Datasets:

medarc

/

clinical_pile_v1_minhash_deduped

like 1

Follow

MedARC 16

ambient_synthesis_of_an_iminium-linked_covalent_organic_framework_for_synergetic_rna_interference_an

## Abstract: Small interfering RNA (siRNA)-mediated gene silencing is a promising therapeutic approach. Herein, we report the ambient synthesis of a positively charged iminium-linked covalent organic framework by a three-component one-pot reaction. Through anion exchange and siRNA adsorption, the resulting multifunctional siRNA@ABMBP-COF, which possesses both the HK2 inhibitor 3-bromopyruvate and SLC7A11 siRNA, exhibits powerful synergistic antitumor activity against fibrosarcoma via the ferroptosis and apoptosis pathways. ## Introduction Small interfering RNAs (siRNAs) are powerful laboratory tools that can specifcally inhibit targeted gene expression. 1,2 The clinical translation of siRNA therapeutics is limited due to their negative charge, high molecular weight (approximately 14 kDa), ease of degradation, and low transmembrane uptake. 3 Various delivery vectors, including viruses, 4 proteins, 5 liposomes, 6 polymers, 7,8 metal-organic frameworks, inorganic nanoparticles, 12 and extracellular vesicles, 13 have been developed to transport siRNAs into cells. However, the limited loading amount, insufficient lysosome escape, and difficulty in synergizing with other therapeutics greatly hinder their use in tumour treatment. 14,15 Therefore, designing next-generation vectors for efficient siRNA delivery is urgent and important. Since the pioneering work of Yaghi et al. in 2005, 16 covalent organic frameworks (COFs), which are a class of porous materials, have shown great potential in drug delivery, protein encapsulation, phototherapy, and immunotherapy. In principle, COFs can adsorb nucleic acids to generate nucleic acid@COFs for oncotherapy. However, nucleic acid@COFs have never been used in antitumor treatments, which might also result from extremely low nucleic acid loading. We hypothesize that this bottleneck could be overcome by synthesizing positively charged COF-based carriers in which the loading amount of negatively charged therapeutic siRNA could be signifcantly improved via electrostatic interactions. Furthermore, the counterions in cationic COFs could be replaced with negatively charged metabolic inhibitors and chemotherapeutic drugs via ion exchange. 39 Thus, multifunctional COF-based siRNA delivery and metabolic therapy could be logically achieved. To date, the reported cationic COFs have been typically synthesized from positively charged monomers, including ethidium bromide, 40,41 propidium iodide, 42 imidazolium, quaternary ammonium, 46 and guanidinium, 47 under harsh solvothermal conditions. This energy-intensive and tedious approach is not conductive to large-scale synthesis. Herein, we report the ambient synthesis of the iminium-linked cationic ABMI-COF via a three-component one-pot reaction (Scheme 1A). Through ion exchange of the iodide counterion with 3bromopyruvate, a hexokinase 2 (HK2) inhibitor, 54 multifunctional ABMBP-COF was generated. Both ABMI-COF and ABMBP-COF possess high siRNA adsorption capacity (greater than 1 nmol mg 1 ) and can escape the lysosome. More importantly, after being loaded with solute carrier family 7 member 11 (SLC7A11) siRNA, 55 siRNA@ABMBP-COF could silence SLC7A11 and inhibit HK2, consequently achieving antitumor effects in vitro and in vivo through ferroptosis and apoptosis (Scheme 1B). ## Results and discussion Inspired by the organic reaction reported by Raston et al., 56 the reaction of 1,3,5-tris(4-aminophenyl)benzene (TAPB), benzene-1,3,5-tricarbaldehyde (BTA), and iodomethane in CH 3 CN with acetic acid produced a 78% yield in ABMI-COF in after 24 h at room temperature (Fig. S1A †). Elemental analysis and inductively coupled plasma-mass spectrometry (ICP-MS) indicated that the obtained ABMI-COF had the molecular formula C 33 H 21 N 3 (CH 3 I) Cite this: DOI: 10.1039/d2sc02297d All publication charges for this article have been paid for by the Royal Society of Chemistry Thermogravimetric analysis (TGA) showed that ABMI-COF was thermally stable up to approximately 380 C (Fig. S1C †). The crystal structure of ABMI-COF was determined using Materials Studio software based on the measured powder X-ray diffraction (PXRD) pattern, in which a series of observed peaks at 2q ¼ 5.7 , 9.9 , 11.5 , 15.2 , and 26.1 were assigned to the (010), (120), (020), (130), and (001) facets, respectively (Fig. 1A). The results indicated that ABMI-COF possessed a 2D network with an eclipsed AA stacking mode (Fig. 1B). The Pawley refnement showed a negligible difference between the simulated and experimental PXRD patterns. ABMI-COF was assigned to the space group P3 with optimized parameters of The type I N 2 adsorption-desorption isotherm at 77 K of ABMI-COF showed that the Brunauer-Emmett-Teller (BET) surface area was S BET ¼ 608 m 2 g 1 and the total pore volume at P/P 0 ¼ 0.99 was 0.38 cm 3 g 1 , confrming its porosity (Fig. 1C). The pore size distribution was determined by nonlocal density functional theory (NLDFT) analysis and indicated that it possessed a narrow pore diameter distribution centred at approximately 1.1 nm, which was consistent with the simulated structure. The formation of ABMI-COF was also confrmed by spectroscopic methods. The Fourier transform infrared spectrum showed the characteristic peak of C]N + at 1666 cm 1 , and the appearance of peaks at 1248 and 1197 cm 1 were due to C-N + (Fig. S1D †). The symmetrical and asymmetrical stretching vibrations of the CH 3 group were located at 2872 and 2948 cm 1 , respectively. Weak peaks of residual CHO and C]N were observed at 1697 and 1626 cm 1 , respectively, indicating the presence of bonding defects. 57 The observed carbon resonances in its 13 C solid-state nuclear magnetic resonance spectrum showed that ABMI-COF contained iminium (182 ppm), methyl (53 ppm), and aromatic (100-150 ppm) species (Fig. S1E †). 58 XPS analysis of ABMI-COF in the N1s region was deconvoluted into a C]N + peak at 401.1 eV and a C]N peak at 398.3 eV (Fig. S1F †). 59 Furthermore, two peaks with a well- separated spin-orbit component of 11.5 eV located at 619.2 and 630.7 eV were assigned to the iodide ion (Fig. S1G †). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images showed a uniform spherical morphology of ABMI-COF with a diameter of approximately 230 nm (Fig. 1D and S1H †). The dynamic light scattering (DLS) measurements showed a z-average size of 232.7 nm with a polydispersity index of 0.120 in phosphate-buffered saline, indicating its good dispersion (Fig. S1I †). The iodide counterions within ABMI-COF could be partly exchanged by 3-bromopyruvate in a weakly basic triethylamine solution to generate ABMBP-COF, which had a molecular formula of C 33 H 21 N 3 (CH 3 ) 2.80 I 0.48 (C 3 H 2 BrO 3 ) 2.32 based on ICP-MS and elemental analysis. As shown in Fig. S2, † ion exchange did not cause changes in crystallinity, structure, micromorphology, or dispersibility but resulted in a slight S BET decrease (ABMBP-COF, 567 m 2 g 1 ). Due to the iminium linkage, ABMI-COF and ABMBP-COF had positive zeta potentials of +28.2 and +25.0 mV, respectively, which endowed these nanoparticles with a high capacity to adsorb negatively charged siRNA. Not surprisingly, after adsorbing siRNA, the zeta potential of the nanoparticles decreased to approximately 60% of that before adsorption, while the hydrodynamic diameters based on DLS measurements were almost unchanged, indicating that siRNA adsorption did not lead to signifcant particle coagulation (Fig. S3A †). Furthermore, after adsorption, the fluorescence of surfaceadsorbed Cy3-labelled siRNA (siRNA-Cy3) was effectively quenched via fluorescence resonance energy transfer caused by the spectral overlap between the Cy3 donor emission and COF acceptor absorption (Fig. S3B-D †). Fluorescent quantitative experiments showed that the saturated adsorption capacities of ABMI-COF and ABMBP-COF for siRNA were up to 1.2 and 1.1 nmol mg 1 , respectively, which are signifcantly higher than those of electroneutral COFs (Fig. S3E †). 60 Unsurprisingly, due to the cytomembrane affinity and proton sponge effect caused by the positive charges, 7,61 the obtained siRNA@ABMI-COF and siRNA@ABMBP-COF readily entered HT-1080 cells within 4 h via pinocytosis (Fig. S4 †) and then escaped from lysosomes into the cytoplasm (Fig. 2A). Their transfection efficiencies were superior to those of commercially available polyethylenimine and calcium phosphate and were comparable to those of Lipofectamine 2000 (Fig. S5 †). siRNA@ABMBP-COF, which contains the HK2 inhibitor 3bromopyruvate, can cause oxidative stress and consequent cell death by inhibiting aerobic glycolysis and mitochondrial oxidative phosphorylation. 54 Theoretically, the antitumor effect of 3-bromopyruvate could be further enhanced by blocking the biosynthesis of glutathione (GSH), which is the major intracellular response to oxidative stress. 62 To examine this possibility, SLC7A11, 55,63 a key transporter that is upstream of GSH biosynthesis, was selected as a therapeutic target. siRNAmediated knockdown of SLC7A11 could inhibit cellular uptake of cystine, thereby blocking GSH synthesis and enhancing oxidative stress. 64 According to CCK-8 cell viability assays, siRNA@ABMBP-COF (40 mg mL 1 , COF equiv.) reduced HT-1080 cell viability to 34.5 AE 2.2% compared to the untreated group, which was signifcantly better than siRNA@ABMI-COF (70.3 AE 8.8%), ABMBP-COF (73.8 AE 2.8%), and ABMI-COF (90.5 AE 6.6%), suggesting a combined effect of SLC7A11 siRNA and 3-bromopyruvate (Fig. 2B). Clonogenic analysis was performed, and siR-NA@ABMBP-COF-treated HT-1080 cells had the lowest clone formation compared with the other treatment groups, further supporting the obtained results (Fig. S6 †). The cell death mechanism induced by the cationic COFbased nanodrugs was investigated. After SLC7A11 siRNA was loaded, siRNA@ABMI-COF and siRNA@ABMBP-COF decreased SLC7A11 expression (Fig. 2C and S7 †), which subsequently blocked GSH synthesis. As a result, a series of cellular biological changes were examined at 48 h, including decreases in the GSH concentration (Fig. S8A †), increases in cytoplasmic Fe 2+ levels (Fig. S9 †), reactive oxygen species (ROS) production (Fig. S10 †) and lipid peroxidation (Fig. S11 †), decreases in glutathione peroxidase 4 (GPX4) expression and activity (Fig. 2C and D), increased malonaldehyde concentrations (Fig. S8B †), and mitochondrial membrane potential loss (Fig. S12 †). These results were consistent with the characteristics of ferroptosis. Furthermore, the expression of ferroptosis suppressor protein 1 (FSP1) and acyl-coenzyme A synthetase long-chain family member 4 (ACSL4) was intact (Fig. 2C and S7 †), suggesting that HT-1080 cells triggered ferroptosis via the cyst(e) ine-GPX4-GSH axis. 68,69 Notably, ABMBP-COF treatment contributed to GSH depletion (Fig. S8A †), ROS upregulation (Fig. S10 †), and mitochondrial damage (Fig. S12 †) but did not upregulate malonaldehyde content (Fig. S8B †) or downregulate GPX4 expression (Fig. 2C and S7 †). These results suggested that 3-bromopyruvate induced cell death through an additional mechanism. After treatment with ABMBP-COF and siRNA@ABMBP-COF for 48 h, HK2 activity in HT-1080 cells decreased to less than 30% of that in the control group (Fig. 2D), and caspase 3 activation was detected by immunofluorescence staining (Fig. S13 †), suggesting that the released 3-bromopyruvate induced apoptosis by triggering cellular energy stress. Interestingly, the combination of ABMBP-COF-induced energy stress and siRNA-induced ferroptosis was more effective in reducing GSH and elevating ROS than either treatment alone, emphasizing the distinct advantage of synergistic treatment (Fig. S8A and S10 †). Ferroptosis and apoptosis were further validated by cell rescue experiments in which different functional molecules were added to the media and cultured with siRNA@ABMBP-COF treated HT-1080 cells (Fig. 2E). Ferrostatin-1 (Fer-1)a ferroptosis inhibitor-alleviated cell death, and direct supplementation with raw materials for GSH biosynthesis, such as glutathione ethyl ester (GSH-EE), N-acetyl-L-cysteine (NAC), and 2-mercaptoethanol (2-ME), restored cell viability to varying degrees, 70,71 suggesting that GSH depletion promoted ferroptotic cell death. In addition, pyruvate, which is a fnal product of the glycolytic pathway, unblocked glucose metabolism and restored cell viability, 72 and the apoptosis inhibitor Z-VAD-FMK (zVAD) inhibited cell death, 73 suggesting that energy stress induced apoptosis. The necroptosis inhibitor necrostatin-1 (Nec-1) and the autophagy inhibitor 3-methyladenine (3-MA) had negligible effects on cell viability; thus, necroptosis-and autophagy-related cell death were excluded. 73 Encouraged by the obtained results, in vivo antitumor activity was evaluated using an HT-1080 human fbrosarcoma xenograft model implanted in BALB/c nude mice. Tumours were collected on day 10 after intratumoral injection of the nanodrugs (0.8 mg mL 1 , COF equiv.), and the results showed that the antitumor therapeutic efficacy was enhanced in the following order: ABMI-COF, siRNA@ABMI-COF, ABMBP-COF, and siRNA@ABMBP-COF (Fig. 3A and B). Specifcally, siR-NA@ABMBP-COF reduced the tumour volume to approximately 60% of that before treatment, while siRNA@ABMI-COF and ABMBP-COF exerted worse antitumoral effects, and ABMI-COF had almost no antitumor effect (Fig. 3C). Histopathological analysis of haematoxylin-eosin (H&E)-stained tumour tissues collected at the end of the treatment showed that the histological morphology of siRNA@ABMBP-COF-treated tumours was signifcantly different from that of the control group, as indicated by extensive cell membrane rupture, nuclear contraction, and loosely arranged cells, indicating cellular damage (Fig. 3D). Ki67 is a nuclear antigen associated with cell proliferation and cancer prognosis and is a cellular marker for measuring the proliferative potential of cancer cells. 74 The immunohistochemical staining results (Fig. 3E) showed that siRNA@ABMBP-COF resulted in a lower ratio of Ki67-positive cells than ABMBP-COF and siRNA@ABMI-COF, indicating the suppression of tumour proliferation. These experimental results are consistent with the trend in the tumour growth curve. The ferroptosis inhibitor liproxstatin-1 (Lip-1) clearly counteracted the tumour treatment effect induced by siR-NA@ABMBP-COF (Fig. 3A-E) and the terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labelling (TUNEL) assay demonstrated slight increases in the proportion of apoptotic cells in the groups of ABMBP-COF and siRNA@ABMBP-COF (Fig. S14 †), suggesting ferroptosis-and apoptosis-related antitumor mechanism in vivo. Furthermore, the levels of ferroptosis-and apoptosis-related metabolites and enzymatic activities were determined on day 4. Compared with monotherapy with ABMBP-COF and siRNA@ABMI-COF, the siRNA@ABMBP-COF-induced combination treatment resulted in signifcant intratumoral GSH downregulation and malonaldehyde upregulation, suggesting the presence of oxidative stress and ferroptosis in (Fig. S15A and B †). Additionally, siRNA@ABMBP-COF inhibited intratumoral HK2 activity, but there was no signifcant difference in GPX4 activity (Fig. S15C and D †), suggesting the presence of a compensatory mechanism in solid tumours, 75 which was different from the in vitro observations. Although siRNA@ABMBP-COF achieved an obvious antitumor effect, it is clear that the GPX4-related compensatory mechanism in vivo is unfavourable for tumour therapy. We believe that antitumor therapy can be further optimized by combining GPX4 inhibitors 76 or radiotherapy, 77,78 which will be investigated in the future. The systemic toxicity of the nanodrugs to nude mice during the treatment was negligible, which was confrmed by a lack of signifcant weight loss in mice during the treatment (Fig. 3F) and H&E staining of major organs collected at the end of treatment (Fig. S16A †). Routine blood and biochemical examinations showed negligible adverse effects of the nanodrugs on liver function, kidney function, and the blood system in healthy nude mice (Fig. S16B and C †). Therefore, the nanodrugs have no obvious acute toxicity and have acceptable biosafety. ## Conclusions In conclusion, we reported the synthesis of an iminium-linked positively charged COF by a three-component one-pot reaction under ambient conditions. Through anion exchange and siRNA adsorption, the resulting multifunctional COF-based nanodrug exerts potent combined antitumor activity against HT-1080 tumour cells and tissues via ferroptosis and apoptosis. This study not only enriches COF synthetic methodology but also highlights cationic COF as a promising platform for siRNAmediated combination therapy. ## Synthesis of ABMI-COF The mixture of TAPB (562.3 mg, 1.6 mmol), BTA (259.4 mg, 1.6 mmol), acetonitrile (200 mL), acetic acid (32 mL), and iodomethane (32 mL) was stirred at 800 rpm and 25 C for 24 h. The precipitate was collected by centrifugation at 12 000 rpm (14 800 g) for 30 min at 4 C and washed 3 times with acetonitrile and then 3 times with ethanol. Finally, the precipitate was dried under supercritical CO 2 to obtain ABMI-COF as an orange-red powder. The yield was 1.1 g (78%). ## Synthesis of ABMBP-COF ABMI-COF (50 mg) was dispersed in an aqueous solution (100 mL) containing 3-bromopyruvic acid (83.5 mg, 0.5 mmol) and triethylamine (100 mL, 0.7 mmol). The mixture was stirred at 600 rpm and 25 C for 12 h. The precipitate was separated by centrifugation at 12 000 rpm (14 800 g) for 30 min at 4 C. The dispersion-stirring-centrifugation process was repeated 4 times. The obtained precipitate was washed three times with water and once with ethanol and was dried under vacuum to obtain ABMBP-COF as an orange-red powder. The yield was 50 mg. ## Animal experimentation All animal procedures were reviewed and approved by the Ethics Committee of Shandong Normal University (Jinan, China; application number AEECSDNU2021009). Further information regarding experimental procedures are stated in the ESI. †

