zinc using total reflection X-ray fluorescence (TXRF) spectrometry in age-and sex-matched AD and cognitively normal (CN) individuals. TXRF measures all the isotopes of an element and detection limits can be comparable to atomic absorption spectroscopy (AAS), inductively coupled plasma optical emission spectroscopy (ICP-OES) or ICP-MS 9 . Matrix effects are minimised enabling higher sensitivity and low detection limits in the order of parts per billion. A key advantage of TXRF is the relatively simple sample preparation required compared to other techniques, contributing to its robust nature, with errors relating to pipetting small volumes rather than extraction efficiency of metals from biofluids or matrix effects. Other advantages include smaller analytical volumes needed (as low as 10 µl) 10 ; simple quantification using an internal standard; and simultaneous multi-element trace analysis without bias towards expected elemental compositions 11,12 . We aim to determine whether plasma phosphorus, calcium, iron, copper, selenium and zinc; and phosphorus/calcium ratio, differentiates between CN and AD individuals and may be putative peripheral biomarkers of AD. This is the first study that we are aware of, applying TXRF to the analysis of blood for biomarker discovery in AD, and proposes a possible role of TXRF instrument in the clinical chemistry laboratory. ## Materials and Methods subjects. Plasma samples from AD and CN controls were selected from the AddNeuroMed cohort 13 . Informed consent was obtained from all participants according to the Declaration of Helsinki (1991), and protocols and procedures were approved by The Joint South London and Maudsley and the Institute of Psychiatry NHS Research Ethics Committee. The diagnosis of probable AD was made according to the Diagnostic and Statistical Manual for Mental Diagnosis, fourth edition and National Institute of Neurological, Communicative Disorders and Stroke-AD and Related Disorders Association criteria, with Mini-Mental State Examination (MMSE) score of <24 for cognition. The APOE single nucleotide polymorphisms (SNPs) rs429358 and rs7412 were genotyped using Taqman SNP genotyping assays (determined by allelic discrimination assays based on fluorogenic 50 nuclease activity) and the allele inferred. In total, we examined 88 subjects: n = 44, AD; n = 44, CN; age-and sex-matched. Blood Collection. Blood collection was performed in the morning after an overnight fast. Venous blood was collected into lithium heparin tubes (Becton and Dickinson) and centrifuged. Plasma was separated, frozen and stored at −80 °C prior to TXRF analysis. ## TXRF of Plasma. Plasma were thawed and 20 μl of a gallium standard solution (2 mg/l, Kraft GmbH, Germany) was added to each sample volume of 20 µl. After thorough mixing, 10 μl of each solution was put onto TXRF quartz glass carriers and air dried: duplicates of each sample were performed. TXRF data was collected over 1000 s on a TXRF spectrometer (S2 PICOFOX, Bruker Nano GmbH, Germany) with a molybdenum tube excitation source operating at 50 kV/600 μA. TXRF spectra were inspected, and all elements identified, prior to deconvolution of the spectra using the PICOFOX TM software. Elemental concentrations of phosphorus, calcium, iron, copper, zinc and selenium were calculated by reference to the gallium standard in each sample. Accuracy of the TXRF measurements was validated by analysing two certified reference samples (Seronorm TM serum levels 1 and 2, Sero, Norway). The lower limit of detection (with three-sigma detection) for an element e (LLD e ) was calculated according to the following equation: where C e is the elemental concentration, and A e and A bg are the areas of the fluorescence peak for element e and the background subjacent to the fluorescence peak for e, respectively. The measurement time for the calculation of LLD e was also 1000 s, the same as for the plasma samples. Each Seronorm ™ standard was measured ten times. The method detection limit (MDL) was also calculated. The MDL is defined as 'the minimum concentration of a substance that can be measured and reported with 99% confidence that the analyte concentration is greater than zero and is determined from analysis of a sample in a given matrix containing the analyte' . The MDL e for an element e was calculated (according to the U.S. Environmental Protection Agency, Title 40 Code of Federal Regulations Part 136, Appendix B, revision 1.11) using the following equation: where t (n=1,99%) is the students' t-value for a 99% confidence level and a standard deviation (SD) with n − 1 degrees of freedom. statistical Analysis. Data analysis was performed using SPSS IBM version 24.0. Normal distribution was checked graphically using box plots, histogram, Q-Q plots, stem and leaf diagrams and numerically using Shapiro-Wilk's test. The following variables were square-root transformed to normality: phosphorus, copper, selenium; and calcium, iron and zinc were log transformed. Analysis of Co-Variance (ANCOVA) was used to determine elemental differences between the CN and AD groups with co-variates: diagnosis, sex, APOE status, diagnosis x sex, diagnosis x APOE, diagnosis x sex x APOE and sex x APOE. A p-value of ≤ 0.05 was considered significant. The phosphorus/calcium ratio was also calculated as in previous studies and underwent ANCOVA testing. Absolute concentrations of metals/elements in serum have been shown to be highly variable 14 and possibly interdependent. Thus, partial correlation analysis adjusted for age, sex, APOEε4 and diagnosis status was performed to determine interdependencies between the elements: phosphorus, calcium, iron, copper, zinc and selenium. A p-value of ≤ 0.05 was considered significant. Percentage concentrations were also calculated for phosphorus, calcium, iron, zinc, copper, selenium by dividing the concentration of each element by the total concentration of these elements and multiplying by 100%. As percentages of the elements were not normally distributed, the Mann-Whitney U test was performed to determine differences between the CN and AD, with p-value ≤ 0.05 being considered significant. As phosphorus and calcium levels were comparatively much higher than that for iron, zinc, copper and selenium, percentages for the latter elements were re-calculated with the exclusion of phosphorus and calcium and again tested for significance. Binary logistic regression model adjusted for age, sex and APOEε4 status was used to calculate predictive probabilities of elemental percentages. These standardized predictive probabilities were then used to calculate the receiver-operator curves (ROC) and areas under the curve (AUC). For regression models, we ensured normal distribution of residuals and absence of multi-collinearity. A p-value of ≤ 0.05 was considered significant. ## Results Demographics, clinical data, and plasma elemental concentrations for CN and AD groups are shown in Table 1 and a dot plot of the elemental concentrations shown in Fig. 1. TXRF measurements of elemental concentrations were validated using Seronorm ™ Levels 1 and 2 (Table 2). Of the elements studied, selenium had the lowest abundance in the plasma and levels as low as 0.012 mg/l could be measured. ## Absolute plasma elemental Concentrations and Interdependencies. Plasma phosphorus (p = 1.33 × 10 −12 ) and calcium (p = 0.025) were significantly higher in AD compared to CN (Tables 1 and 3). Further, a higher phosphorus/calcium (p = 2.55 × 10 −14 ) ratio was observed in AD patients (Tables 1 and 3). Absolute plasma concentrations of iron, zinc, copper and selenium were similar between the groups (Table 1). By partial correlation analysis, plasma levels of elemental phosphorus, calcium, iron, zinc, copper and selenium were shown to be strongly interdependent (Fig. 2). Hence, further statistical analysis between the CN and AD groups was performed on percentages of these elements. percentages of plasma elemental Concentrations. Of the six elements quantified, the percentages of phosphorus and calcium were much higher in the plasma than that for iron, zinc, copper and selenium (Table 4). Further, the percentage of phosphorus in the plasma was higher in AD than CN Table 4), whereas the percentage of calcium (p = 6.04 × 10 −11 ), iron (9.85 × 10 −9 ), zinc (p = 1.20 × 10 −5 ), copper (p = 1.63 × 10 −10 ) and selenium (p = 9.22 × 10 −12 ) were lower in AD (Table 4). As phosphorus and calcium were at much higher levels than the other elements, their inclusion for the calculation of elemental percentages could potentially bias the analysis. Thus, the analysis was repeated but percentages of iron, zinc, copper and selenium were re-calculated with the exclusion of phosphorus and calcium. Again, plasma iron was significantly lower in the AD patients compared to CN subjects (Table 4; p = 0.028). However, zinc was higher in AD than CN (Table 4; p = 1.74 × 10 −4 ), whereas plasma copper and selenium levels were comparable between CN and AD. ## Logistic Regression Analysis. ROC based on the binary logistic regression model were used to determine the sensitivity of percentages of the elements in the plasma to predict CN or AD. The AUC of the base model (age, sex and APOEε4 genotype) was 0.643 (Fig. 3A), which was considerably improved by the addition of phosphorus (AUC = 0.930, p = 1.2 × 10 −5 ; Fig. 3B) or calcium (AUC = 0.927, p = 1.6 × 10 −5 , Fig. 3C). The AUC also improved with addition of iron, copper or selenium (AUCs: 0.863, 0.912 and 0.937; p = 1.96 × 10 −3 , 6.00 × 10 −5 and 6.00 × 10 −5 , respectively; Figs 2E,F and 3D). Zinc had no effect on the base model AUC (p = 0.120, Fig. 3G), while the phosphorus/calcium ratio significantly improved the AUC (0.926, p = 2 × 10 −5 , Fig. 3H). Table 2. Validation of Total Reflectance X-Ray Fluorescence (TXRF) measurements of plasma elemental concentrations using Seronorm ™ Level 1 and 2 standards, alongside lower limit of detection (LLD, 3σ detection) and method detection limit (MDL) for each element. Values are mean ± standard deviation (mg/l). [~Reference value measured by inductive-coupled plasma-sector field mass spectrometry]. ## Discussion We report a pattern of increased percentage of phosphorus but decreased calcium, iron, copper, zinc and selenium in the plasma from AD subjects compared to CN. High absolute levels of phosphorus and calcium were best able to discriminate between demented and non-demented cases, especially phosphorus. We demonstrate strong interdependencies between plasma levels of phosphorus, calcium, iron, copper, selenium and zinc. These findings extend and support previous studies that suggest the involvement of metal dyshomeostasis in AD but that there appears an even greater contribution from those systems regulating phosphorus and calcium metabolism. Non-parametric test (Mann-Whitney U test) was performed to determine differences between plasma levels of elements between cognitively normal (CN) and AD subjects, with elements expressed as a percentage of all the elements but also with @ exclusion of phosphorus and calcium. A p < 0.05 was considered significant. (NA, not applicable). The higher plasma phosphorus (both absolute and percentage levels) observed in AD patients in our study is consistent with a recent study showing high serum phosphate was associated with increased risk of AD, vascular dementia and Lewy Body dementia 6 . Binary logistic regression modelling demonstrated the importance of phosphorus to distinguishing AD from CN subjects. Dementia is a recognised comorbidity with chronic kidney disease in which serum phosphate is often raised 15 . Accompanying elevated plasma phosphate is an increase in fibroblast growth factor-23 (FGF-23) which itself is associated with cerebral small vessel disease that strongly drives cognitive decline and dementia 16 . Serum phosphate and FGF-23 can be increased by increasing dietary intake 17 ; food, especially processed foods, commonly contain phosphate additives 18 . High phosphorus intake itself is also associated with increased cardiovascular risk and mortality 19 . Interestingly, high serum phosphate has been shown to be independently associated with the pro-inflammatory cytokine, IL-6 20 , with high phosphate intake inducing inflammation and increased serum TNFα in uremic rats 21 . Increased peripheral pro-inflammatory cytokines is characteristic of AD 22 and high circulating phosphate may contribute to the increased inflammatory milieu in AD. Whether reducing phosphorus intake and lowering blood phosphate levels attenuates dementia remains to be investigated. Note that we have assumed that the elemental phosphorus measured by TXRF is mainly from phosphate, which is the usual form of phosphorus assessed by conventional clinical chemistry methods. Increased plasma calcium in AD is concomitant with the increased phosphorus in our study, although decreased if expressed as a percentage of the panel of six elements which illustrates the difficulties in interpreting elemental concentrations. Alternatively, expressing the ratio of phosphorus/calcium may aid interpretation, with the ratio being higher in AD than CN. Calcium dyshomeostasis has been consistently implicated in AD 23 . Serum calcium levels are modulated by APOE, appearing to increase neuronal calcium influx to increase Aβ and associated with worse cognitive function in APOEε4 carriers, but this interaction is lost in APOEε2 carriers 24 . However, calcium changes can be mediated by non-APOE mechanisms. Plasma calcium levels are regulated by vitamin D and PTH. Hyperparathyroidism, characterised by elevated PTH, with high or elevated calcium, usually with low vitamin D levels, has been associated with impaired cognitive function and dementia 25 . Elevated blood phosphate and calcitriol (synthesized from vitamin D) deficiency may increase PTH secretion 26 and explain the higher plasma calcium in AD observed in this study. Lower percentage of plasma selenium in AD in this study is, consistent with previous studies. However, absolute levels were comparable between AD and CN or when comparing percentages calculated with exclusion of phosphorus and calcium, and explains the lack of consistency in alterations in blood selenium levels in AD 27 . Inclusion of selenium as a percentage of all elements studied gives the highest sensitivity and specificity to AD in this study. Selenium is often a cofactor in antioxidant molecules/pathways and the low peripheral levels in AD observed may contribute to increased susceptibility to oxidative stress, a major aspect of AD pathology. Indeed, a selenium and/or vitamin E supplementation trial had been undertaken to determine if supplementation can attenuate AD 28 . AD was not attenuated/prevented by selenium supplementation, but the study was underpowered, only involved men, and supplementation was of relatively short duration. However, selenium status has been shown to be determined by local geology and if sufficient, selenium has little impact on AD 29 . Zinc has been reported to be decreased or unchanged in the serum and in CSF of AD patients 30 but decreased serum zinc has been proposed to be due to ageing rather than AD 31 . Additionally, a recent ICP-MS study suggested serum zinc is increased with subjective memory complaint but decreased in MCI and AD 14 . Absolute zinc levels were comparable between CN and AD in this study but was significantly lower in the latter when normalised to the total concentrations of the panel of six elements. Conversely, zinc was higher in AD than CN if the percentage was calculated with the exclusion of phosphorus and calcium. The strong interdependences observed here and in other studies 3,32 may contribute to the disparate findings in studies which usually only measure absolute levels of one or two metals. By considering panels of elements/metals rather than one or two of them, variations in exposure and nutrition may be accounted for somewhat (see below). Anemia and attenuated hemoglobin levels are associated with an approximately two-fold increased risk for developing AD, establishing anemia as a risk factor for cognitive loss 8,33 . Significantly lower percentage plasma iron observed in the present study could arise due to deficient iron-loading into transferrin 8,34 by dysfunctional ceruloplasmin, a ferroxidase that facilitates iron loading into transferrin 35 , reported in serum and CSF of AD (and PD) subjects 36 . Low peripheral iron levels may compromise hemoglobin production, leading to a significant decline in hemoglobin levels in AD 8 . Previous evidence has implicated transferrin and its receptor alleles as risk factors for AD 37 . Copper dyshomeostasis has been implicated in AD, with reduced plasma copper and ceruloplasmin levels correlated to impaired cognitive performance in AD subjects 38 . Overexpression of APP in Tg2576 mice lead to copper deficiency, associated with impaired superoxide dismutase and ceruloplasmin prior to the appearance of amyloid neuropathology 39 . Copper supplementation has been proposed to be beneficial in AD by normalizing reduction of plasma copper 38 . In this study, we show that the percentage of copper is lower in AD than CN. However, absolute copper levels and copper percentages with the exclusion of phosphorus and calcium, were similar in CN and AD and consistent with larger published studies 40 . Ageing is associated with a chronic inflammatory state, both peripherally and centrally and even more so, in AD 21,22 . Inflammation induces hepcidin secretion from the liver which promotes ferroportin degradation to increase iron retention in enterocytes and peripheral macrophages 41,42 . Thus, peripheral inflammation in AD may explain the decreased percentage of circulating iron levels observed in this study. Iron deficiency directly stimulates bone production and cleavage of FGF23 as well as via enhanced renal erythropoietin production 17 . Elevated peripheral inflammatory cytokines will also stimulate of FGF23 production and cleavage 16 and as mentioned above, elevated FGF23 levels are associated with increased risks of cardiovascular disease and cognitive impairment/dementia. Further research is required to understand the true meaning of plasma values in the context of AD. Considering the variability in exposure to environmental and/or nutritional factors (particularly in the context of malnourishment) which are significant variables, it is difficult to make precise conclusions on the liaison between the elements measured and AD. We and others 2,3 , have attempted to account for environmental factors and/or nutritional status to some degree by normalising values to a panel of elements. Changes in peripheral circulating levels of elements may reflect decreased or increased body intake/absorption and/or re-distribution between cellular and tissue compartments, e.g., increased brain accumulation. Our dataset does not include information on dietary intake of the various elements analysed in the study or their excretion from the body; any co-morbidities; and/or levels of proteins that integrate the metabolism of the elements measured. However, our cohort was age-matched, and sex balanced. There is also the concern that plasma mineral and metal measurements may reflect weight loss and malnutrition, common in AD, particularly at end stage disease. Thus, our patient cohort comprise only mild-moderate AD (MMSE ~ 20), where only 3% of individual would be expected to be malnourished 43 . We are unable to establish causality in this observational study but propose pathways worthy of further investigation in AD. In the present study, we demonstrate altered distribution of minerals and metals in the plasma of AD and considerable interdependency in their circulating levels. Future studies should take the nutritional and environmental elemental exposure status into account to more accurately reflect the association between the elements and AD. We propose the plasma signature of high absolute levels of phosphorus and calcium, a high phosphorus/ calcium ratio with a low percentage of iron may be indicative of cognitive impairment and/or AD. Further investigations are needed to determine if this pattern is common to dementia, useful for assessing the prodromal stage of AD, and/or monitoring AD progression.

chemsum

{"title": "Pattern of Altered Plasma Elemental Phosphorus, Calcium, Zinc, and Iron in Alzheimer\u2019s Disease", "journal": "Scientific Reports - Nature"}

hybridized_local_and_charge-transfer_excited_state_fluorophores_enabling_organic_light-emitting_diod

## Abstract: Pure organic emitters with full utilization of triplet excitons are in high demand for organic light-emitting diodes (OLEDs). Herein, through modulation of electron donors and introduction of phenyl rings as p spacers, we present three pure organic fluorophores (BCz, BTCz and BPTCz) with the hybridized local and charge-transfer (HLCT) excited state feature for OLED fabrication. Importantly, the introduction of p spacers in BPTCz not only enhances locally excited character with a fast radiative decay but also promotes intermolecular interactions to suppress non-radiative decays, contributing to a high solid-state fluorescence efficiency over 90%. Significantly, BPTCz not only endows its doped OLEDs with an external quantum efficiency (EQE) up to 19.5%, but also its non-doped OLED with a high EQE of 17.8%, and these outstanding efficiencies are the state-of-the-art performances of HLCT-based OLEDs. ## Introduction The exploration of high-efficiency organic emitters has drawn much attention due to their crucial applications in organic light-emitting diodes (OLEDs) since the frst work reported by Tang and co-authors. 1 In OLEDs, the recombination of holes and electrons can form singlet and triplet excitons in a ratio of 1 : 3 according to spin statistics. 2 However, approximately 75% triplet excitons decay through non-radiative processes because of the spin-forbidden transition between the singlet and triplet, leading to the limitation of the external quantum efficiency (EQE) to 5% in conventional fluorescent devices. Therefore, the effective utilization of triplet excitons in organic emitters is very essential for constructing high-efficiency OLEDs. Recent efforts have paved a way to utilize the triplet excitons in thermally activated delayed fluorescent (TADF) molecules, 3,4 which possess a narrow singlet-triplet energy splitting (DE ST ) because of the charge transfer ( 1 CT) feature. The up-conversion with a spin-flip process of the lowest dark triplet (T 1 ) to the lowest emissive singlet state (S 1 ) is activated via thermally assisted reverse intersystem crossing (RISC), leading to a maximal EQE with a breakthrough of the spin statistical limitation. Nevertheless, the frst-order coupling between 1 CT singlet and 3 CT triplet is inefficient in TADF molecules 5 (Fig. 1a), which results in slow up-conversion and thereby efficiency roll-off in TADFbased OLEDs. In contrast, the coupling between 1 CT and the locally excited-state ( 3 LE) is much more effective due to the large spin-orbit coupling (SOC). Consequently, the second-order coupling with an intermediate state ( 3 LE) is introduced to achieve fast RISC rates in TADF molecules (Fig. 1b) which contain multiple donor (D) and acceptor (A) moieties. 6 In order to take full advantages of LE and CT excitons, Ma and colleagues have established a distinctive type of organic emitters in which the lowest singlet state exhibits a hybridized local and charge-transfer (HLCT) feature. 