chemsum

{"title": "Ambient synthesis of an iminium-linked covalent organic framework for synergetic RNA interference and metabolic therapy of fibrosarcoma", "journal": "Royal Society of Chemistry (RSC)"}

inducing_social_self-sorting_in_organic_cages_to_tune_the_shape_of_the_internal_cavity

## Abstract: Many interesting target guest molecules have low symmetry, yet most methods for synthesising hosts result in highly symmetrical capsules. Methods of generating lower-symmetry pores are thus required to maximise the binding affinity in host-guest complexes. Here, we use mixtures of tetraaldehyde building blocks to access low-symmetry imine cages. Whether a low-energy cage is isolated can be correctly predicted from the thermodynamic preference observed in computational models. The stability of the observed structures depends on the geometrical match of the aldehyde building blocks.One bent aldehyde stands out as unable to assemble into high-symmetry cages-and the same aldehyde generates low-symmetry socially self-sorted cages when combined with a linear aldehyde.We exploit this finding to synthesise a family of low-symmetry cages containing heteroatoms, illustrating that pores of varying geometries and surface chemistries may be reliably accessed through computational prediction and self-sorting. ## INTRODUCTION Controlled host-guest recognition is of crucial importance to biological processes and artificial supramolecular systems alike. Cage-like compounds have been developed to exploit such hostguest interactions to achieve pollutant remediation, gas storage, anion binding, biomimetic guest recognition, and molecular separations. Advantages of organic cage hosts include their improved solubility over framework materials, making them excellent candidates for both liquid-or solid-phase applications. Furthermore, cages offer synthetic handles that can be used to finely tune their cavity shape and electronic properties, and hence potential interactions with guest molecules. [7, However, the synthesis of molecular cages is often challenging, particularly where multiple bonds must be formed selectively. To avoid this problem, many molecular cage syntheses take advantage of dynamic covalent chemistry, in which reversible reactions provide an error correction mechanism to ensure the thermodynamic cage product is obtained. In the case of imine-based cages, multiple amines and aldehydes must react to yield a single cage species instead of oligomeric mixtures of imines. Ideally, high-fidelity self-sorting biases the formation of the cage product over the many other possible products, enabling selective isolation of the target molecule. A limitation of self-sorting strategies is that they often result in the formation of highly symmetrical products. Reducing the symmetry of the host may induce anisotropy in the solid state, improve the binding of low-symmetry guests, or enable more controlled and directional post-synthetic modification. For example, fine-tuning of the cavity of an organic cage has been shown to afford precise control over the selectivity of the resultant solid-state material. Stepwise syntheses exploiting orthogonal reactivities can afford low-symmetry organic cages, but this limits the scalability of the resulting materials. Alternatively, low-symmetry architectures may be obtained by purification of complex mixtures, but this is a laborious process and may be unachievable on a preparative scale due to reconfiguration of the desired products. A recent study showed that low-symmetry cages can be formed using a lower-symmetry aldehyde precursor, but the presence of multiple structural isomers precluded the unambiguous characterisation of the cage products. We sought to avoid these problems by designing an alternative single-step route to low-symmetry iminebased cages. We used mixtures of multiple aldehyde precursors with different geometries to investigate their self-sorting behaviour, screening for combinations that led to the selective formation of low-symmetry cages. We recently reported a series of Tet 3 Di 6 tubular organic cages that were prepared through imine formation between three pseudo-linear tetratopic aldehydes ("Tet") and six ditopic amines ("Di"), and selected the linear tetraaldehydes as a starting point for these studies. Reacting a mixture of two tetraaldehydes with a single diamine can produce three distinct sorting outcomes (see Figure 1(i)): narcissistic self-sorting, in which only cages incorporating a single aldehyde precursor are observed; social self-sorting, in which only cages incorporating both aldehyde precursors are observed; and scrambling, in which a mixture of different sorting outcomes are observed. However, it is extremely hard to predict -either intuitively or computationally -which outcome will be observed for a given pair of aldehyde reactants. For simple cases, such as two linear aldehydes, one could expect that narcissistic self-sorting is likely due to the mismatch in the aldehyde lengths and strain in the resultant cages. [46, It is much more difficult to predict the outcome when a linear aldehyde is combined with a bent aldehyde, such as B1 (see Figure 1(ii)). We hypothesised that the greater conformational degrees of freedom of non-linear aldehydes compared to linear aldehydes would aid social self-sorting or scrambling by accommodating a wider range of options for the cage geometry. Figure 1 (i) Illustration of narcissistically and socially self-sorted systems as opposed to non-sorted scrambled outcomes for an imine-based organic cage forming reaction using linear (green) and bent (orange) aldehydes in presence of (1R,2R)-trans-1,2-cyclohexanediamine (blue); (ii) structures of the bent (B1, B2) and (iii) linear (L1-4) tetraaldehydes used in this work. To test our hypothesis, we studied imine-based cages formed from two bent tetratopic aldehydes B1-2 and four linear aldehydes L1-4 of varying length (see Figure 1(iii)). First, we sought to confirm that all the aldehydes individually form cages with (1R,2R)-trans-1,2-cyclohexanediamine (R,R-CHDA) in the presence of trifluoroacetic acid catalyst. Each bent aldehyde was then reacted sequentially with the series of linear aldehyde partners and R,R-CHDA to assess their cage-forming and self-sorting behaviour. All reactions were characterised by ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) and 1 H NMR spectroscopy. Where cage species could be isolated, crystal structures were sought to confirm their identities and assess their stable conformations. The sizes of isolable low-symmetry cages were further investigated in solution by diffusion-ordered spectroscopy (DOSY NMR) and ion-mobility spectrometry-mass spectrometry (IMS-MS). Aldehyde B1 was found to induce social self-sorting in the studied cages, thus heteroatom-containing analogues of B1 were synthesised and reacted using the same methods to test whether the self-sorting behaviour was retained. We previously used density functional theory (DFT) formation energies to explain the thermodynamically preferred cage topologies in the dynamic imine-based self-assembly processes. [39, Comparing the thermodynamic stabilities of potential cage products can be a good guide to selectivity, but the reaction outcome can also be affected by factors such as reaction kinetics, [37, solvent effects, and the solubilities of the species involved in the equilibrium. In parallel with the synthetic efforts, we used computational techniques to predict the stability of the different homo-and heteroleptic structures originating from aldehydes B1-2 and L1-4. The experimentally observed outcomes agreed with the relative gas-phase formation energies of the possible Tet 3 Di 6 products, showing the predictive power of the simple model for the self-sorting behaviour of imine-based organic cages. ## Single aldehyde systems Aldehydes B1-2 and L1-4 were synthesised via Pd-catalysed cross-coupling reactions (see ESI Section S3 for the synthetic details). The reactions of L2 and L3 with R,R-CHDA have been reported to give tubular covalent cages [3L2] and [3L3], respectively (see Figure 2 for the single crystal structures). The reaction of L1 with R,R-CHDA afforded a complex mixture of imine condensation products that could be purified by recrystallisation to yield [3L1]. Reactions of L4 and B1 with R,R-CHDA both result in single cage products [3L4] and [3B2], respectively. The single crystal structures of [3L1] and [3B2] could be elucidated. Unlike for the other aldehydes, reaction products of B1 with either R,R-or S,S-CHDA could not be identified and attempts to grow single crystals from such reaction mixtures were unsuccessful. However, co-crystallisation of the opposite-handed reaction products led to reequilibration of the building blocks and provided a pseudo-C3h-symmetric [3B1-RS] cage. Investigation of the crystal structure of [3B1-RS] revealed incorporation of equal amounts of each enantiomer of CHDA into the cage structure, which is reminiscent of the CHDA self-sorting observed in a previously reported organic cage CC3-RS. This result prompted us to computationally explore the thermodynamic preference for the formation of pseudo-C3h cages incorporating both CHDA enantiomers against the corresponding enantiopure Tet 3 Di 6 cages (see ESI Section S2 for the computational details). Indeed, the DFT formation energy for ## Mixed aldehyde systems To explore self-sorting in the system, we combined aldehydes from the L and B families in single-pot reactions with R,R-CHDA under cage-forming conditions. We expected the Tet 3 Di 6 topology to be favoured in all cases based on our previous work. This assumption reduced the space of the possible structures to a number that could be systematically explored by computational methods. As imine formation is reversible under the reaction conditions used here, the observed product distributions are expected to relate to the thermodynamic minima. Therefore, formation energies can be predictive of the range of products seen. If all the possible cages have similar formation energies, we predict that multiple cage products will be formed or that the self-sorted products will be selected by solvation and entropic effects. Conversely, if one or more cages are much lower in energy than the other possibilities, we predict that those structures will dominate the product distribution. For each reaction studied, we manually constructed the Tet 3 Di 6 cages in all possible stoichiometries of the linear and bent aldehydes. We then applied a workflow consisting of high-temperature molecular dynamics simulations with OPLS3e force field, followed by further geometry optimisations at the PBE-GD3/TZVP-MOLOPT-GTH level of theory to find the expected gas-phase conformations of the resulting cages (see ESI Section 3 for more details). Single point energies were calculated for the modelled structures at the M06-2X/6-311G(3df, 3dp) level of theory and the resulting formation energies are summarised in Table 1 (for B1) and Table 2 (for B2). These calculations are performed on isolated molecules in the gas phase, which does not consider solvent effects, and hence large energetic differences are needed to predict solution-phase structures with confidence. In parallel, we attempted to address the problem synthetically by targeting the cage stoichiometry of [L + 2B] (+ 6 R,R-CHDA omitted for clarity). A deuterated variant of aldehyde L3 was available from a previous study and used in reactions with B1 to allow discrimination of the resultant cages by UPLC-MS. No such deuterated analogue was available for the mixture of L4 and B2, and the reaction was characterised primarily by 1 H NMR chemical shifts and UPLC retention times. Tables 1 and 2 summarise which structures were experimentally observed. Reactions of B1 with aldehydes L1-4 and R,R-CHDA all resulted in cage compounds (corresponding to entries marked with asterisks in Table 1 (see ESI Section S3. 4 for screening details and raw spectra). For the shortest linear aldehyde L1, the major cage product was a pseudo-D3 low-symmetry [L1 + 2B1] cage, which was readily isolated via recrystallisation. The structure of this compound was elucidated by single crystal X-ray diffraction (see Figure 3) and was the lowest energy structure predicted for that system. When the elongated L2 was used instead, the two major products observed by UPLC-MS were a low-symmetry [2L2 + B1] cage and the previously described homoleptic [3L2], which again were the two lowest-energy predicted structures. For the even longer aldehyde L3-d, a complex mixture was observed by UPLC-MS (see ESI Figures S16-S21 and Tables S3-S4). The major product was identified as the [3L3-d] cage, in agreement with the computational models. Lower-mass peaks corresponding to ## [2L3-d + B1] and [L3-d + 2B1 ] could also be detected, both structures being of comparable DFT formation energies. The relative proportions of the products could not be determined due to insufficient chromatographic separations. For the longest aldehyde trialled, L4, the major product was the symmetrical tubular [3L4] cage, which was of significantly lower DFT energy than any competing structure in this system. In all cases, other species could be detected by mass spectrometry as trace products, including the chiral cage [3B1], but could not be isolated. The distribution of the products is affected by the length of the linear aldehydes L1-4 in a seemingly unpredictable way, but the observed structures agree with predicted trends in the DFT formation energies. The length of L1 appears to be suitable to relieve strain in a cage containing two B1 moieties; L2 is of suitable length to relieve strain in a cage containing one B1 moiety, but L3 and L4 seem to be too long and do not form stable mixed Tet 3 Di 6 cages with B1. Reactions involving aldehyde B2 and aldehydes L1-4 also all resulted in the formation of cage compounds (see Table 2 and ESI Section S3.4). For aldehydes L2 and L3, the outcomes of the reactions were scrambled and all possible Tet 3 Di 6 cages were observed, which were all of comparable DFT formation energies. For L1 and L4, we observed narcissistic self-sorting. In the case of L4, signals corresponding to [3B2] and [3L4] can be seen by 1 S5). However, as all products in this system have the same mass, it was not possible to unambiguously characterise the self-sorting behaviour with mass spectrometry. While formation energies of [3L1] and [L1 + 2B2] are comparable, and the formation energy of [3L4] is higher than that of the sociallysorted cages [nL4 + mB2], we propose that the clean narcissistic self-sorting in those cases is a result of antagonistic coupling between the homoleptic and the heteroleptic cages in these libraries. Formation of the stable [3B2] cage removes free B2 from solution, thus favouring the formation of Table 1 Side-and top-views of the DFT-optimised structures (PBE-GD3/TZVP-MOLOPT-GTH) of the possible Tet 3 Di 6 outcomes for the reactions using mixtures of B1 and L1-4 under cage forming conditions. Underneath are the single point formation energies (M06-2X/6-311G(3df, 3dp)) in kJ mol -1 . Entries marked with asterisks are the experimentally observed outcomes. Building blocks are coloured according to Figure 1, nitrogen atoms are dark blue, hydrogen atoms are omitted. Table 2 Side-and top-views of the DFT-optimised structures (PBE-GD3/TZVP-MOLOPT-GTH) of the possible Tet 3 Di 6 outcomes for the reactions using mixtures of B2 and L1-4 under cage forming conditions. Underneath are the single point formation energies (M06-2X/6-311G(3df, 3dp)) in kJ mol -1 . Entries marked with asterisks are the experimentally observed outcomes. Building blocks are coloured according to Figure 1, nitrogen atoms are dark blue, hydrogen atoms are omitted. ## Heteroatom-containing [L1 + 2B1X] systems To investigate whether social self-sorting would also be observed with analogues of B1 we synthesised thiophene (B1S) and pyridyl (B1N) derivatives that have structurally related geometries (see ESI Section S3.2 for the synthetic details). Due to the incorporation of heteroatoms and differently sized rings in their cores, these aldehydes were expected to produce cages with different pore geometries and electronic properties. experiments, which demonstrated the hydrodynamic radii are similar for all three structures (see ESI Section S4). Further evidence was obtained from IMS-MS experiments, which indicated that the drift times for the three cages are similar (see ESI Section S5), supporting the conclusion that the subtle differences in the linker structures have little effect on the overall molecular size in these systems. We performed analysis of the shapes and electronic structures of the internal cage cavities to probe the effect of using different aldehydes. Cage geometries were optimised as described previously. These structures were used to calculate the total electron density and the electrostatic potential at the M06-2X/6-311+G(d,p) level of theory. We found the 0.0004 a.u. density isosurface and selected a subsurface approximating the internal cavity of each cage (see ESI Section S2.5 for the algorithm and the implementation).Figure 4 shows the mapping of the electrostatic potential onto the cavity surface. The potential around the main window between the two non-linear aldehydes is most affected by the neighbouring heteroatoms, while the entire cavity surface becomes narrower and elongated in the case of the more expanded thiophene linker in [L1 + 2B1S]. The heteroatoms themselves have little effect on the shape of the void, but do affect its electronic properties, providing a subtle yet important distinction that may have consequences for guest binding and selectivity. was predicted to be more stable than the corresponding homoleptic cages [3L1] and [3B1], and was indeed preferentially formed. Two heteroatom-containing analogues of [L1 + 2B1] were formed using this strategy, demonstrating the generality of the social self-sorting approach to synthesis of organic cages of low-symmetry. The slight change in the aldehyde geometry and the incorporation of heteroatoms did not affect the overall size of the cage molecules, while allowing for tuning of the shape and electronic properties of the internal cavity. We hope that these results will aid the design of more anisotropic organic cages for challenging separations and the selective encapsulation of biologically relevant low-symmetry guests.