7 In this excited system, the RISC process occurs at upper excited states via a "hot exciton" channel (Fig. 1c). Due to the strong coupling between the 1 HLCT state and 3 LE or 3 CT, the RISC is promoted with a very fast rate which avoids long-lived components and thus benefts the reduction of efficiency roll-off in OLEDs. Additionally, the radiative rate of the HLCT state is much higher than the radiative rate of 1 CT in TADF molecules, allowing the radiative process to compete with non-radiative pathways in the frst place. The typical HLCT molecules are composed of D and A segments and thus the well-regulated dihedral angles of D-A can modulate the character of excited states and energy-level arrangement to promote the radiative decays and RISC process. Therefore, the performance of HLCT molecules greatly depends on the ability of donors, acceptors, and p spacers. In order to enable high-performance OLEDs, the building blocks of HLCT emitters should be carefully selected to regulate appropriate electron push-pull strength. Here, we report three D-(p)-A emitters, namely BCz, BTCz, and BPTCz (Fig. 1d), which are based on a combination of benzothiadiazole and carbazole derivative groups. The relationship between the molecular structure and HLCT excitedstate character is well understood through UV-Vis absorption, photoluminescence (PL), transient spectra, and theoretical simulations. Importantly, the influence of p spacers between D and A moieties is also studied, which plays a pivotal role in HLCT emitters. BPTCz with p spacers exhibits fairly efficient RISC of "hot exciton", due to the large oscillator strength between S 1 and S 0 (efficient radiation), narrow energy split of T 2 and S 1 (fast RISC), and large gap between T 1 and T 2 (low interconversion). Therefore, BPTCz affords both doped and nondoped OLEDs with high EQEs close to 20%, which are the highest efficiencies of HLCT-based OLEDs reported so far. ## Photophysical properties The synthetic routes for BCz, BTCz and BPTCz are shown in Scheme S1, † and meanwhile the three compounds possess excellent thermal properties (Fig. S13 and Table S1 †) that are favorable for OLED applications. The UV-Vis absorption spectra and PL spectra of BCz, BTCz and BPTCz flms are displayed in Fig. 2a. The main absorption peak is located at 463, 477 and 430 nm for BCz, BTCz and BPTCz, respectively. In comparison with the PL spectrum of BCz with a PL peak (l max ) of 590 nm, when the tert-butyl groups are introduced into the donor unit, the resulting BTCz demonstrates a spectral red-shift (l max ¼ 614 nm), due to the enhanced donating ability of the tert-butyl groups. However, BPTCz with phenyl bridges displays a spectral blue-shift (l max ¼ 564 nm), in spite of enlarged conjugation. The same spectral shifts are also observed from their UV-Vis absorption and PL spectra in solution state (Fig. S14 †). The variations of the UV-Vis absorption and PL spectra imply that the modifcations of the donor moiety and p spacer can greatly affect the photophysical properties of organic emitters. Furthermore, the intramolecular charge transfer (ICT) feature of the three compounds can be verifed by the bathochromic shifts of their PL spectra as the solvent polarity increases (Fig. S16 †), along with considerably broadened full-wavelength at half-maximum. 8 As for the solid-state BCz, BTCz and BPTCz, their neat flms exhibit a PL quantum yield (PLQY) of 73.0%, 45.5%, and 90.6%, respectively. To better understand the fluorescence behaviors of the three compounds, transient PL measurements were performed. As shown in Fig. 2b and S16-S18, † the BCz, BTCz and BPTCz neat flms demonstrate a single exponential decay feature with a prompt lifetime of 15.0, 14.0, and 6.1 ns, respectively, which is related to the fluorescence emission from a sole excited state. Additionally, temperature-dependent transient decays were also performed. As shown in Fig. S17-S19 and Table S2, † the lifetimes of the emitters are hardly changed while the temperature varies. These results confrm the fluorescence feature of the excited states and thus exclude the TADF process in these molecules. Moreover, the blue shifted emission and shortened lifetime suggest an enhanced LE component of the lowest singlet state in BPTCz when compared with that of BCz and BTCz. 9 Additionally, the radiative decay rate constants (k r ) of fluorescence and nonradiative decay rate constants (k nr ) are estimated from the PLQY and lifetime (Fig. 2c and d). It is found that BPTCz with p spacers not only exhibits a much higher radiative rate than BCz and BTCz, but also a lower non-radiative decay rate, which are benefcial for achieving high-performance OLEDs. ## Theoretical studies In order to examine the excited states of BCz, BTCz and BPTCz, theoretical calculations were conducted by using the density functional theory (DFT) and time-dependent DFT (m062x/6-311g(d)). As plotted in Fig. S20, † the highest occupied molecular orbitals (HOMOs) of the three compounds are mostly distributed on the whole molecules, the lowest unoccupied molecular orbitals (LUMOs) of BCz and BTCz are mainly located on the benzothiadiazole acceptor, while the LUMO of BPTCz is located on the benzothiadiazole core and also extended to the adjacent phenyl rings. Moreover, natural transition orbital (NTO) analyses were also performed to explore the excited states and transition characters in these D-(p)-A molecules (Fig. S21-S23 †). As displayed in Fig. 3a, the S 1 of the three molecules present a hybrid feature of LE and CT, demonstrating their HLCT features of the excited-states. In BPTCz, the S 1 state exhibiting a stronger LE character can render faster radiation decay, as the LE state is more efficiently radiative than the CT state. 10 The much higher oscillator strength (f ¼ 0.8681) of BPTCz is well consistent with the faster radiative decay through experiment (k r ¼ 1.48 10 8 s 1 ). The NTO results demonstrate that the introduction of phenyl rings as p spacers can well modulate the excited state (S 1 ) with more LE component so as to enhance radiative decay and lead to a higher PLQY in BPTCz. In a further set of experiments, the energy-level arrangements and spin-orbit coupling (SOC) matrix elements of the three molecules were studied. As illustrated in Fig. 3b, all the compounds exhibit very low energy levels of T 1 and large energy gaps between T 1 and T 2 , which can suppress interconversion (IC) transition from T 2 to T 1 . Importantly, a small energy difference between S 1 and T 2 is found for the three molecules, which triggers the spin-flip at higher-lying excited states, and thus results in a fast RISC process occurring from T 2 to S 1 (Fig. 3b) to avoid concentration quenching of triplet excitons. 11,12 In detail, BPTCz exhibits the narrowest energy difference between S 1 and T 2 , implying that there is a higher possibility for the occurrence of the RISC process. Meanwhile, BPTCz with p spacers shows the largest energy difference between T 2 and T 1 , suggesting great reduction of the IC rate. Consequently, the RISC process is more efficient to enable exciton distribution on S 1 . It should be noted that all the molecules present a relatively high SOC constant ($0.5 cm 1 ) between S 1 and T 2 , which is likely associated with the various heteroatoms in the molecules. 13 With the help of the narrow gap and large SOC constant, the spin-flip between T 2 and S 1 is very efficient to take full utilization of the triplet excitons. Generally, excited-molecular motions can be suppressed by intra-and intermolecular interactions to render a lower k nr . Firstly, reduced density gradient (RDG) isosurface maps of the molecules were studied to investigate the intramolecular interactions. As illustrated in Fig. 4a, relatively stronger intramolecular interactions are observed in BCz rather than BTCz and BPTCz. To gain further insights into conformation changes of ground state (S 0 ) and excited state (S 1 ) in the molecules, root-mean-square-deviation (RMSD) calculations were carried out to reflect the non-radiation behavior induced by conformation changes (Fig. 4b). 17 In the case of BCz, the conformation changes of S 0 and S 1 are mainly induced by the rotation of the carbazole moieties. As for BTCz with incorporation of tert-butyl units, the conformation changes are mainly contributed by the rotation of carbazole groups as well as the peripheral tert-butyl units. Accordingly, BTCz shows larger conformation changes than BCz. The higher RMSD values indicate the existence of more molecular motions that could induce faster non-radiative transitions (k nr ¼ 1.80 10 7 s 1 for BCz and k nr ¼ 3.89 10 7 s 1 for BTCz through experimental results), leading to a lower PLQY in BTCz. However, BPTCz has a large RMSD value of 0.636 but shows a higher PLQY (90.6%) and a lower k nr (1.54 10 7 s 1 ), which implies that there are some other factors affecting the emission behaviors of BPTCz rather than the intramolecular interactions, such as intermolecular interactions. In this light, the molecular packing in a single crystal was further examined to reconcile the conflict. As shown in Fig. S24, † BPTCz adopts a parallel arrangement as the p-p stacking mode in the crystal, and efficient D-A electrostatic interaction between adjacent molecules improves the intermolecular interaction (Fig. S25 †), which can effectively restrict the molecular excited state vibration and suppress the non-radiative pathways. Meanwhile, the presence of steric hindrance of the tert-butyl groups prevents further molecular close-packing of adjacent D and A moieties, thereby inhibiting the notorious aggregation that causes quenching. In contrast, close p-p stacking (d ¼ 3.352 A) is formed in BCz due to the lack of tert-butyl groups (Fig. S26 †). Nevertheless, the three compounds exhibit obvious intra-and inter-molecular interactions to suppress vibration motions of excited molecules, and thereby restrict non-radiative transitions. As a result, the molecules are endowed with high fluorescence efficiencies that are benefcial for constructing high-performance OLEDs, especially BPTCz with p spacers. ## Device performance Electroluminescence (EL) performances of the three compounds were further investigated. By a thermal evaporation approach, doped OLEDs with the device structure as: ITO/ PEDOT:PSS (30 nm)/MCP (20 nm)/EML (28 nm)/TPBI (40 nm)/ LiF (1 nm)/Al (100 nm) were fabricated, wherein m-bis(N-carbazolyl)benzene (MCP) and 1,3,5-tris(N-phenylbenzimidazol-2yl)benzene (TPBI) work as the hole transporting layer and electron transporting layer, respectively. As for the EML, the emitters were doped into a common host (CBP) with a doping ratio of 40 wt%. The EL performances are summarized in Table 1 and displayed in Fig. 5. As seen from Fig. 5a, the doped BPTCz device achieves a maximal luminance of 33 190 cd m 2 , which is 18 150 and 7680 cd m 2 in the doped BCz and BTCz devices, respectively. Meanwhile, a maximal EQE of 9.3% and 7.3% is obtained in the doped BCz and BTCz devices, respectively. Impressively, a maximal EQE reaching up to 19.5% is realized in the doped BPTCz device (Fig. 5b). To our knowledge, such a high EQE is the highest efficiency of HLCT-mechanism based OLEDs (Table S3 †), which cover HLCT-based OLEDs showing different color emissions. Given that promising studies on HLCT just appeared a few years ago, high-efficiency HLCT emitters are still rare; notably, this work is the frst example to realize high EQEs close to 20% in HLCT-based OLEDs. It is also worth mentioning that the high EQEs afforded by the BPTCz-based devices are even comparable to those of the best-performing TADF-based OLEDs. 4, These results confrm the practicability of HLCT materials as high-efficiency OLED emitters. Moreover, it is found that BCz, BTCz and BPTCz also endow their non-doped OLEDs with excellent device performances (Fig. S29 †), given the fact that fast k r and k RISC of the emitters help to guarantee high efficiency in neat flms. For instance, the non-doped BPTCz device exhibits a maximal EQE as high as 17.8%, while that of the non-doped BCz and BTCz devices is 9.1% and 7.4%, respectively. Therefore, with regard to BCz, BTCz and BPTCz, each compound enables its doped and non-doped OLEDs with comparable device performances, although the doped OLEDs show slightly higher device performance, which is likely due to more effective suppression of exciton annihilations in the doped EML. In brief, these results verify that BCz, BTCz and BPTCz are attractive HLCT emitters allowing both doping and non-doping techniques to fabricate highly efficient OLEDs, especially BPTCz endowing high EQEs comparable to that of high-efficiency TADF emitters. The steric hindrance of tert-butyl groups in molecules can help to enhance molecular solubility and decrease selfquenching of aggregated excitons in neat flms by restricting excited-molecular motions and p-p stacking. 24 Therefore, solution-processed OLEDs based on BCz, BTCz, and BPTCz were also fabricated to evaluate their solution-processability. As displayed in Fig. S30, † the solution-processed BPTCz device achieves a maximal EQE of 6.0%, which is rarely reported in solution-processed OLEDs based on HLCT emitters (Table S4 †). 25,26 With further optimization of device structure such as exploiting appropriate hole transporting materials and host materials, performance enhancement can be expected. ## Conclusions In summary, three pure organic HLCT emitters (BCz, BTCz and BPTCz) were systematically studied through structural modulation to take full utilization of triplet excitons. It was found that, on one hand, the phenyl rings acting as p bridges in BPTCz brought about more LE character and a higher radiative rate; on the other hand, the tert-butyl units in BPTCz contributed to enhanced intermolecular interactions and suppressed nonradiative decays. As a result, BPTCz neat flm exhibited a much higher PLQY (90.6%) than BCz and BTCz. Remarkably, BPTCz enabled its non-doped and doped OLEDs with an impressive maximal EQE of 19.5% and 17.8%, respectively, which are the highest efficiencies of HLCT-mechanism based OLEDs reported so far. The excellent device performances confrm the effectiveness of our design strategy, and also the great potential of HLCT materials as pure organic highefficiency emitters similar to the well-known TADF emitters for OLED applications.

chemsum

{"title": "Hybridized local and charge-transfer excited state fluorophores enabling organic light-emitting diodes with record high efficiencies close to 20%", "journal": "Royal Society of Chemistry (RSC)"}

designing_singlet_fission_candidates_from_donor-acceptor_copolymers

## Abstract: Singlet Fission (SF) has demonstrated significant promise for boosting the power conversion efficiency (PCE) of solar cells. Traditionally, SF is targeted as an intermolecular process, however its dependence on crystal packing makes molecular design difficult. In contrast, intramolecular SF (iSF) enables the exploration of tunable bi-chromophoric systems following well-defined structure-property relationships. In this work, we propose a set of parameters to screen conjugated donor-acceptor copolymer candidates with potential iSF behaviour. We focus our analysis on the E(S1)>2E(T1) thermodynamic condition and on the appropriate charge transfer (CT) character of S1. We map the CT character with respect to the frontier molecular orbital (FMO) energies of the constituent monomers, providing a cost-effective protocol for an accelerated screening of promising iSF donor-acceptor pairs, while minimizing the number of computations. These parameters are applied to a chemically diverse, curated library of 81 truncated dimers of synthetically feasible donor-acceptor copolymers. From our dataset, four candidates are flagged for iSF, two of which were previously experimentally reported. This protocol is envisioned to be scaled up for the high-throughput screening of large databases of donor-acceptor dimers for the design and identification of conjugated polymers capable of iSF. ## INTRODUCTION First described in 1965, singlet fission (SF) is the spin-allowed conversion of a high-energy singlet to two lower-energy triplets. 1 To be energetically possible, the excited singlet energy needs to be at least twice that of the triplet (i.e. 𝐸(𝑆 ) ≥ 2𝐸(𝑇 )). By definition, SF is a multi-excitonic process: upon the absorption of light, the absorbing singlet splits into two independent triplets (T1) through a correlated triplet-triplet pair ( 1 TT) according to the following scheme: 2 ## 𝑆 + 𝑆 𝑆 + 𝑆 ⇄ 𝑇𝑇 ⇄ 𝑇 + 𝑇 In organic photovoltaic devices, this theoretically leads to a doubled photocurrent if both excitons are separated at a donor-acceptor interface. In this way, materials exhibiting quantum efficiencies above 100% and power conversion efficiencies (PCE) beyond the thermodynamic (Shockley-Queisser) limit of 33% become accessible. 3 Figure 1. Mechanisms for SF after absorption: direct S1 to 1 TT conversion following the blue arrows, or indirect conversion mediated by charge transfer (CT) states following the brown arrows. SF involves two centers: following singlet excitation in one, there is energy transfer to the second, such that one triplet is formed at each center. 2 Both direct 4 and charge-transfer mechanisms have been proposed for this (Figure 1). The formation of the triplet pair can proceed through either an intermolecular or an intramolecular process. In the former case, the centers are located on two separate molecules, while in the latter the two centers are covalently bound. Intermolecular SF has been extensively studied in molecular crystals. 1, However, its success highly depends on the coupling between the separate units, 13 which ultimately relies on the molecular packing, and as such can be difficult to predict and control. 