chemsum

{"title": "Inducing social self-sorting in organic cages to tune the shape of the internal cavity", "journal": "ChemRxiv"}

a_flexible_and_scalable_scheme_for_mixing_computed_formation_energies_from_different_levels_of_theor

## Abstract: Phase stability predictions are central to computational materials discovery efforts and have been made possible by large databases of computed properties from high-throughput density functional theory (DFT) calculations. Such databases now contain millions of calculations at the generalized gradient approximation (GGA) level of theory, representing an enormous investment of computational resources. Although it is now feasible to carry out large numbers of calculations using more accurate methods, such as meta-GGA functionals, recomputing the entirety of a database with a higher-fidelity method is impractical and would not effectively leverage the value embodied in existing calculations. Instead, we propose in this work a general procedure by which higher-fidelity, low-coverage calculations (e.g., meta-GGA calculations for selected chemical systems) can be combined with lower-fidelity, high-coverage calculations (e.g., an existing database of GGA calculations) in a robust and scalable manner to yield improved phase stability predictions. We demonstrate our scheme using legacy GGA(+U ) calculations and new r 2 SCAN meta-GGA calculations from the Materials Project and illustrate its application to solid and aqueous phase stability. We discuss practical considerations for constructing mixed phase diagrams and present guidelines for prioritizing high-fidelity calculations for maximum benefit. ## I. INTRODUCTION The advent of large databases of computed material properties, such as the Materials Project , AFLOW , the Open Quantum Materials Database (OQMD) , and the Joint Automated Repository for Various Integrated Simulations (JARVIS) , has paved the way for a new era of data-driven materials science . These databases now contain computed properties derived from millions of individual calculations, the vast majority of which employ density functional theory (DFT) due to its efficient compromise between computational cost and accuracy. For example, the Materials Project contains computed formation energies for more than 140,000 materials calculated using the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) functional, with a Hubbard U value and empirical energy corrections applied to some chemical systems. This data is widely used in machine learning and computational materials screening efforts in which the thermodynamic (meta)stablility of a material is often among the first selection criteria . Despite its versatility and historical success, PBE has well-documented systematic errors related to electron * kapersson@lbl.gov; https://materialsproject.org/ self-interaction that are particularly notable in diatomic gases and transition metal compounds with localized electronic states . PBE also fails to capture medium-and long-range dispersion interactions , which are important for describing the properties of weakly-bound systems. Even when adjusted using empirical correction schemes, the mean absolute error (MAE) in formation energies predicted by this level of theory are still on the order of 50-200 meV/atom [10,14,15, , although the error in energy differences among polymorphs is typically lower (e.g. 25 meV/atom) . Today, more than a decade after most materials databases were established , theoretical advances and growth in computing power have made it feasible to compute large numbers of formation energies at higher levels of theory , which could substantially increase their accuracy. For example, we recently showed that the restored-regularized Strongly Constrained and Appropriately Normed (r 2 SCAN) meta-GGA functional reduced the error in predicted formation energies of strongly-bound and weakly-bound materials by 50% and 15%, respectively, compared to the PBEsol GGA functional, while simultaneously exhibiting reliable convergence . The original SCAN functional on which it is based has also been shown to predict volumes, lattice constants, and ground-state structures of many materials more accurately than PBE [22, . Carrying out enough higher-fidelity calculations to comprehensively cover technologically-relevant chemical spaces, as is required for the construction of compositional phase diagrams and the discovery of new structure-property relationships, could clearly benefit the many materials discovery efforts that depend on such data. However, replacing all of the existing lowerfidelity (GGA) calculations in large materials databases with higher-fidelity (e.g., meta-GGA) calculations would consume an enormous amount of energy and computing time, since SCAN and r 2 SCAN, for example, have 4-5× the computational cost of PBE . Even if resources were unlimited, there is likely to be little benefit in recomputing materials that are highly unstable (i.e., far from the convex energy hull), since predicting (meta)stability is of primary importance. Furthermore, meta-GGA calculations will not improve formation energy predictions to an equal extent for all materials. For example, SCAN has been shown to be slightly less accurate than PBE in predicting the formation energies of weakly-bound materials (e.g., intermetallics ), and r 2 SCAN improves the predictions for these materials to a much lesser extent than for strongly-bound materials . Therefore, instead of recomputing materials en masse, higher-fidelity calculations should be targeted at the materials for which they are likely to improve the accuracy of the phase diagrams the most (e.g., strongly-bound materials, materials close to the energy convex hull). Adopting this strategy will economize future use of resources and preserve the massive investment embodied in existing lower-fidelity calculations. Notably, however, such an approach will require phase stability predictions to be based on a mixture of formation energies computed at different levels of theory. The most straightforward way to systematically improve upon GGA phase diagrams in a high-throughput manner, which we refer to here as "naive mixing", is simply to build each phase diagram using formation energies from lower level calculations, and then replace them with higher level calculations whenever they are available. However as we will show, constructing mixed phase diagrams this way can result in severe distortions to the shape of the convex energy hull and dramatically worsen phase stability predictions. As an alternative to naive mixing, we propose in this work a scheme to construct phase diagrams that mix calculations from different density functionals comprising a lower-fidelity, higher coverage and a higher-fidelity, lower coverage set of calculations (here, PBE(+U ) and r 2 SCAN) with minimal risk of distortion. By defining the reference state at each point in compositional space as the ground state PBE(+U ) structure, we build a framework in which energies from any two functionals can be mixed in a robust and scalable manner that preserves the shape of the convex hull. After presenting our mixing scheme, we assess how a transition from PBE(+U ) to r 2 SCAN affects predicted polymorph stability and energy above hull by analyzing a set of approximately 33,900 r 2 SCAN calculations and discuss strategies for prioritizing r 2 SCAN calculations such that the mixed phase diagram most closely approximates the full r 2 SCAN phase diagram. We conclude by using our mixing scheme to analyze solid and aqueous phase stability in two example systems. The mixed phase diagrams presented in this work, along with the 33,000+ new r 2 SCAN calculations, are made publicly available in the Materials Project database to increase the accuracy of future computational material science efforts. ## II. THEORETICAL FRAMEWORK FOR MIXING FORMATION ENERGIES FROM DIFFERENT FUNCTIONALS A. Mixing rules The computed energy of formation for a material, ∆H f , is defined with respect to the elements by where E 0K,DFT is the total electronic energy computed from DFT at 0 K, subscript 'el' represents each of the constituent elements in the material, and n are stoichiometric coefficients. We note that E can include empirical corrections and that this formulation assumes that differences in finite-temperature enthalpy between materials are negligible. Electronic energies E are not intrinsically meaningful, and their energy scales differ substantially among functionals for the same material. However, the differences in electronic energy among materials and elements, and hence the value of ∆H f , define a consistent, physically-meaningful quantity that can be compared among different levels of theory. In this manuscript, we consider mixing GGA(+U ) and r 2 SCAN calculations; although the scheme we present here can be used to mix energies from any two functionals. Note that we use "GGA(+U )" to refer to the mixture of empirically-corrected PBE and PBE+U calculations that currently populate the Materials Project Database. Specifically, the Materials Project uses PBE (i.e., GGA) for all materials except those containing Co, Cr, Fe, Mn, Mo, Ni, V and W, which are calculated with a Hubbard U value . These GGA and GGA+U calculations are combined using the mixing scheme of Jain et al. to yield a consistent set of formation energies. In this work, we use this combination of adjusted GGA and GGA+U calculations as our high-coverage, low-fidelity set of calculations, while "r 2 SCAN" denotes unadjusted meta-GGA energies that comprise the lowcoverage, higher-fidelity calculations. Additional details regarding the computational methods for each calculation type are provided in Section V. As noted in the Introduction, the most straightforward approach to constructing mixed r 2 SCAN / GGA(+U ) phase diagrams is "naive mixing", where we simply replace GGA(+U ) formation energies with r 2 SCAN formation energies whenever r 2 SCAN calculations are available, while using GGA(+U ) formation energies everywhere else. There are two drawbacks with naive mixing. First, in chemical systems where r 2 SCAN predicts significantly smaller or larger formation energies than GGA(+U ) for most compounds, inserting a single r 2 SCAN formation energy onto a GGA(+U ) phase diagram can either cause that single phase to move off the hull ), when neither would occur on a full r 2 SCAN phase diagram. Second, in many cases r 2 SCAN stabilizes a different ground-state structure for the elements than GGA(+U ) (see Appendix B). Naive mixing of formation energies in chemical systems containing one of these elements is not rigorously consistent, because the formation energies are being referenced to different structures. To circumvent these issues, we build our mixing scheme by considering all formation energies to be the sum of a reference energy and a relative energy. We define the reference energy, E ref , for each functional as the formation energy of the GGA(+U) ground state structure at each point in compositional space, where the formation energy is calculated with respect to elemental end-points relaxed with that functional. The energy of any material in either functional may then be expressed as a difference, ∆E ref , relative to the corresponding reference energy. Note that ∆E ref is calculated directly from the difference in polymorph energies, and hence does not depend on the energies of the elemental endpoints. Our mixing scheme is similar in spirit to the previous GGA/GGA+U mixing scheme as well as the combined computational-experimental Pourbaix diagrams of the Materials Project ; however extends these approaches to be applicable to any two functionals without relying on pre-fitted energy correction parameters. Using this framework, we propose two "mixing rules" that define our scheme for constructing mixed r 2 SCAN / GGA(+U ) phase diagrams. These mixing rules are summarized schematically in Figure 1 and briefly elaborated below. In sections that follow, we will illustrate each rule with an example. 1. Beginning with a GGA(+U ) hull, replace GGA(+U ) energies with r 2 SCAN energies by adding their ∆E ref to the corresponding GGA(+U ) reference energy. 2. Construct the convex energy hull using ∆H r 2 SCAN f only when there are r 2 SCAN calculations corresponding to every reference structure (every GGA(+U ) stable structure). In this case, add any missing GGA(+U ) materials by adding their ∆E ref to the corresponding r 2 SCAN reference energy. Rule #1 provides a means to introduce r 2 SCAN energies onto GGA(+U ) phase diagrams when r 2 SCAN calculations are only available for one or a few compositions. In Figure 1, polymorph A represents the reference structure (GGA(+U ) ground state). Since the reference structure is, by definition, on the GGA(+U ) hull, the r 2 SCAN relaxed structure corresponding to this reference structure (as determined by the pymatgen StructureMatcher algorithm) is assigned ∆H GGA(+U) f . Polymorphs B and C, calculated in r 2 SCAN, are assigned energies that maintain their energy difference with respect to the reference structure, ∆E ref . For example, if polymorph C is 10 meV/atom higher in energy than polymorph A in r 2 SCAN, it would be assigned an energy ∆H GGA(+U) f + 10 meV/atom. It is also possible for this method to place a polymorph below the GGA(+U ) hull. Polymorph B is unstable with respect to the reference structure in GGA(+U ) but is lower in energy than the reference structure in r 2 . If it were 10 meV/atom lower than the reference structure in r 2 SCAN, it would be assigned an energy ∆H GGA(+U) f − 10 meV/atom, slightly changing the shape of the hull in the mixed phase diagram compared to GGA. Finally, polymorphs that do not have a r 2 SCAN energy (such as polymorph D) maintain their energy with respect to the GGA(+U ) hull. When r 2 SCAN calculations become available for every reference state, then, by Rule #2, the convex hull is computed directly with r 2 SCAN formation energies (∆H r 2 SCAN f ). Now, any unstable GGA(+U ) phases that have not been calculated in r 2 SCAN can be added to the diagram by adding their ∆E ref to the H r 2 SCAN f of the corresponding reference structure. In other words, we invert Rule #1 so that r 2 SCAN structures become the reference structures. Note that a central assumption of our mixing scheme is that r 2 SCAN energies are always preferable to GGA(+U ) energies. This assumption is well-justified by the generally superior accuracy of SCAN and r 2 SCAN formation energies reported in many studies [18,22, . In general, application of our mixing scheme should be restricted to pairs of functionals where one has an a priori reason to prefer one energy over another. In addition, we note that in principle it is possible to use our framework to mix energies from more than two functionals, provided that reference energies are available within each functional and that a clear hierarchy can be established among them. ## B. Mixed diagrams for relative polymorph stability (Rule #1) We illustrate the motivation behind Rules #1 and 2 using the Sn-Br phase diagram, which is shown in Figure 2. In general, when constructing phase diagrams we seek to determine 1) the shape of the convex energy hull (i.e., stable compositions and their formation energies), and 2) the stable polymorph at each composition. Figure 2a and 2b compare the Sn-Br phase diagram with the formation energy of all phases calculated in GGA(+U ) and r 2 SCAN, respectively, and show that the accuracy of both aspects is improved by r 2 SCAN. GGA(+U ) incorrectly predicts the ground-state polymorph of SnBr 2 as rocksalt (spacegroup P 3m1) and overpredicts the magnitude of ∆H f as -1.136 eV/atom, whereas the experimental value is estimated at -0.84 -0.92 eV/atom (indicated by the shaded band in Figure 2). By contrast, r 2 SCAN correctly predicts the SnBr 2 ground state as P nma and makes a substantially more accurate prediction of its formation