2,4 This limitation is circumvented in intramolecular SF (iSF). 14,15 Such is the case of covalently-linked dimers, in which synthetic modification of the linking units allows fine-tuning the spatial orientation between the sites. However, precisely due to the proximity of the two implicated fragments, the triplets in these systems recombine quickly and rarely become independent. Molecules with extended conjugation, such as polyenes and carotenoids, have also shown iSF, 20 but their large structural flexibility makes nonradiative decay pathways readily available. A few studies have demonstrated iSF in conjugated polymers, particularly in donor-acceptor copolymers, leading to some very promising candidates. On the one hand, Busby et al. designed a poly(benzodithiophene-alt-thiophene-1,1-dioxide) (BDT-TDO) copolymer with a triplet quantum yield of 170%, which highlighted the importance of i) strong intramolecular donor-acceptor interactions, and ii) an acceptor core with a low triplet energy. On the other hand, Zhai et al. reported SF character in thin films of poly(phenylene-alt-vinylene) albeit not in solution, indicating that for certain polymers SF may involve inter-chain processes. 23 Given the limited number of copolymer-based materials undergoing iSF reported so far, clear performance trends could not be established. To date, research exploring the mechanisms of SF has been restricted to the small number of materials in which this process was experimentally observed. 2,14,15,26 Computations of iSF in polymers have been done retroactively to rationalize SF reported in existing materials, 21,27 but there has been a lack of effort to locate new iSF copolymer materials using computational tools. 28 Only very recently a computational screening of intermolecular SF candidates, based on crystal structures, has been reported. 26 Certainly, the discovery of novel iSF systems will largely benefit not only from large-scale screening but also from the development of new molecular design principles. In this work, we take advantage of the well-established modular chemistry of conjugated polymers, and their demonstrated potential for iSF, to explore their chemical space using computational screening techniques. Through systematic modulation of the donor and acceptor units in truncated dimers, we sought to establish design rules that link the monomer and dimer characteristics to the iSF potential of the resulting copolymer. In this way, we provide an accelerated computational screening framework that allows to explore a wide range of potential conjugated copolymers from in silico donor-acceptor combinations. From a curated database of 81 systems, we identify four promising iSF candidates; in two of these iSF has been previously reported. 21,24 Altogether, we discuss both the rational and large-scale strategies of molecular design that will enable the discovery of new iSF materials. ## Criteria to achieve iSF and Design Strategy The main conditions that SF candidates need to fulfill are the following: (1) the energy of S1 is greater than or equal to twice the energy of T1: 𝐸(𝑆 ) ≥ 2𝐸(𝑇 ); 28 (2) the coupling between the two chromophores involved is strong, in order to promote 𝑆 → 𝑇𝑇 ; 14 and (3) the correlated triplet pair ( 1 TT) must evolve into two independent triplets (T1) that can physically separate from one another and escape recombination. These criteria are referred to as the (1) energetic, (2) coupling and (3) separation criteria, respectively. Within the framework of donor-acceptor copolymers, the design strategy consists in combining a donor core, which acts as the main photon absorption site and whose S1 has a dominant CT contribution to the acceptor, with an acceptor featuring a low lying triplet state (see Figure 2). 24 In this way, the strong CT character of S1 is expected to promote an efficient splitting to 1 TT (coupling criterion), while the spatial separation between the two triplets on nearby acceptors, separated by the donor unit, is expected to diminish the possibility of triplet-triplet recombination (separation criterion). ## Database construction Our database includes nine donors and nine acceptors that are commonly found in the literature of conjugated polymers. These contain cyclic, fused and bridged derivatives of thiophene, benzene, pyrrole and other heterocycles (see Figure 3). Wellestablished chemical motifs were prioritized to ensure that potential SF candidates that emerge from this database are synthetically feasible, as well as units that are amenable to multiple polymerization techniques and that can be synthesized with high atom economy in few steps. Units reported in previous works to display iSF in conjugated polymers were included: thiophene-1,1-dixoide (TDO) 24 , benzodithiophene (BDT) 22 , phenylene 23 , vinylene 23 (in the form of (E)-2-(2-(thiophen-2-yl)vinyl)thiophene), TVT), cyclopentadithiophene (CPDT) 21 , benzothiadiazole (BT) 21 and isoindigo (iI). 25 Each donor and acceptor core was encoded as a SMILES string. 35 The dimer set was generated by linking the nine donors with the nine acceptors through a covalent carbon-carbon bond to form the 81 donor-acceptor pairs. The resulting SMILES strings of the dimer were then converted to cartesian coordinates using the gen3d operation in OpenBabel, 36 which includes a conformational search and a geometry optimization at the force field level. Tighter convergence criteria were then applied by reoptimizing the geometries at the Density Functional Theory (DFT) level. The full method used for dataset construction is detailed in section S1, and all data are made available in the Materials Cloud repository. Common abbreviations for the cores, when available, are given in brackets. Structures discussed in section 3.1 are highlighted in red and blue. ## Computational Details DFT computations were performed using the Gaussian16 package (Revision A.03). 37 Ground state geometries were optimized at the ωB97X-D/6-31G* level of theory. Normal mode analysis confirmed that the stationary points were minima with all real frequencies. To evaluate the energetic criterion, vertical excitations and excited state geometry optimizations were computed using time-dependent density functional theory (TD-DFT), within the Tamm-Dancoff approximation (TDA) to correct for the triplet instabilities reported in TD-DFT. 38 The range-separated hybrid functional ωB97X-D was used, given its accurate treatment of excited states, in particular with respect to its description of charge transfer character. 39,40 Full details of functional and basis set benchmarking are given in section S2. For computations with solvent, the solvent cavity reaction field (SCRF) was used with a conductorlike polarizable continuum model (CPCM), which was found to give nearly equivalent results to the SMD continuum model (see section S2.3 for comparison of solvent models). Gaussian output files were parsed with cclib 41 and TheoDORE (version 1.7.2) 42,43 to assess the coupling and separation criteria by means of the local and CT character of the electronic transitions. This is done through the 𝛺 values, which quantify the amount of hole (h + ) and electron (e -) transition density located in the different molecular fragments (i,j). In the present case, we considered two fragments: the donor (D) and the acceptor (A) cores of the donoracceptor dimer. Accordingly, the electronic transition is decomposed into a matrix containing four omega values, in which the diagonal elements (Ω → and Ω → ) quantify intra-fragment contributions (i.e. the hole and electron are formed on the same fragment, i=j), while the off-diagonal elements are the charge-transfer components (i≠j), in which the electron density is transferred from the donor fragment to the acceptor fragment (Ω → ) or vice versa (Ω → ). For each transition, the sum of the four omega values is 1. ## RESULTS The Results are presented in four sections. In section 3.1. we analyse the excited state characteristics of the BDT-TDO copolymer, which previously exhibited iSF. In section 3.2 and 3.3 we establish specific numerical thresholds to efficiently screen the energetic, coupling and separation criteria (section 2.1) from a curated dataset of 81 donoracceptor dimers (section 2.2). Finally, in section 3.4 we map the CT character of S1 with respect to the frontier molecular orbital (FMO) energies of the constituent monomers, and test it on 25 substituted bithiophene-benzothiadiazole pairs. ## BDT-TDO copolymer Copolymers made of thiophene-1,1-dioxide (TDO, shown in red in Figure 3) acceptors, and benzodithiophene (BDT, shown in blue in Figure 3) donors have shown good SF quantum yields and triplet pair lifetimes. 24 Being a prototypical SF copolymer with excellent properties, we selected it as a representative test-case to establish a costeffective computational strategy to evaluate iSF design criteria (section 2.1). Vertical and adiabatic S1, T1, T2 and Q1 energies were computed at the ωB97X-D/TZVP level considering a D-A dimer and a D-A-D-A tetramer model of the BDT-TDO copolymer (see Table S3). While the energetic criterion ( 1) is not fulfilled for the dimer model at the Frank-Condon point (∆𝐸 = 𝐸(𝑆 ) − 2𝐸(𝑇 ) = −0.72 𝑒𝑉), this value becomes much closer to zero in the adiabatic picture (∆𝐸 = −0.18 𝑒𝑉), and even slightly positive for the extended tetramer model (∆𝐸 = 0.04 𝑒𝑉). This highlights the impact of adiabaticity in predicting potential SF behaviour, while shows that the dimer-to-tetramer extension has a much less meaningful impact on the energetic criterion. Fragment-based decomposition analysis of hole and electron density in the excited states revealed that charge transfer from the donor core (BDT) to the acceptor (TDO) at the Frank-Condon point is the primary contribution to the S1 excitation (Ω → = 0.46), following the coupling criterion, whereas a local excitation within the acceptor dominates in T1 state (Ω → = 0.42), matching with the separation criterion (see section 2.1). Similar values were obtained for the adiabatic states and, in all cases, the three other possible contributions to excitation character are smaller (see Tables S4-S5). In order to be efficient, SF needs to overcome triplet-triplet annihilation (TTA) paths, that is, recombination of the two T1 states to higher excited states such as T2 or Q1. For the recombination paths to be energetically unfavourable, both 𝐸(𝑇 ) − 2𝐸(𝑇 ) and 𝐸(𝑄 ) − 2𝐸(𝑇 ) should be positive. 14 While low-lying T2 or Q1 states do not necessarily prevent the singlet splitting, they may reduce the rate of SF. The computed adiabatic energy of T2 resulted in 0.4 eV below S1 when evaluated in gas phase conditions. Remarkably, this difference is significantly reduced to 0.1 eV when including polar solvent effects. This diminution originates in the strong CT character of S1, which has negligible contributions to the mainly local T1 or T2 states. Finally, we found that Q1 is consistently above both S1 and T2 in all cases. In summary, our computations correctly predict (1) thermodynamic adequacy, (2) D-to-A CT character as the largest contribution to S1, and (3) T1 being primarily localized on the acceptor in BDT-TDO copolymer. Adiabaticity plays an important role on ∆𝐸 prediction and thus, empirical rules to correct cost-effective vertical energies of D-A dimers are required. In the next section, we exploit this approach using a curated database of D-A dimers. ## Excited States Energies Threshold for the energetic criterion. We sought to establish a computationally efficient method to evaluate ∆𝐸 (i.e. the energetic criterion), which bypasses the structural optimization of S1 and T1. To do so, the S1 and T1 energies of 81 donor-acceptor dimers in our dataset (see section 2.2) were evaluated at the S0 geometry and at their excited state minima to establish an empirical trend. We found that the relationship between the vertical and adiabatic energies for both S1 and T1 is linear (see Figures S9 and S10), and thus the correlation between the vertical and adiabatic ∆𝐸 is also linear (see Figures 4 and S11). The vertical T1 energies are found to be consistently higher than those obtained from adiabatic computations. As a result, all dimers with ∆𝐸 > 0 𝑒𝑉 in adiabatic computations are also above −1 eV when computed vertically (shown as dotted lines in Figure 4). It is therefore possible to approximate the energy conservation criterion computed adiabatically to: This provides a simple and cheap method to estimate ∆𝐸 from vertical computations by means of systematically shifting the threshold value corresponding to the energetic criterion. It is noted that the linear relationship between the vertical and adiabatic values fails when ∆𝐸 is below −2 eV. However, this will not bias our identification of potential iSF candidates based on this criterion, as this loss of correlation occurs well below the established threshold of −1 eV. To assess if dimer models are representative of larger oligomeric (and polymeric) systems, we evaluated the vertical and adiabatic excited state energies of a subset of 21 tetramers. We considered systems that span the entire range of excitation energies and omega values, with each donor and acceptor moiety represented at least once. The correlation between the dimer and tetramer vertical excitation energies is linear for the entire range of excitation energies and the y-intercept is close to zero (see Figures S14-S15). Consequently, negligible deviations in ∆𝐸 and ∆𝐸 are obtained (see Figures S16-S17). Dihedral analysis of the conjugated dimers and tetramers. An important structural feature of conjugated copolymers is the degree of coplanarity and how it is affected in the excited state. To address this question we evaluated the variation in the dihedral angle (φ) between the donor and acceptor units upon excited state geometry optimization. Comparison of the dihedral angles in the S0 and S1 minima reveal little changes upon adiabatic relaxation (see Figure S22), other than the dimers becoming slightly more coplanar (φ → 0° or 180°) at the S1 minimum in most cases. 44 Consequently, the effective conjugation and character of the excited states remains very similar in the vertical and adiabatic pictures (see Figures S12-S13), indicating that the excited states character of the D-A dimers can be efficiently captured through vertical excitation computations, which are addressed in section 3.3. The dihedral angle between the central donor and acceptor units in the tetramers is virtually equivalent to those in the analogous dimers (see Figure S23). Interestingly, for most materials the change in this dihedral angle upon excited state optimization in S1 is smaller in the tetramer than in the dimer (see Figure S24). This observation suggests that the polymer chain extension restricts the dihedral angle change induced in the excited state. As a consequence, very similar excited state character is obtained in the adiabatic and vertical excitations (see Figures S18-S19), strengthening the assertion that vertical computations at the S0 geometry describe accurately the expected behaviour of larger adiabatic systems. Comparison between the excited state character of S1 and T1 in the dimer and tetramer shows only small deviations while following a nice linear trend (see Figures S20-S21). Considering the structural diversity of our library, these results demonstrate unequivocally that the dimer model, while representing a computational simplification, nonetheless offers an accurate picture of the excited state behaviour and iSF potential of larger donoracceptor polymeric systems. Dependence of energy splitting on monomer structure. The computed ∆𝐸 and ∆𝐸 values of the 81 D-A dimers are represented with respect to their acceptor unit in Figure 4. Considering the nine sets of same-acceptor pairs, the dimers can be classified into two categories: those in which the energetic criterion mainly depends on the acceptor unit (acceptor-dependent), and those which have a broad distribution of ∆𝐸 depending on both the donor and acceptor constituents (donor-tuning). In the former category, all dimers containing the same acceptor (DPP, iI, TDO and NDI) have approximately the same energy splitting values regardless of the donor. In the latter category are the dimers containing BT, F4, TPD, bithiazole, and BDO acceptors, for which certain donors modulate the excited state energy levels