energy (-0.833 eV/atom). As we have discussed, it is not always feasible to recompute an entire chemical system using r 2 SCAN (as we have done to construct Figure 2b). When improving predictions of polymorph stability is a primary research objective, it makes sense to prioritize r 2 SCAN calculations for all known polymorphs at the composition of interest. However, if we were to apply this strategy to SnBr 2 and replace all GGA(+U ) formation energies of SnBr 2 polymorphs with r 2 SCAN energies using naive mixing (Figure 2c), SnBr 2 would no longer be predicted as stable. This occurs because the entire hull is shallower (smaller magnitude of ∆H f ) in r 2 SCAN than in GGA(+U ), and hence using an r 2 SCAN formation energy for SnBr 2 causes it to move off the hull. Instead, we must apply Rule #1 to make the r 2 SCAN energies compatible with the GGA(+U ) hull. We do so by positioning r 2 SCAN formation energies relative to the GGA(+U ) ground state polymorph (P 3m1), as shown in Figure 2d. Because we maintain the energy differences relative to this reference energy, the correct polymorph is now stabilized. Compared to naive mixing of formation energies, applying Rule #1 preserves the overall shape of the GGA(+U ) convex hull while enabling improvement in phase stability predictions using as few as two r 2 SCAN calculations (one for the polymorph of interest and one for the reference structure). However, because r 2 SCAN stabilizes the P nma polymorph instead of the P 3m1 polymorph stabilized by GGA(+U ), ∆H f is lowered (and made less accurate) by 37 meV/atom, which is the difference in energy between the P 3m1 and P nma polymorphs in r 2 SCAN. Hence, although use of Rule #1 for study of a single composition may yield more accurate relative polymorph energies, it carries the risk of making the magnitude of the formation energy slightly less accurate compared to a full r 2 SCAN phase diagram. ## C. Mixed diagrams for formation energy (Rule #2) When identifying stable compositions or predicting accurate formation energies is the primary research objective, it makes sense to prioritize recomputing all GGA(+U ) ground states in r 2 SCAN, as shown in and all other materials are in GGA; d) the same set of energies as c), but employing our mixing scheme (Rule #2); e) r 2 SCAN for all reference states (i.e., GGA(+U ) ground states) and GGA(+U ) for all other materials; f) r 2 SCAN for all materials within 20 meV/atom of the GGA(+U ) convex hull. The numerical value in parentheses indicates the energy above hull of the experimental ground state P nma polymorph of SnBr2. The shaded blue regions represent the estimated range of experimental formation energies for SnBr2 . Tabulated r 2 SCAN and GGA(+U ) energies for all materials are provided in Appendix E. convex hull to be constructed using ∆H r 2 SCAN f . Unstable GGA(+U ) polymorphs are then positioned relative to the corresponding reference structures. Several unstable polymorphs of Sn that were mixed in this manner are visible in Figure 2e. In this chemical system, recomputing the hull when only the GGA(+U ) ground states have been calculated in r 2 SCAN will still not recover the exact r 2 SCAN for-mation energy for SnBr 2 , because GGA(+U ) stabilizes the incorrect ground state, and Rule #2 treats this incorrect ground state as the reference energy. Hence, the formation energy of SnBr 2 predicted by the mixed phase diagram in Figure 2e is too high by 37 meV/atom (the difference in energy between the P 3m1 and P nma polymorphs in r 2 SCAN). Because there is no way to know a priori whether r 2 SCAN will stabilize a different ground state than GGA(+U ), a more robust strategy is to compute all polymorphs within some tolerance of the GGA(+U ) hull. Computing both ground states and slightly metastable polymorphs with r 2 SCAN makes it more likely that the shape of the convex energy hull in the mixed phase diagram will be identical to that in a full r 2 SCAN diagram. We apply this strategy in Figure 2f, in which we mix r 2 SCAN energies for all materials within 20 meV/atom of the GGA(+U ) hull, and use GGA(+U ) energies for all other materials. The value of 20 meV/atom is motivated by analysis presented later (see Section III B) indicating that materials with higher ∆E ## GGA(+U) hull are rarely stabilized by r 2 SCAN. With this strategy, the shape of the resulting energy hull (Figure 2f) exactly matches that of the pure r 2 SCAN hull (Figure 2b). Comparing Figures 2a, 2c, and 2e illustrates the importance of Rule #2 when mixing r 2 SCAN and GGA(+U ) calculations. Formation energies cannot be naively mixed without carrying a substantial risk of over-or understabilizing certain compositions. The hull must remain in GGA(+U ) until there are r 2 SCAN calculation corresponding to every reference state. Even in that case, it is preferable to include slightly unstable polymorphs in order to achieve better accuracy in cases where r 2 SCAN stabilizes different polymorphs. ## III. PRACTICAL CONSIDERATIONS A. Mixed diagrams for ternary and higher systems Ternary or higher-dimensional chemical spaces present special challenges for mixing energies between functionals, because strict application of mixing rules #1 and #2 can introduce inconsistencies between the full phase diagrams and those of constituent subsystems. For example, consider a case in which all of the GGA(+U ) ground states in chemical system A-B are computed in r 2 SCAN. According to Rule #2, the binary A-B phase diagram would be constructed using ∆H r 2 SCAN f . This may result in different formation energies and/or predicted stable phases than the GGA(+U ) phase diagram, as illustrated previously for the Sn-Br system. Now suppose that we wish to construct a ternary phase diagram for the A-B-C system, in which there are multiple ternary ground states that have not been computed in r 2 SCAN. Since the A-B-C system does not satisfy the requirements for Rule #2, we would construct this ternary phase diagram using ∆H ## GGA(+U) f . This could result in the ternary A-B-C diagram predicting different formation energies and/or stable phases in the A-B subsystem than the binary A-B diagram. Such an inconsistency may be problematic depending on the use case. Note that if all reference energies in the full A-B-C system have been recomputed in r 2 SCAN, then strict application of the mixing rules will not result in any inconsistencies. However, due to the much larger number of ternary and higher materials (compared to binaries), it becomes progressively more difficult to recompute all the reference energies needed to apply Rule #2 as the size of the chemical system increases. In cases where consistency between lower-and higherdimension phase diagrams is essential, one may apply the mixing rules individually to each chemical subsystem, in order of increasing dimensionality. To continue the example above, Rule #1 and Rule #2 would be applied individually to each of the A-B, B-C, and A-C chemical systems. The ternary phase diagram would then be constructed by combining these pre-adjusted binary formation energies with GGA(+U ) formation energies. An example of such a diagram is presented in Figure C.2. Applying the mixing scheme to binary subsystems before treating the ternary system amounts to a modified form of naive mixing because it involves directly combining formation energies obtained from GGA(+U ) (for ternaries) with those calculated with r 2 SCAN for binaries, without considering whether r 2 SCAN energies are available for all ternary ground states. As such, mixed phase diagrams for high-dimensional chemical systems that are constructed in this manner should be used sparingly and interpreted with care. However, due to the inherently larger number of phases involved in higher dimensional systems, we expect this modified form of naive mixing to be less likely to cause severe distortions of the hull compared to binary systems. To test this hypothesis, we compared ternary phase stability predictions from approximately 6,000 ternary phase diagrams computed in GGA(+U ) to mixed versions constructed using ∆H r 2 SCAN f values for all binary subsystems and ∆H for all ternary materials (see Appendix C). We evaluated how frequently these "edged" diagrams either 1) destabilized a known experimental ternary phase (i.e., a phase reported in the Inorganic Crystal Structure Database ) that was stable in pure GGA(+U ) or 2) stabilized a known experimental ternary phase that was unstable in pure GGA(+U ). For the majority of chemical systems (83%), experimental ternary materials predicted stable by the pure GGA(+U ) diagram remained so in the mixed phase diagram, while in another 14% of cases exactly one material was destabilized. Similarly, for 94% of chemical systems, experimental materials predicted unstable by the pure GGA(+U ) diagram remain so in the mixed diagram, while for 6% of chemical systems exactly one unstable experimental material was stabilized (see Figure C.1). Thus, although employing modified naive mixing (i.e., "edged" phase diagrams) to achieve consistency between lower-and higher-dimensional phase diagrams carries a modest risk of destabilizing known experimental phases for some chemical systems, there are many other cases in which the mixed diagrams stabilize experimental phases that pure GGA(+U ) does not. Altogether, these results suggest that modified naive mixing is unlikely to severely distort phase stability predictions. ## B. Definition of materials "close to the hull" In Section II we observed that it is preferable to recompute not just GGA(+U ) ground states, but also materials close to the convex energy hull in order to ensure that the mixed energy hull has the correct shape (compare Figures 2e and f). This begs the question of how to define "close to the hull". More specifically, we can rephrase the question as "how likely is r 2 SCAN to stabilize a material that is X meV/atom above the hull in GGA(+U )?" For example, a material that is 500 meV/atom above hull in GGA(+U ) will almost certainly not become stable in r 2 SCAN, but a material that is unstable by 3 meV/atom could (as was the case with P nma SnBr 2 in Section II). Determining an appropriate threshold is necessary to properly target r 2 SCAN calculations. To inform this question, in Appendix A we evaluate the extent to which r 2 SCAN changes the energy above hull of unstable polymorphs. Examining approximately 7300 unstable materials with a GGA+(U ) energy above hull of 50 meV/atom or less, we find that in 95% of cases, the energy above hull either increases or decreases by no more than 19 meV/atom (see Figure A.1). This means that materials more than 19 meV/atom above the GGA(+U ) hull would only be stabilized by r 2 SCAN in rare cases. Hence, we adopt a threshold of 20 meV/atom as our definition of "close to the hull" for purposes of prioritizing calculations. By way of comparison, we note that among 16 systems identified by Yang et al. in which SCAN stabilized the correct ground state and GGA(+U ) did not, the energies above hull of the experimental ground states in GGA(+U ) ranged from 2 to 50 meV/atom. Another study showed that SCAN mispredicted the ground states of TiO 2 and FeS 2 , with misprediction on the order of 50 meV/atom as well . Thus, although based on analysis of a large set of materials, our selection of 20 meV/atom as a "safe" threshold is not guaranteed to capture the r 2 SCAN ground state polymorph in every case. A higher threshold could certainly be chosen if greater confidence in capturing the correct ground states is required. ## C. Failures of structure matching In Section II A we established the need to obtain r 2 SCAN energies of GGA(+U ) ground states, which serve as reference energies for constructing mixed phase diagrams. To obtain the most accurate r 2 SCAN energies, we generally perform r 2 SCAN structure optimizations rather than single-point calculations and then use the pymatgen StructureMatcher algorithm to determine whether the r 2 SCAN-relaxed structure is the same (within tolerances) as the GGA(+U ) starting structure. In the vast majority of cases, the r 2 SCAN-relaxed structure and the GGA(+U ) starting structure match, allowing us to use the r 2 SCAN energy as a reference energy. However in selected cases (some 1% of all materials we have computed thus far), r 2 SCAN will optimize to a structure that is no longer considered equivalent to the starting structure. This is especially common for the crystal structures of diatomic molecules (e.g., H 2 , Cl 2 , O 2 ) in which the different treatment of short-and medium-range interactions by r 2 SCAN compared to PBE is particularly significant. We address this issue in two ways. In some cases, manual inspection of the structures allows us to establish that they represent the same material, and hence that the r 2 SCAN energy can be used as a reference energy. However, manual inspection is not feasible for highthroughput work. Instead, we perform single-point calculations for any materials in which the r 2 SCAN-relaxed structure no longer matches the input structure. The r 2 SCAN single-point calculation is guaranteed to match the corresponding GGA(+U )-optimized structure and provides an r 2 SCAN energy that can serve as a reference energy. Meanwhile, an r 2 SCAN optimization of the same structure (which may no longer be the same according to the StructureMatcher) is guaranteed to have a similar or lower energy than the single point and will be added to the hull at the correct position by application of Rule #1. Performing r 2 SCAN single points also provides a means of obtaining reference energies for large structures that would be impractical to optimize in r 2 SCAN within reasonable computational limits (see Section III D). ## D. Prioritizing r 2 SCAN calculations for maximum benefit We conclude our discussion of practical considerations by considering the best strategy for prioritizing r 2 SCAN calculations, given that computational resources are limited and that its cost is still approximately 5× that of PBE . We can define several levels of "calculation coverage" (meaning, subsets of materials that have all been recomputed with r 2 SCAN, Figure 3) based on the mixing rules we have established. In order to apply Mixing Rule #1, at least two r 2 SCAN optimizations at a single composition are needed: one for the GGA(+U ) ground state and one for another polymorph. To apply Rule #2, we require r 2 SCAN energies for every GGA(+U ) ground state or (ideally) every GGA(+U ) material within 20 meV/atom of the hull. These energies are preferably obtained from structure optimizations, although as discussed above, single-point calculations can be used, with the risk of a slightly less accurate hull shape. The pinnacle of calculation coverage (which may have less value that its computational cost, as noted in the Introduction) is full recomputation of all materials using r 2 SCAN. With a goal of achieving second-or third-level cov- erage, we can identify several strategies for prioritizing which materials to calculate in order to maximize the benefits of the mixing scheme for formation energy prediction. To do so, we classify materials close to the hull as 1) strongly-or weakly-bound and 2) small or large. Previous studies established that SCAN and r 2 SCAN predict substantially more accurate formation energies than PBE or PBEsol for "strongly-bound" materials, i.e., materials whose GGA(+U )-predicted formation energy is lower than -1 eV/atom. The improvement in accuracy for "weakly-bound" materials is more modest. As such, creating mixed phase diagrams for stronglybound systems is likely to improve overall accuracy the most, and hence we assign higher priority to stronglybound materials. With respect to size, experience indicates that optimizations of large structures (e.g., larger than approximately 40 sites) with r 2 SCAN will often exceed typical maximum wall time limits at supercomputing centers (e.g. 48 hr). This is not to say optimization is impossible; rather, in a high-throughput computing context it does not usually make sense to invest an excessive amount of computing nodes or wall time into a single material. As such, we choose to perform single-point calculations for large materials in order to obtain a reference energy (albeit a less accurate one) so that Rule #2 can be applied. Fully-optimized structures can be obtained as computational resources allow, and added into the mixed phase diagrams according to Rule #1. ## IV. EXAMPLES A. Application to a metastable ternary nitride system As a practical example of our complete mixing scheme, we use it to investigate compound metastability in the ternary Zn-Sb-N system. Nitrides remain relatively unexplored compared to other chemical spaces, even though they exhibit the largest range of thermodynamicallyaccessible metastable states among inorganic materials , which is thought to be a consequence of the large cohesive energy of metal-nitrogen bonds that kinetically traps metastable structures . Compared to stable nitrides, metastable nitrides are more likely to contain metal cations in high oxidation states, which imparts unique semiconducting properties that make these materials interesting for electronic and photovoltaic applications, among others . Metastable nitrides are relatively rare in nature and difficult to synthesize experimentally due to the high stability of molecular N 2 . However, the use of reactive nitrogen precursors such as ammonia, azide compounds, or plasma-cracked atomic N allow nitrogen chemical potentials of up to +1 eV/N above the hull to be reached in laboratory synthesis . Recent experimental studies have reported synthesis of several metastable nitrides (Cu 3 N, Sn 3 N 4 , and Tialloyed Sn 3 N 4 ). Ternary Wurtzite-based nitrides, such as MgSnN 2 , ZnSnN 2 , ZnGeN 2 , and ZnSiN 2 , have received specific attention recently as potential alternatives to III-V semiconductors. . Computational screening studies recently predicted three new metastable ternary phases (ZnSb 2 N 4 , Zn 2 SbN 3 , and Zn 3 SbN 3 ) and FIG. 4. Zn-Sb-N phase diagrams illustrating different mixing strategies for GGA(+U ) and r 2 SCAN calculations. a) GGA(+U ) only; b) r 2 SCAN only; c) strict application of Rule #1 and #2 to calculations comprising r 2 SCAN energies for all elements and binary phases with GGA(+U ) energies for all ternary phases d) same set of calculations as c), but using modified naive mixing in which binary hulls are constructed from r 2 SCAN formation energies. Phases labeled "metastable" are phases that can be stabilized by a +1 eV/N increase in the nitrogen chemical potential, which is achievable in laboratory synthesis . Tabulated r 2 SCAN and GGA(+U ) energies for all materials are provided in Appendix E. a new metastable binary phase (SbN) in the Zn-Sb-N chemical space. Zn 2 SbN 3 , the first Sb-based nitride semiconductor ever reported, was experimentally realized and exhibited promising electronic properties for photovoltaic and water splitting applications. SbN was predicted to be relatively close to the metastability limit (requiring +0.8 eV/N to stabilize) and is the subject of ongoing investigations as another potential Sb-based nitride semiconductor. Given the diverse bonding characteristics of nitrogen compounds, computational predictions of metastability can be particularly sensitive to the choice of functional and energy correction scheme (e.g., GGA vs. GGA+U vs. r 2 SCAN). For example, Sun et al. showed that for many binary nitride systems, PBE overstabilizes the nitrogen-rich region of the convex energy hull, while GGA+U overstabilized the nitrogen-poor region. SCAN was found to predict formation enthlapies with good accuracy across both portions of the hull . To expand on this previous work and inform future high-throughput screening studies, we evaluate how the use of a mixed r 2 SCAN / GGA(+U ) ternary phase diagram would affect these predictions. Figures 4a and 4b show the phase diagrams computed entirely using GGA(+U ) and r 2 SCAN calculations, respectively. Both the pure GGA(+U ) and pure r 2 SCAN phase diagrams predict that Zn 2 SbN 3 and SbN are metastable, consistent with the previous studies. However, the energy above hull of both metastable compositions of interest is higher in r 2 SCAN. r 2 SCAN predicts 30 meV/atom above hull for Zn 2 SbN 3 (vs. 20 meV/atom in GGA), reflecting the overstabilization of the N-rich region of the phase diagram in GGA(+U ) noted by Sun et al. . r 2 SCAN predicts 260 meV/atom above hull for SbN (vs. 172 meV/atom in GGA), but in the r 2 SCAN diagram this material falls within the metastable synthesizability limit, consistent with experimental reports . The other notable differences between the GGA(+U ) and r 2 SCAN phase diagrams are that Zn 3 SbN 3 , which has not been synthesized to the best of our knowledge and is not in the ICSD, is predicted stable by GGA(+U ) but unstable by r 2 SCAN, while ZnSb 2 N 4 is unstable in GGA(+U ) but metastable in r 2 SCAN. The stable / metastable / unstable classification for all other compositions is the same in both diagrams. Moving to mixed phase diagrams, we now consider a situation in which only the elements and binary compositions have been computed with r 2 SCAN, while all ternary phases remain in GGA(+U ). We compare two methods of constructing this mixed ternary phase diagram in Fig- ures 4c and 4d. In Figure 4c, we apply Mixing Rule #1 and #2 strictly (i.e., considering the entire phase diagram at once). In this scenario, because we do not have a r 2 SCAN calculation for the GGA(+U ) reference structure Zn 3 SbN 3 , the hull is still calculated using GGA(+U) energies (with the exception of polymorphs stablized by r 2 SCAN, as discussed later). r 2 SCAN polymorphs for each element or binary composition are placed on this GGA(+U ) hull by anchoring to the respective reference states according to Rule #1. Inspection of Figure 4c shows broad similarity to the pure r 2 SCAN diagram (Figure 4b), with a few notable differences. Zn 3 SbN 3 is predicated unstable in the pure r 2 SCAN diagram yet metastable in the mixed diagram. For SbN the reverse is true: this material is predicted metastable in the pure r 2 SCAN diagram but unstable in this mixed diagram. The convex energy hull in Figure 4c is constructed with GGA(+U ) energies, so it is identical to that of Figure 4a with one significant exception. As noted in Section II B, Rule #1 can cause the energy of the convex hull to decrease in cases where r 2 SCAN stabilizes a different polymorph than GGA(+U ). In this case, r 2 SCAN stabilizes a different structure for N 2 (which is a crystalline solid at 0 K) that is 1.8 meV/atom lower in energy than the reference energy (see Figure B.1), causing it to be placed below the GGA(+U ) hull and thereby "lowering" the hull energy of the N-rich region in the mixed phase diagram. This causes Zn 3 SbN 3 , which is predicted stable in the pure GGA(+U ) phase diagram, to move off the hull by 0.4 meV/atom and become metastable. The energy above hull for SbN increases by 1 meV/atom, and hence it retains its classification as unstable consistent with the GGA(+U ) phase diagram. (Figure 4d) presents an alternative phase diagram con-structed by fully applying Rules #1 and #2 to the binary edges and then adding GGA(+U ) energies for ternary phases by modified naive mixing, as discussed in Section III A. Here, the edges of the diagram are identical to those predicted by a pure r 2 SCAN diagram because we have full coverage of all GGA(+U ) ground states and hence Rule #2 applies. In the interior of the diagram, three of the four ternary compositions retain the same stable/unstable/metastable classification they have in the pure r 2 SCAN diagram, while Zn 3 SbN 3 is predicted to be stable in the GGA(+U ) diagram (whereas it is predicted unstable in the pure r 2 SCAN diagram). As noted in Section III A, the modified form of naive mixing employed to construct Figure 4d is thermodynamically less consistent than strict application of Rule #1 and #2 (Figure 4c), and should only be invoked when consistency between binary and higher dimension phase diagrams is essential. In this example, where the metastability of phases is of primary interest, it would be advisable to apply the mixing rules strictly to ensure that the entire convex hull is constructed in a consistent manner. Indeed, among the two mixed diagrams, Figure 4c shows the most consistency with the pure r 2 SCAN diagram. ## B. Application to aqueous phase stability Finally, we demonstrate how the mixing scheme presented here can be used to inform aqueous phase stability predictions. The computational Pourbaix diagram formalism of Persson et al. generates aqueous stability (pH-pE) diagrams by referencing experimental free energies of dissolved ions to DFT-predicted formation energies derived from solid phase diagrams. As such, the mixing scheme presented here can be applied to the creation of Pourbaix diagrams in addition to solid phase diagrams. SCAN-derived Pourbaix diagrams, for example, were shown to be systematically more accurate for transition metal oxides . However, the large number of stable phases needed to build computational Pourbaix diagrams may preclude calculating entire chemical spaces in SCAN or r 2 SCAN, motivating the usefulness of our mixing scheme in this context. We illustrate the mixing scheme on the Se-O system, for which the PBE-derived Pourbaix diagram is known to be inaccurate with respect to experiment . Specifically, it predicts a stable SeO 2 phase that is not observed experimentally (Figure 5a). Creating a computational Pourbaix diagram of this system requires a solid phase diagram of the Se-O-H chemical system, which contains 85 individual materials, according to the Materials Project database. Five of these materials contain more than 40 sites, and hence could be particularly challenging to recompute in r 2 SCAN (see Section III D). Use of the mixing scheme allows us to construct the hull in r 2 SCAN by performing calculations only for the ground states (9 materials) while retaining information from GGA(+U ) about the metastability of other phases. The resulting Pourbaix diagram built from mixed r 2 SCAN and GGA(+U ) energies (Figure 5b) correctly predicts that the oxide phase SeO 2 is unstable (unlike the pure GGA(+U ) diagram), in agreement with Pourbaix diagrams presented by Wang et al. that were prepared from both experimental data and pure SCAN calculations. Hence for this system, our mixing scheme has made it possible to leverage a relatively small number of calculations and achieve similar predictive accuracy as full recalculation of all materials with SCAN. ## C. Summary In summary, we have developed a mixing scheme to enable construction of phase diagrams that combine formation energies from any two DFT functionals. Such a capability is important to high-throughput materials screening efforts because it allows a relatively low-coverage, high-fidelity set of calculations (here, r 2 SCAN) to be used in concert with existing high-coverage, lower-fidelity calculations (here, GGA(+U )) to improve the accuracy of phase stability predictions. Our scheme allows mixing r 2 SCAN and GGA(+U ) calculations when as few as two r 2 SCAN calculations (one corresponding to a GGA(+U ) ground state) are available, and scales smoothly to cases where entire binary, ternary, or higher-dimensional chemical systems are calculated in r 2 SCAN. We identified specific guidelines that can be used to target limited computational resources towards materials where r 2 SCAN calculations are likely to improve accuracy the most, and illustrated how the mixing scheme can be applied to solid and aqueous phase stability predictions. ## V. METHODS We employed the Vienna ab initio Simulation Package (VASP) v.6.1.1 in conjunction with v.54 of the projector-augmented wave (PAW) PBE pseudopotentials for all r 2 SCAN calculations in this work. We employed a two-step high-throughput workflow described elsewhere , which comprises a structure optimization with PBEsol to generate a initial guess of the charge density, followed by a subsequent structure optimization with r 2 SCAN . These calculations were force-converged, with a plane-wave energy cutoff of 680 eV and a bandgap-dependent k -point density , which were developed to achieve a formation energy converged to within approximately 1 meV/atom. We do not apply any energy corrections to the resulting r 2 SCAN energies. For context regarding corrections, in Appendix D we fit energy corrections to diatomic gases and show that the corrections that would be applied to r 2 SCAN are substantially smaller than those that have been widely used for GGA(+U ) . PBE calculations were retrieved from the Materials Project REST API . Calculations for transition metal oxides and fluorides contained a Hubbard U value and incorporated the GGA/GGA+U mixing scheme of Jain et al. , in addition to empirical corrections applied to some chemical systems . We refer to these calculations as GGA(+U ) throughout this work. ## VI. DATA AVAILABILITY All data referenced herein are publicly available in the Materials Project database . At the time of this publication the database contains r 2 SCAN calcula-tions for approximately 33,000 materials, corresponding to 77% of all elements, binary, and ternary materials within 20 meV/atom of the GGA(+U ) convex energy hull. Our computational workflow has been implemented into the pymatgen and atomate packages as of version 2020.1.28 and 0.9.5, respectively, for readers wishing to utilize it in their own work. The mixing scheme described herein is available in the Material-sProjectDFTMixingScheme class in pymatgen as of release 2022. 1.20. was greater than -19 meV/atom, and for 99% this value is -36 meV/atom. This means that compounds that are more than 19 meV/atom above the GGA(+U ) hull are unlikely to appear on the r 2 SCAN hull, while compounds more than 36 meV/atom above hull will only be stabilized by r 2 SCAN in especially rare circumstances. Here, we assess the frequency and magnitude with which r 2 SCAN changes the relative stability of elemental structures compared to GGA, which we define as the change in the energy above hull for a given structure when going from GGA to r 2 SCAN: ∆∆E GGA→r 2 SCAN hull ≡ ∆E r 2 SCAN hull − ∆E GGA hull . If r 2 SCAN yields the same elemental groundstate structure as GGA, then ∆∆E GGA→r 2 SCAN hull = 0 by definition. If GGA predicts elemental structure A to be 2 meV/atom higher in energy compared to stable elemental structure B, but r 2 SCAN predicts elemental structure A to be stable and 2 meV/atom lower in energy than elemental structure B, then ∆∆E GGA→r 2 SCAN hull = −4 meV/atom. Note that because all structures considered in this section are elemental, no U value is applied in this analysis. Figure B.1 highlights the elements that have different ground-state structures between GGA and r 2 SCAN as well as the magnitude of this energy difference, where applicable. Based on the results in Figure B.1, even for the elements where r 2 SCAN stabilizes a different ground-state structure, the energy difference is often relatively small; the average value of ∆∆E GGA→r 2 SCAN hull is -30 meV/atom (or -12 meV/atom if including the 0 meV/atom change in energy above hull for the elements that do not have a different ground state). One pronounced exception is elemental Ce, for which the r 2 SCAN ground state structure is unstable in GGA by 2 meV/atom, but the GGA ground state structure is unstable in r 2 SCAN by 223 meV/atom. It should also be noted that, for some elements (i.e. Eu, Pu, Xe), r 2 SCAN does not a yield a DFT-optimized structure that matches the GGA ground-state structure (as determined by the pymatgen StructureMatcher algorithm and subsequent manual inspection). This is likely due in part to an improved treatment of van der Waals interactions with r 2 SCAN.