towards favourable splitting. In particular, the donors TVT, CPDT, 2,2'bithiophene, thienothiophene and BDT, which all include thiophene moieties, shift ∆𝐸 to more positive values, while the donors that do not have thiophene motifs (Cbz, fluorene and phenylene) are systematically detrimental to the energy conservation condition. Smaller values of ∆𝐸 originate in non-planar dihedral angles between the donor and acceptor units in dimers linked via a benzene ring. This leads to a weaker effective conjugation and, generally, to higher T1 excitation energies. A representative case is shown in Figure 5 for the TPD acceptor. The six donors that are coplanar with the acceptor (φ ≈ 0°) can form intramolecular chalcogen bonds promoted by the thiophene unit, and the resulting dimer possess vertical energy splitting values between −0.78 eV and −1.17 eV. In contrast, the D-A dimers that do not contain thiophene motifs in the donor exhibit higher dihedral angles and more negative ∆𝐸 values. ## Excited State Character In order to identify how the excitation energies and thus, ∆𝐸 , are affected by the different state character of S1 and T1, we performed a fragmentbased analysis of the main local and CT contributions. We focused on the donor-to-acceptor CT component of S1 (Ω → ) and on the local acceptor contribution of T1 (Ω → ) as key requirements for efficient iSF that will potentially favour singlet splitting and prevent fast TTA, respectively (criteria 2 and 3, section 2.1). These are represented for the 81 dimers in Figure 6. The dimers BDT-TDO (discussed in section 3.1) and CPDT-BT, for which iSF has been reported, 21,24 fulfill both criteria, with the key contributions (Ω → and Ω → ) both above 0.4. For that reason, we tentatively select 0.4 as threshold for screening purposes. Remarkably, all other dimers found above this threshold have BT as acceptor, which systematically generates very promising candidates for iSF. In fact, the classification into acceptordependent and donor-tuning D-A dimers discussed for the energy splitting values remains valid for Ω → and Ω → . In particular, DPP-and iI-dimers, which systematically show positive , lead to large Ω → and small Ω → in all cases (due to large Ω → values). Therefore, the dimers involving these acceptors will not undergo iSF, but most likely inter-chain SF based on local acceptor states. Large singlet-triplet energy splitting has been previously associated with local excitations in organic systems. 45 However, new design principles need to be considered when evaluating the iSF capabilities of D-A copolymers. From our results, it is possible to envision a 'modular' design strategy based on the frontier molecular orbital (FMO) energies of the donor and acceptor units to screen the CT or local character of S1 in the dimer. ## Charge Transfer Prediction In this section, we correlate the CT character of S1 in the 81 D-A dimers with the FMOs of the 18 constituent monomers (collected in Table S6). This is represented schematically in Figure 7, where it is shown how the CT excitation (D→A) competes with local excitations in either the donor or the acceptor depending on the relative ordering of the FMOs. Within this approximation, the ratio between the local orbital gaps and the resulting CT energy difference defined as and will estimate favourable (>1) or unfavourable (<1) CT excitations with respect to local excitations. The approximation of considering orbital gaps as oneelectron transitions is possible because the relationship with the local excited energies is rather linear (see Figures S25 and S26). This shows that the exciton binding energy, defined as the difference between the orbital gap and the excitation energies, is relatively constant for all donors and acceptors considered. Figure 8 associates the computed FMO ratios with the computed Ω → values. It can be seen that the donor-acceptor monomer pairs with FMOs best suited for CT are located above 1.0 in both axes. In contrast, dimers with FMO ratios below 1 correctly predict minor CT character (Ω → <0.2). This numerical comparison using monomer FMO energies is therefore a robust metric for eliminating poor potential iSF candidates. To illustrate the direct impact of monomer FMO energies on Ω → , we generated 25 substituted bithiophene-BT donor-acceptor pairs. The monomer energy levels are substantially modulated through functionalization of the conjugated backbone with electron-donating (-OH), electron-withdrawing (-CN) and halide (-F, -Cl) moieties. As a result, the CT character of S1 in the dimer (Ω → = 0.44 when unsubstituted) becomes as low as 0.26 when an electron-withdrawing group is placed on the donor unit, and as high as 0.75 when electron-donating (withdrawing) groups are attached to the donor (acceptor) moiety. These variations are correctly captured by the FMO ratio model (Figure 8), which reproduces the general increase of Ω → as the FMO ratio increases. This example demonstrates that monomer functionalization can be used to optimize the properties necessary for iSF. ## SCREENING PROTOCOL The protocol established to evaluate and screen promising iSF candidates among donor-acceptor copolymers consists of the following steps: Step 1. Compute the ground state FMOs of all donor and acceptor monomer cores, and evaluate the FMO ratios for the donor-acceptor dimers according to the expressions Step 2. For the candidate donor-acceptor combinations resulting from step 1, generate the dimers and compute the vertical S1 and T1 excited energies. Apply the energetic criterion associated with vertical energies: Step 3. Determine the character of the vertical S1 and T1 states, and apply the threshold for criteria 2 and 3 based on the following omega values: 9) for deeper study. Remarkably, two of these four candidates have already been reported to undergo iSF (BDT-TDO and CPDT-BT), supporting the convenience of this protocol. 21,24 Note that the cut-off values used here were established at the ωB97X-D level of theory, and that these parameters may shift depending on the functional used. ## CONCLUSION We have developed a cost-effective computational protocol to perform large-scale screening of donoracceptor copolymers with promising features for intramolecular singlet-fission. Using a structurally diverse database of donor and acceptor units, we have established a simplified yet robust computational strategy to evaluate the energy splitting criterion and the charge-transfer requirements of the D-A candidates from conventional vertical excited state computations. In the context of accelerated screening, we have proposed an expression to predict the excited state character of D-A dimers from the FMO energies of their constituent donor and acceptor units. This drastically reduces computational time in initial screening stages, as the number of computations is reduced from ND*NA to ND+NA, ND and NA being the number of donor and acceptors considered. This protocol correctly identified two donoracceptor pairs that exhibit singlet fission behaviour experimentally, 21,24 and proposed two promising new candidates, thieno [3,2-b]thiophene-BT and 2,2'bithiophene-BT, that have not been studied to date. Benzothiadiazole (BT) in particular shows promise as an acceptor unit in iSF D-A materials. Altogether, these findings pave the way for high-throughput screening of large, chemically diverse databases of D-A conjugated polymers as a mean to bolster the collective library of SF materials. ## ASSOCIATED CONTENT Details of database construction, benchmarking results, and supplementary figures pertaining to excited state character, tetramers, monomer bandgaps and excitonic effects are made available in the Supporting Information. ## DATASET Raw data used to generate all figures (dimer excitation energies and state character, monomer FMO energies and excitation energies, substituted monomer FMO energies, and substituted dimer excitation energies and state character) will be made available in a Materials Cloud repository upon publication. The collection of all output files from Gaussian, Turbomole and TheoDORE computations is available at the same location.

chemsum

{"title": "Designing Singlet Fission Candidates from Donor-Acceptor Copolymers", "journal": "ChemRxiv"}