chemsum

{"title": "A flexible and scalable scheme for mixing computed formation energies from different levels of theory", "journal": "ChemRxiv"}

proximity_effect_in_crystalline_framework_materials:_stacking-induced_functionality_in_mofs_and_cofs

## Abstract: Metal-organic frameworks (MOFs) and covalent organic frameworks (COFs) consist of molecular building blocks being stitched together by strong bonds. They are well known for their porosity, large surface area, and related properties. The electronic properties of most MOFs and COFs are the superposition of those of their constituting building blocks. If crystalline, however, solid-state phenomena can be observed, such as electrical conductivity, substantial dispersion of electronic bands, broadened absorption bands, formation of excimer states, mobile charge carriers, and indirect band gaps. These effects emerge often by the proximity effect caused by the van-der-Waals interactions between stacked aromatic building blocks. This Progress Report shows how functionality is imposed by this proximity effect, that is, by stacking aromatic molecules in such a way that extraordinary electronic and optoelectronic properties emerge in MOFs and COFs. After discussing the proximity effect in graphene-related materials, its importance for layered COFs and MOFs is shown. For MOFs with well-defined structure, the stacks of aromatic building blocks can be controlled via varying MOF topology, lattice constant, and by attaching steric control units. Finally, an overview of theoretical methods to predict and analyze these effects is given, before the layer-by-layer growth technique for well-ordered surface-mounted MOFs is summarized.Received: ((will be filled in by the editorial staff))Revised: ((will be filled in by the editorial staff)) ## Introduction Molecular framework materials, including metal-organic frameworks (MOFs), coordination polymers, and covalent organic frameworks (COFs), provide an intriguing bridge between chemistry and solid-state physics. They are composed of molecular units that may carry the whole range of functional groups known to chemistry. These molecular building blocks are stitched together by strong bonds. This setup provides much higher chemical, thermal, and mechanical stability as compared to molecular crystals and, thus, allows the formation of large pores which can reach up to 10 nm in diameter, as well as extremely large internal surface areas (see, e.g., and references therein). These unique properties have motivated intense MOF and COF research during the past years: they allow high gas uptake capacities and, thus, application in methane and hydrogen storage. Coupling these properties, intrinsic to porous materials, with molecular functionalities integrated into nodes and linkers can yield to multifunctionality: selective uptake, CO2 capture , hydrogen isotope separation, as well as switching permeance and selectivity via optical or electrical switching. Catalysis also profits from the presence of large pores and channels, which allow transport of reactants and products, and, if coupled with functional groups, turns the voids in these materials into selective nanoreactors. Even though the IUPAC recommendations for nomenclature explicitly specifies that these materials are not necessarily crystalline, high crystal order is observed for many of them. Crystallinity allows high-quality structural analysis by experimental methods, e.g., via x-ray diffraction. As a consequence, theoretical work can be carried out in a straightforward fashion and can then be compared directly to experimental results, thus, allowing for a direct validation of computational approaches. The well-defined structure also makes a proper physico-chemical characterization of these materials possible and provides, thus, the basis for the high level of understanding of their structure and electronic structure. At the same time, crystal order is the reason for solid-state effects that are caused by the translational symmetry, such as magnetic order, indirect band gaps, and ballistic charge transport. As in other crystalline solids, defects play a critical role for certain properties, including catalytic activity, electronic and optical properties, thus, a proper characterization of defect types and their density is crucial. This point is particularly important, as the properties discussed below are a direct consequence of the crystallinity of the framework materials. Intrinsic magnetic order suffers from the large distances between the spin centers in a MOF and the resulting weak couplings in the order of few meV. Thus, they are only observed at cryogenic temperatures. Ballistic transport is typically hindered either by the large effective masses of the typically dispersionless bands or by the chemical composition of the frameworks, where non-carbon centers (for example oxygen, boron or nitrogen) can effectively block electron conjugation. Only recently, ballistic transport, facilitated by electron conjugation, has been demonstrated in layers of two-dimensional (2D) MOFs and COFs. There is, however, a more subtle way to implement strong band dispersion in crystalline molecular framework materials: Controlled stacking of aromatic molecules, incorporated into the materials as linkers or as pillars with suitable intermolecular distance, is subject to the proximity effect, more precisely, the molecules interact via π -stacking, which causes strong alterations of the electronic structure of the framework materials. The proximity effect is the basis of the recent breakthrough in van-der-Waals physics of 2D materials (see, e.g., [31-40]). While this type of stacking is rather obvious for the class of layered COFs with atomically thin layers and with aromatic connectors or linkers, it can also be achieved in MOFs with suitable crystal structure. A further option of MOFs and, to some extent, also COFs is to grow superstructures by applying layer-by-layer procedures (MOF-on-MOF) and to fabricate structurally well-defined organic/organic heterointerfaces. In this Progress Report, we elaborate the impact of the proximity effect in MOFs and COFs with suitable crystal structures. We will first discuss the importance of the proximity effect for the case of graphene, which serves as role model for the stacked aromatic moieties in MOFs and COFs. In order to reach a wide audience with background in chemistry, we will introduce key concepts of solid-state physics, when appropriate. We will then approach the closest relatives to graphene, stacked layered COFs, before we explore pathways to tailor these interactions by means of crystal structure and functional side groups in both layered and threedimensional MOFs. The conclusions include our perspective of further exploiting the proximity effect in crystalline porous framework materials and, thus, exploiting the fusion of most recent advances in the condensed matter physics and materials chemistry. In the experimental section, we give an overview of theoretical approaches, rationally design and to computationally explore these materials, and summarize the layer-by-layer (lbl) growth technique to synthesize surfacemounted MOFs with controlled stacks of aromatic linker molecules. ## Stacking of Graphene Layers In order to understand the potential impact of stacking of aromatic moieties in molecular framework materials, it is useful to discuss the proximity effect in graphene. Graphene is by far the most widely studied 2D material up to date and, at the same time, it can be considered as the prototype of a 2D polymer. The extraordinary properties of graphene are distinct from those of its parental 3D system, graphite: massless Dirac fermions, exceptionally high electrical-conductivity, long-distance spin-transport, and half-integer quantum Hall effect. The Dirac points (see Figure 1a) in the Brillouin zone of graphene, situated at each K point, cause these very interesting features. Stacking two or more single graphene layers on top of one another may, however, change most of the electronic properties of graphene. In the solid state, Bernal graphite is the most common form, where the graphene layers are stacked in the ABAB fashion, such that centers of 6-fold rings of one layer are situated on top of carbon atoms in the second layer, forming the so-called staggered structure. Bernal graphite requires four carbon atoms in the unit cell, which results in four bands at the K point of the Brillouin zone close to the Fermi level, however, they are now parabolic rather than linear, and, thus, give rise to massive electrons in a zero-gap semiconductor (see Figure 1b, see also Refs. ). On the other hand, an almost linear (more accurately, a parabolic dispersion with very large curvature) and very small effective mass can be seen at the H point of the Brillouin zone, with the Fermi level slightly below the top of the valence band and a band gap of about 5 meV, thus, a few orders of magnitudes larger than in graphene. ), and (c) eclipsed graphite (AAAA stacking, hypothetical). Single layer of graphene, which exhibits Dirac points (linear dispersion relation) at K points and a band gap of about 1 eV, changes its electronic properties when stacked differently into 3D structures: ABAB graphite has quadratic and nearly linear dispersion relations at the K and H points, respectively, and band gap at H of 5 meV; AAAA graphite has two Dirac points (at K and H points) above and below Fermi level (blue dashed horizontal lines). Simulations at the DFT/PBE-D3(BJ) level of theory with TZP basis set and spin-orbit coupling as implemented in AMS/BAND software. Pictures of structures made with VESTA. In a hypothetical structure of eclipsed graphite (AAAA stacking of layers, Figure 1c), an intriguing electronic structure emerges. Two atoms per unit cell and one layer are necessary to represent the smallest possible 3D structure. There are two nearly linear bands at the K point close to the Fermi level, similar to graphene. However, the two Dirac points without band gaps at K and H points are above and below the Fermi level, respectively. Thus, the  and  * bands are not symmetric with respect to the Fermi level. This demonstrates that the proximity effect depends both on the distance between the layers and on the mutual shift of the crystal planes. Even more strikingly, strong effects are observed if two adjacent graphene layers are twisted with respect to each other. The beautiful moiré structures of the lattice are reflected also in the electronic structure (in presence of magnetic field resulting in the Hofstadter butterfly). Most notably, in 2018, Cao et al. showed that at a magic twist angle of 1.05°, correlation is strongly enhanced and a Mott insulating state is formed, which results in a superconducting state. This effect was shown initially in 2011 by Bistritzer and MacDonald, who reported that the Dirac-point velocity vanishes at the twist angle of 1.05° and that this effect is accompanied by a very flat moiré band, which results in a sharp peak in the Dirac-point density of states (see Figure 2). The difference between flat, conjugated bands and dispersionless bands should be noted here: the former result from the symmetry of the lattice (e.g., kagome) with extended  conjugation, while the latter indicate localized, non-interacting states without electron conjugation. This short summary of electronic properties of graphene, bilayer graphene, and graphite illustrates the importance of the proximity effect: even though the graphene layers are subject to only weak van-der-Waals forces, which leave the atomistic structure essentially unaffected, the interlayer interaction has significant impact on the electronic structure, which is so strong that it can even cause the electronic phase transition from a zero-gap semiconductor to a superconductor. Similar stackings are present in layered COFs and in MOFs with suitable geometries. The impact of the proximity effect in these materials will be discussed in the next chapters. ## Stacking of 2D Covalent Organic Frameworks In 2005, Côte and co-workers introduced a new family of layered and porous materials which are referred to as covalent organic frameworks (COFs). In these systems, organic linker molecules are connected via covalent bonds with connectors, formed, e.g., by oxygen and boron atoms (see Figure 3b for building blocks of COF-5). Regular frameworks, formed by stitching together these linkers and connectors, can have honeycomb topology within each layer, as in the case of COF-5 (see Figure 3d), but are not restricted to this topology. For example, using porphyrins as tetratopic organic linkers leads to a square topology. Note that even though these materials are referred to as "2D COFs", in the literature, they are in fact rather layered materials, where single 2D sheets form three dimensional structures with porous channels normal to the basal planes. Nowadays, many different COF systems have been synthesized and investigated, and several reviews or perspective articles are available in the literature. Here, we would like to focus on the aspect of layer stacking and its effect on the electronic properties of selected COFs. In 2011, we have calculated the stacking order of a few layered COFs from first-principles. We focused especially on the structures of COF-1, -5, -6, and -8. Back in 2011, the experimentally reported layer stacking of these systems was either AAAA (eclipsed) (cf. Figure 3e top) or ABAB (staggered). We showed that these systems are considerably more energetically stable (see Figure 3a) if their stacking arrangement is either serrated (AA'AA', that is, the 6-fold rings in the linkers and connectors are stacked as in Bernal graphene, cf. Figure 3g top for COF-5) or inclined (AA'A''A''', cf. Figure 3f top). In these arrangements, the layers are shifted with respect to one another by about one C-C bond length, ~1.4 , compared with the eclipsed stacking. Note that many other stackings, possibly random ones, are possible. Our calculated PXRD patters fit very well to the experimental ones (see Figure 3c). We also showed that the proposed stackings of layers affect the pore geometries. Pores in COFs are the functional regions and we calculated that the inclined and serrated arrangements account for an increase in the surface area by 6%, estimated for the interaction with He, or 3% for the interaction with N2, compared with eclipsed stacking. Moreover, the polarity of pores increases for the two stackings, because both oxygen and boron atoms are exposed to the pore surface. After our report on the stacking structure of COFs, the proposed stacking arrangements, especially the serrated one, were commonly adapted in the community. In 2017, Fan et al. have demonstrated for the first time inclined stacking of layers in 2D COFs by introduction of steric substituents between the layers. PXRD patterns of all the stacking arrangements. (d-g) (Top) structures and (bottom) electronic band structures (lowest two panels show zoom-ins of conduction band maximum (CBM) and valence band minimum (VBM)) of (a) monolayer (1L), (b) eclipsed (AAAA), (c) inclined (AA'A''A'''), and (d) serrated (AA'AA') stackings of COF-5. Single layer COF-5 has fairly flat bands close to the Fermi level (blue dashed horizontal lines). Prominent dispersion of bands in the direction perpendicular to the layers can be observed upon stacking single layer to multilayer COF-5, the strongest in case of AAAA (eclipsed) stacking. Calculations performed on the optimized structures from Ref. at the DFT/PBE level of theory with TZP basis set as implemented in AMS/BAND software. Figure adapted from Ref. . Pictures of structures made with VESTA. We have simulated the electronic band structure of single layer of COF-5, as an exemplary 2D COF structure, and compared it to these of bulk COF-5 with different stackings: eclipsed, inclined, and serrated (see Figure 3 bottom panels for full band structure and zoom-ins of conduction and valence bands. The band structure of the single layer of COF-5 exhibits almost flat bands when compared to the 3D stackings. However, close examination of the valence and conduction bands in the vicinity of the Fermi level of the COF-5 single layer reveals a typical signature of the so-called (Archimedean) kagome lattice (see Figure 3d CBM zoom). In the valence band (see Figure 3d VBM zoom), these are two bands of linear dispersion, crossing at the K point, similar to graphene. These bands are also sandwiched between two flat bands, forming so-called ruby bands. Thus, the top of valence band is very flat throughout the Brillouin zone. The conduction bands are a bit more disperse, with minimum at the  point. The single layer of COF-5 is, thus, a direct-gap semiconductor with a band gap of about 2.7 eV. Our results are in good agreement with previously reported theoretical work of Liu et al., who obtained a band gap of 2.5 eV at a similar level of theory. Stacking of COF-5 layers into a 3D structure with different arrangements (Figure 3d-f) changes electronic properties of the materials. While the eclipsed stacking keeps almost unchanged inplane signatures of the kagome lattice, interlayer interactions induce strong band dispersion in the direction perpendicular to the layers. The band dispersion between  and A points in the Brillouin zone reaches almost 1 eV. As discussed above, this stacking is, however, energetically unfavorable. The two most stable high-symmetry structures of COF-5, inclined and serrated, strongly alter the electronic structure of the monolayer. In both cases, the in-plane kagome characteristic is no longer present along the -K-M- path. For serrated case, some kagome signature is visible for A-L-H-A path, however, these bands are not forming the band edges, as in the monolayer or eclipsed form. Out-of-plane dispersion is also present, but smaller than in the eclipsed stacking. For the serrated case, it is less than 0.5 eV. Band gap values and characters also change compared with the monolayer case: for the eclipsed structure, we calculated an indirect band gap of about 1.3 eV between  and A points, for the inclined structure, we obtained an indirect band gap of about 2.2 eV between X and  points, while serrated stacking gives also an indirect band gap of about 2.2 eV between M and  points. Band dispersion is of course desirable for nanoelectronic applications, because curved bands at the band edges result in light charge carriers (holes and electrons). While COF-5 single layer offers light electrons and heavy holes, which is interesting for production of short transport channel devices, the stacking introduces also light holes, which could be interesting for standard transistor applications. This survey on COF-5 shows that, just as in graphene, the proximity effect has a strong influence on the electronic structure of molecular framework materials. After the reports of COF-1 and COF-5, many routes to form 2D COFs have been developed. While a comprehensive summary of the literature would not fit into the scope of this article, we highlight here the most important steps that are crucial for the development of van-der-Waals science in 2D COFs. First, a large variety of coupling reactions has been developed in the field of 2D polymers (the term 2D polymers is commonly used for single layers of 2D COFs). We show a selection in Figure 4. Second, the large diversity of aromatic structures offers, besides honeycomb and square, also many other structural forms, including hexagonal and kagome. For detailed reviews, see Refs. [74-77]. Moreover, even the symmetry of the pores can be controlled by proper choice of linkers. This illustrates that there is no doubt that the potential structures offered by 2D COFs significantly exceed those of dense layered materials that are available in nature or have been synthesized. Third, in order to enable ballistic electron transport also inside the individual COF layers, it is important to remove the electron scattering centers, which typically are the atoms at the coupling sites. If they are perturbing the conjugation in the COF, electrons cannot flow and the electronic system is not much different to that of a molecular crystal. As shown in the groups of Jiang and Feng, fully conjugated 2D COFs are possible, e.g., if aromatic building blocks are coupled by using the Knoevenagel reaction. These works demonstrate that 2D COFs with fully conjugated electronic π-system can be realized in experiments, which means that it is possible to construct high-mobility single layer 2D COF semiconductors. It remains to be explored which consequences the proximity effect has, if these 2D COFs are stacked upon one another, both in homogeneous and heterogeneous stacks. Up to date, a huge variety of 2D COF structures has been synthesized and studied, e.g., in the direction of photoactivity and photosensitivity. [61,78, However, electronic structures are seldom discussed and, even then, mostly only single layers are considered. [60, Thomas et al. investigated electronic structures of 2D -conjugated polymers with three-and fourarm connections, which correspond to kagome and Lieb (Archimedean) lattices, and explained occurrence of flat and disperse bands. However, no stacking was taken into account. On the other hand, Er et al. reported an excellent carrier mobility and photoconductivity along the vertical direction of a DA-COF (D -donor, A -acceptor). The authors found that these properties depend on the number of layers and the stacking of layers. The conduction was achieved by electron hopping between layers, in which the donor and the acceptor groups were aligned with respect to each other. This arrangement would correspond to the AAAA stacking discussed above. In other works on COF systems available in the literature, some of the COFs were reported to have high charge carrier mobilities, as in the case of COF-366 and COF-66, with covalently linked porphyrin units offering extended planar -electron system. ## Layered, atomically-thin metal-organic frameworks MOFs are prototypes of the reticular chemistry concept, in which well-defined building blocks are stitched together to form porous 3D frameworks. In the simplest case, two building blocks are used, linkers (typically di-or higher-topic organic molecules) and connectors (typically metal ions or metal/oxo clusters). Up to date, a huge variety of MOFs has been synthesized and characterized, with the most frequent applications in the fields of gas adsorption and separation, sensing, drug-release, proton conductance in fuel cells, catalysis, water splitting, and many others. In contrast to COFs, where the 2D versions are more prominent than their 3D counterparts, 2D MOFs with atomically thin layers coupled by van-der Waals-interactions areto daterather exotic and have been reported in few cases only. [24, However, the instances of such layered system reported so far promise an interesting bridge between chemistry and physics, offering materials with interesting electronic and magnetic properties. Similar to COFs, and even sometimes better, MOFs also offer extended 2D π-conjugation. For example, a recent paper by Yang et al. reports on a semiconducting MOF magnet. K3Fe2[PcFe-O8] (Pc -phthalocyanine) exhibits spontaneous magnetization and full in-plane πd conjugation, which results in room temperature carrier mobility of about 15 cm 2 V -1 s -1 . The long-range magnetism arises from the magnetic coupling between iron centers via the πelectron system. This MOF exhibits strong out-of-plane band dispersion, however, different to many COFs and MOFs, it also shows in-plane dispersion. The K3Fe2[PcFe-O8] MOF consists of layers being only one atom thick (see Figure 5a). Such atomically-thin planar MOFs, with extended in-plane π-conjugation, are currently found to be the most conductive frameworks known up to date. For example, in the case of Ni3HITP2 (HITP -2,3,6,7,10,11hexaiminotriphenylene; see Figure 5b), the highly π-stacked and extended conjugation of the network results in high electrical conductivity. This semiconducting MOF was initially investigated by Dinca and co-workers and later by Zeng and co-workers. It has an interlayer distance of about 3.3 and a conductivity of 40 S cm -1 at room temperature. In another MOF-like coordination polymer based on benzenehexadelenolate, ([Cu3(C6Se6)]n), with the π-d conjugation extended in-plane, the layers are again stacked in the AA'AA' fashion and have Cu metal centers in a square planar configuration (see Figure 5c). The interlayer distance is about 3.6 , very close to that in graphite. The material exhibits metallic character, similar to many other atomically-thin MOFs, and very high room-temperature conductivity of 110 S cm -1 . It should also be noted that, using sophisticated forms of MOF post-synthesis modification, MOFs can be transformed into COFs by crosslinking the primary linkers with suitable secondary linkers and then removing the metal ions form the lattice. This approach can also be used to fabricate effectively 2D COF structures. ## Proximity effect in Metal-Organic Frameworks: the SURMOF showcase Several groups have introduced layer-by-layer methods to deposit MOF thin films on substrate, first, in 2007, Wöll and Fischer reported on the layer-by-layer route to MOF synthesis, and later H. Kitagawa and coworkers. Such surface-mounted MOFs are referred to as SURMOFs. They exhibit high crystallinity and can be investigated using virtually all surface science techniques, see also Section on SURMOFs synthesis in the Experimental Section. SURMOF-2, being an isoreticular series based on MOF-2, is one of the simplest MOF architectures suited for layer-by-layer growth. They are derived from MOF-2, a bulk framework material based on paddle-wheel units with four dicarboxylate groups and typically Cu 2+ or Zn 2+ -dimers connected to ditopic organic linkers of different length, the shortest one being 1,4-benzene dicarboxylate. The length of the linkers determines the lattice constant and, thus, the pore size of the resulting SURMOFs, where up to 4 nm in diagonal have been reached up thus far. Layers of such SURMOFs form square lattices and, theoretically, could be stacked together in three different arrangements. The most symmetric P4 variant is the eclipsed stacking with linkers and connectors in one layer directly on top of linkers and connectors in another layer (see Figure 6a). This stacking is found in all SURMOF-2 derivatives discussed in this chapter and leads to a less stable system due to unfavorable Coulomb and van-der-Waals interactions. Computational investigations showed that two other stackings are energetically more favorable. These are slipped and inclined stackings (see Figure 6b and c), with P2 or C2 symmetries, respectively, which emerge in bulk synthesis. However, in the SURMOF approach, the metastable P4 symmetry with eclipsed stacking is enforced by the anchoring of the first MOF layers to the nucleating surface. at the DFT/PBE0-D3 level of theory with all electron basis set (double zeta with polarization on light elements and triple zeta with polarization for Cu atoms) as implemented in Crystal14. The Cu atoms were antiferromagnetically coupled in the present simulations (bands corresponding to  and  electrons are identical). Slipped and inclined stacking of SURMOF-2 layers induces out-of-plane dispersion in band structure, due to increased van-der-Waals and preferable Coulomb interactions. Pictures of structures made with VESTA. We have calculated the corresponding electronic band structures for all three stackings of Cu-SURMOF-2 (see Figure 6). While the in-plane bands are nearly dispersionless, the stacking induces dispersion in the out-of-plane directions ( to Z). The conduction band is formed by dispersionless single state in all cases, which corresponds to the Cu d-orbitals. On the other hand, the valence band is dominated by C p-orbitals of the organic linkers. The P2 or C2 systems exhibit heavy electrons throughout the Brillouin zone, similar as in P4. However, light holes emerge due to stacking-induced band dispersion in the direction perpendicular to the layers. The first report on indirect band gap formation in MOFs was published in 2015. Thin films of epitaxial MOFs have been studied, and photoinduced charge-carrier generation was observed. The investigated Zn-paddle wheel SURMOF-2 derivative utilized Pd-porphyrinoid linkers (Pd-PP-SURMOF, see Figure 7a). As parent SURMOF-2, this structure is also a square lattice inplane and layers are stacked in the AAAA fashion. Such a system results in fairly dispersionless bands of the electronic structure (see Figure 7b), however, zoom-in to the conduction and valence bands reveals dispersion of a few meV. This value corresponds to a mobility of about 0.003 cm 2 V -1 s -1 , which at that time was larger than for any other MOF. This MOF exhibits an indirect band gap, which should result in suppressed electron-hole recombination and improved photovoltaic properties in such organic-semiconductor based devices. . Bands are fairly flat, however, small out-of-plane dispersion occurs in the direction perpendicular to the layers. The dispersion is in the limit of a couple of meV. Pictures of structures made with VESTA. The photovoltaic efficiency of the reported PP-SURMOF amounts to only 0.2%, and is thus far too low for realizing a competitive device. Another issue is the absorbance of the PP itself: the strongly absorbing Soret band is in the ultraviolet, while the Q bands, which are located in the visible spectrum, are only weakly absorbing. PP functionalization can strongly enhance the absorbance of the Q bands. Among the large number of possibilities, three particularly interesting ones have been identified by combining rational design with computational screening. (1) Distorting the planarity alters the selection rules and, thus, enhances the intensity of the Q bands. This can be achieved by bromination of the PP core (twisted octabromo porphyrin). ( 2) The π-conjugation of the PP can be extended by adding a phenyl-acetylene (PA) group. ( 3) The π-system can be affected by the presence of electron-withdrawing fluorinated phenyl substituents. All three strategies show an effect on the isolated PP linker molecules, in particular, the Q band intensity increases. Incorporated to a PP-SURMOF, these three linkers, however, show very different absorbance. This can, again, be attributed to the proximity effect. For the brominated linker, the geometrically distorted building blocks fail to arrange themselves into well-ordered stacks. As a result, loosely packed linker stacks with distances of ~6.1 , too far to cause a significant proximity effect, are obtained. Hence, the incorporation of the brominated PP (1) into the MOF lattice does not affect the absorber properties significantly. For the fluorinated species (3), also the formation of a well-ordered lattice is reported. However, Coulomb repulsion keeps the linkers at the widest possible distance (~6.3 ) and the resulting absorption spectrum is very similar to that of the individual molecules. If large and aromatic linkers are used (as in case of PA, (2)), the attractive London dispersion interaction fosters linker rotation to the extent that the PP molecules form well-ordered stacks with intermolecular distance of ~3.3 , very close to that in graphite. Consequently, the band structure shows strong dispersion, which results in a red shift of the Soret band and enhances and broadens the Q bands. These results are summarized in Figure 8 and reported in Ref. . . Strong band dispersion observed for the PA-Zn-SURMOF in the stacking direction, due to enhanced London dispersion interactions between the PP linkers. Pictures of structures made with VESTA. To what degree is the spatial extension of the aromatic system of the individual porphyrinic linkers relevant to the properties, which are mainly caused by the proximity effect in the linker stack? To answer this question, we analyzed the band dispersion as function of the rotation angle between the PP moiety and the PA substituent (see Figure 9). This is achieved by rotating parts of the linker (PA and the benzene rings connected to the paddle wheel). Indeed, such a rotation results in a strong manipulation of the electronic bands, from almost flat conduction band in a hypothetical structure with all likers at 0 with respect to the PP to very strong dispersion in both band edges for rotations of both parts by 110. We believe that such rotations can be achieved by proper selection of functional groups (steric control units, SCUs, see below) in the PP linkers. corresponds to the case of i = 0° and i = 110°. Calculations at the same level of theory as in Ref. . Pictures of structures made with VESTA. In the previous paragraphs, we have shown that well-arranged stacks of aromatic molecules can provide additional functionality to MOFs. The precise control of stacking of MOF layers can be tuned by introducing so-called steric control units (SCU; see Figure 10). As demonstrated in recent work, when restricting the size of these units to values smaller than the MOF pore size, the packing of aromatic moieties within the MOF linkers can be varied without changing overall lattice parameters of the MOF material. In particular, by first creating libraries in-silico using appropriate computational workflows (as described in ), the structures of interest can be identified by computational screening of large libraries, before linkers equipped with the "best" SCU are synthesized and used to assemble the corresponding MOFs or SURMOFs. Figure 10. Employing crystal engineering to tune excitonic coupling in chromophoric SURMOFs using steric control units (SCUs). (a) Without SCUs, the naphthalene-diimide (NDI) linkers yield a non-luminescent assembly (H-aggregate). A theoretical analysis revealed that increasing the angle Q controlling the stacking of the linkers (b) modifies the coupling such that luminescence is recovered. (c) The workflow used to screen the library of possible linkers is indicated in the lower left. (d) TDM-TDM coupling changes sign (TDMtransition dipole moment). Figure adapted from Ref. . Additionally, the individual MOF layers provide strong anisotropic transport properties for excitons and other optical excitations. In some systems, exceptionally large exciton diffusion lengths have been determined experimentally. In that work, it was proposed that the observed experimental value of around 100 nm is actually only limited by the domain size of the investigated SURMOFs. Based on theoretical considerations, the intrinsic exciton diffusion length should reach the micrometer-regime, which would make this thin layer MOF one of the best exciton transport materials. ## Conclusion This Progress Report shows that, in addition to molecular functional groups, undercoordinated metal sites, porosity, and large surface areas, a further possibility of property control can be incorporated into crystalline framework materials, such as COFs and MOFs: If aromatic molecules are placed in well-controlled stacks, the proximity effect gives raise to strong electronic effects. If the intermolecular distance between the basal planes of the aromatic molecules is in the range of the interlayer distance of graphene (~3-3.5 ), disperse electronic bands emerge, resulting in a ballistic charge carrier transport with appreciable mobilities. Thus, while the electronic properties of most framework materials are merely the superposition of the electronic properties of the constituting molecular building blocks, a suitable stacking of aromatic building blocks can turn them into semiconducting materials with particular electronic and optoelectronic properties. Without steric control and sufficient flexibility, van-der-Waals interactions result in selfassembly of stacks with strong proximity effect. Mutual shift and twist between the basal planes of the aromatic molecules and intermolecular distance have a strong impact on the resulting electronic structure change, due to the proximity effect. It is possible to control the stacking by strong interlayer interactions, functional groups or steric control units. These control mechanisms are still beyond the state -of-the-art and subject of the authors' ongoing research efforts. As recent results of the physics community on twisted bilayer graphene and two-dimensional van-der-Waals heterostructures demonstrate, many new effects are possible by exploiting the proximity effect. These so-called quantum materials include Mott insulators, superconducting states, Majorana fermions, and topological states. The rich structural and chemical variety of MOFs and COFs opens the door to exploit these effects in materials that emerge from chemistry. It will pave the way to a wealth of opportunities, e.g., exploiting topological states to drive chemical reactions as catalysts or to control molecular separation, to form high-density wellordered molecular wires, and to form interfaces between quantum materials and liquids or nanoparticles. ## Computational Methods This section gives an overview of methods that are useful for computationally tackling layered COFs and MOFs, from constructing the atomistic structures to obtain high-level electronic structure data. It includes the methods that have been used to obtain the results discussed above. The purpose is to allow newcomers to quickly enter the exciting field of framework materials. Structure builder. Many MOFs and COFs have non-trivial geometry and are composed of rather large, sometimes bulky, molecules. Several methods have been developed to stitch these building blocks together to form a framework, including AuToGraFS , zeo++ , and MOFplus. Coupling framework generation to a force field is essential to produce starting structures that should be followed by higher-level methods. In this work we employed the Universal Force Field (UFF) extended for MOFs and hydrogen bonds. There are various other forcefield alternatives available for MOFs, which typically perform superior for a particular MOF subclass, but do not provide the almost complete range of elements that are covered by extended UFF. Some can, however, produce superior structures that are comparable to those of quantum chemical approaches. For an overview on alternatives, we refer to a recent review by Boyd et al. For COFs, incorporating only 1 st and 2 nd row atoms, a large variety of standard force fields are available. DFTB. The DFTB method is an approximation to density-functional theory (DFT). It employs a minimal atomic-orbital valence basis and involves two-center approximations to the Kohn-Sham matrix elements, thus, allowing the Slater-Koster technique for pre-calculation of all integrals, which is the key ingredient to allow a performance boost of about three orders of magnitude with respect to DFT. It is available in three variants, the original DFTB form, DFTB0, does not correct for interatomic charge transfer. The self-consistent charge (SCC) approximation, SCC-DFTB, also known as DFTB2, improves the transferability, in particular to polar systems. It has been later improved to the third-order level (DFTB3). Most widely applied parameters include the mat-sci parameters for DFTB0, the mio parameters for SCC-DFTB, and more recently the 3ob parametrization, applicable to DFTB2 and DFTB3. Spin polarization and spin-orbit coupling (SOC) can also be included. London dispersion interactions are commonly treated using an a posteriori correction or the many-body dispersion (MBD) approach. DFTB2 has been successfully applied to 2D COFs, [27,59,78, DFTB3 has not yet been tested carefully with SOC and with periodic boundary conditions. Popular codes covering MOFs and COFs in periodic boundary conditions are dftb+ and AMS/DFTB. The method allows to determine the structure, stacking, electronic, including topological, properties in a first screening approach. An interesting approach to tackle large MOFs and COFs without interlayer conjugation is the Fragmented Molecular Orbitals approach (FMO-DFTB) developed in the Irle group and available in Gamess-US. For more elaborate investigations, DFTB should be substantiated by DFT. DFT for periodic systems. DFT has become the working horse for the calculations of structures and properties of MOFs and COFs. Unfortunately, there is no DFT software available that simultaneously offers all necessary features for this materials class. Periodic systems are typically efficiently treated using plane wave basis sets in conjunction with the projected augmented wave (PAW) method. They are readily available in commercial (VASP, CASTEP ) and free (QuantumEspresso ) software. Large pores and the possibility to directly compare to molecular systems makes also local basis function approaches interesting, with typical representatives being AMS/BAND, CRYSTAL, cp2k, and FHI/aims. They also offer access to computationally affordable hybrid functionals PBE0 and HSE06. London dispersion interactions need to be accounted for, most popular is the D3 approach by Grimme or the MBD approach. Local basis functions should have triplezeta quality with polarization functions to allow for the correct description of the decay of the electron density into the pore centers. For geometry optimizations and low-level band structure calculations, a gradient corrected local functional is sufficient, most prominently the PBE functional is employed, while band gaps should be improved using a hybrid approach, such as PBE0 or HSE06. For heavier nuclei, scalar relativistic effects can be treated either using the zero-order regular approximation (ZORA), relativistically corrected pseudo-potentials or effective-core potentials. DFT allows also the calculation of spin-spin couplings, which are necessary to account for magnetic ordering. In very demanding cases, wave-function based methods, such as complete active space self-consistent field (CASSCF) calculations, have been employed to accurately describe excited states. Beyond DFT. There are well-known limitations for electronic structure calculations using DFT, most notably the too-narrow band gaps. A thorough validation of using methods beyond DFT is, therefore, desirable, even though often impossible due to the computational costs. Options include the quasi-particle model GLLB-SC, which is available in AMS/BAND, and GW and RPA methods, as provided in FHI/aims and VASP. Layer-by-layer growth technique. While the bulk powder form of MOFs is only of limited interest with regard to the fabrication of stacked layers, the layer-by-layer (lbl) or quasiepitaxial growth of SURMOFs offers a number of interesting options. Briefly, the lbl process proceeds by repeating a four-step cycle comprising the sequential immersion into the two solutions of the individual reactants, separated by rinsing with a solvent (typically ethanol) (Figure 11). Thus, instead of mixing the reactants like in the conventional solvothermal bulk MOF synthesis, the metal or metal/oxo source is kept apart from the linker solution. Another important difference is that the lbl synthesis is carried out at temperatures much lower than these used in the solvothermal synthesis, in some cases growth already occurs at room temperature. To nucleate the growth of SURMOFs on the substrate, the exposed surface first needs to be functionalized with specific functional groups, thus, yielding -COOH, -OH or pyridyl-terminations. The first step typically is an exposure to the metal source (e.g., Cu-or Zn-acetate). Subsequently, the other steps (rinsing with ethanol, exposure to organic linker, rinsing with ethanol) are repeated until the desired thickness is reached. After synthesis, the SURMOF structure can be rigorously characterized by out-of-plane and in-plane x-ray diffraction. Figure 11. Layer-by-layer growth of a HKUST-I SURMOF on a quartz-crystal microbalance (QCM) crystal. From the frequency shift of the QCM substrate, the deposited mass can be determined in a quantitative fashion. Note that the growth is self-terminating, e.g., for the metal source the increase of mass saturates after a few minutes. Data are taken from Ref. . With regard to the topic of this review, the fabrication of stacked layers, it is important to note that SURMOF heterolayers, i.e., stacked 2D crystalline coordination networks, can be realized in a straightforward fashion by switching to different linkers and/or nodes during the lbl process. Thus, a programmed assembly of arbitrary sequences of MOF-heterolayers becomes possible. While in most cases stacked heterolayers are achieved by switching the organic linkers, in a few cases, also the metal nodes have been varied, e.g., in case of Cu/Zn. Particularly interesting, with regard to realizing mixed-metal MOFs, are lanthanide-based SURMOFs, since the very similar coordination chemistry of these metal ions allows switching between the different elements without pronounced changes of the MOF structure. Crystalline framework materials offer an additional means for functionality design: controlled van-der-Waals stacks of aromatic building blocks are subject to the proximity effect and raise phenomena such as wide absorption bands, mobile charge carriers, or indirect band gaps. Such stacks are present in layered MOFs and COFs, and can be created via steric control in 3D frameworks.

chemsum

{"title": "Proximity effect in crystalline framework materials: stacking-induced functionality in MOFs and COFs", "journal": "ChemRxiv"}

an_efficient_implementation_of_semi-numerical_computation_of_the_hartree-fock_exchange_on_the_intel_

## Abstract: Unique technical challenges and their solutions for implementing semi-numerical Hartree-Fock exchange on the Phil Processor are discussed, especially concerning the single-instruction-multipledata type of processing and small cache size. Benchmark calculations on a series of buckyball molecules with various Gaussian basis sets on a Phi processor and a six-core CPU show that the Phi processor provides as much as 12 times of speedup with large basis sets compared with the conventional four-center electron repulsion integration approach performed on the CPU. The accuracy of the semi-numerical scheme is also evaluated and found to be comparable to that of the resolution-of-identity approach. ## Introduction The density functional theory(DFT) is one of the most efficient quantum mechanical methods for chemical and material studies. In a practical DFT application the computational cost has two major components. One is the numerical integration of the exchange-correlation (XC) potential with a DFT XC functional . The other is the computation of the Coulomb and Hartree-Fock (HF) exchange matrices based the electron-repulsion integrals (ERIs) . The latter is typically the ratedetermining step for most HF and DFT calculations. Between the HF exchange and Coulomb, the latter can be treated efficiently with various techniques , leaving the HF exchange as the most time-consuming step in a hybrid DFT calculation. Hence it is important to develop and implement efficient algorithms for fast HF exchange calculation. One way to make DFT computation more productive is to take the advantage of new computer technologies, such as the General-Purpose Graphics Processing Units(GPGPU) and the Intel Many Integrated Core(MIC) architecture. The GPGPU extends the use of the graphics processing unit to perform general purpose computation such as scientific calculations. Compared with a traditional CPU, a GPGPU card possesses more processing cores and overall has much higher theoretical peak floating point operations per second(FLOPS). On the other hand the Intel MIC technology is a redesign of the previous generation of Celeron and Pentium cores and condenses these cores onto one processor. Although the MIC architecture has far fewer cores than the GPGPU(currently about 60-70 cores per processor), it introduces the Vector Processing Units (VPU) to each core, which can execute 8 double-precision floating point operations per CPU instruction cycle . With the VPU working as a multiplier the MIC processor is capable of providing similar computing capacity as that of the GPGPU card. The first generation of MIC architecture, also known as Xeon Phi, was introduced as "Knight Corner"(KNC) by Intel in 2012. It is a "coprocessor" and can only work through the PCI-express lane to assist the calculation load. The KNC coprocessor has 61 cores, each with one VPU at frequency of 1.0 -1.2 GHz. The total theoretical FLOPS count for KNC coprocessor is about 1 teraFLOPS in double precision. In comparison, the Nvidia Tesla K40 GPGPU card has a theoretical peak performance of 1.43 teraFLOPS in double precision. In 2016, the second generation of Xeon Phi "Knight Landing"(KNL) was unveiled. The KNL was available in two forms, either as a coprocessor or as a host processor (CPU). A KNL processor has 64-72 cores, with two VPUs on each core at a frequency ranging from 1.3 to 1.5GHz. The KNL processor is able to perform about three teraFLOPS in double precision, which is comparable to the Nvidia Tesla K80 GPGPU card. The unprecedented potential computing power of these new types of hardware have attracted significant attention of quantum chemistry software community, especially GPGPUs. Yasuda and Ufimtsev etc. did a pioneer work on using the GPGPU card to calculate the four-center ERI to form the HF exchange and Coulomb matrices for HF and DFT applications. Their study was mostly concentrated on the low angular momentum integrals involving S and P type Gaussian functions. Later Asadchev etc. extended the ERI calculation to high angular momentum cases up to the G-type Gaussian function . The combination of the GPGPU and CPU for the ERI calculation brought 2-3 times speedup for the HF exchange and Coulomb matrix calculation in terms of the S, P and D types of Gaussian functions compared with an eight-core CPU server. Other four-center ERI implementations on the GPGPU for the HF exchange and Coulomb calculation were also reported with noticeable speedups compared with the traditional CPU implementations. GPGPU implementations for post-HF wavefunction methods have also been reported besides the calculation of Coulomb and HF exchange . Less development efforts have been reported for the Phi processor in computational quantum chemistry , perhaps because it came about more recently than the GPGPU. Leang et al studied the efficiency of matrix operations on the Phi coprocessor and showed that the Phi coprocessor can yield up to three times speedup compared with the host CPU. Apra etc. employed the Phi coprocessor for CCSD(T) calculation , and the implementation also achieves 2-3 times speedup by utilizing the Phi coprocessors. On the other hand, Reid etc. modified CP2K Program for the Phi coprocessor and made performance comparison with a dual Xeon CPU(16 cores) system for energy calculations, and found that the code ran about 5.43 times slower on the Phi coprocessor than on the CPU-only system, even though the theoretical FLOPS count for the dual CPUs is about 0.8 teraFLOPS versus 1 teraFLOPs for the Phi processor. The second generation of Phi processor supports the direct compilation and execution of the software on the processor. Such an attempt for the quantum chemistry software LSDalton showed that it is still several times slower on the KNL than on a dual Xeon CPU system. The difficulty was mostly attributed to the inefficient use of VPUs, which are the SIMD(Single instruction, multiple data) processing units in the Phi processor . Unfortunately, the native ERI algorithm belongs to the MIMD(multiple instruction, multiple data) type of operation and it is technologically challenging to optimize for the Phi processor. Although the Phi processor may have more cores than the host CPU, each core inside the Phi processor is much weaker than the normal core in the CPU. One needs to use the VPU effectively in order to take the full advantage of the Phi processor. Therefore, it is necessary to rewrite software significantly so that it can be efficiently executed on an SIMD architecture. We have recently published the implementation of a semi-numerical integration for computing the HF exchange energy and matrix with self-consistent field(SCF) for conventional CPUs . In this work we describe the implementation of this scheme for the MIC architecture. The scheme is similar to the COS-X algorithm and pseudo-spectral scheme and has been described in a separate publication . We have shown that the semi-numerical scheme is more efficient for large basis sets and large molecules due to quadratic scaling with respect the basis set size. Furthermore, it requires fewer temporary variables therefore fit better the Phi Processor that has much smaller cache size than a standard CPU. ## The Semi-numerical Algorithm for HF Exchange The semi-numerical algorithm to calculate the HF exchange matrix has been discussed in detail in our previous paper . We briefly summarize it here. The HF exchange matrix is derived through the differential of the HF exchange energy with respect to the spin-resolved density matrix P σ µν : where ϕ represents a general Gaussian basis function, µ, ν, λ and η are basis function indexes. σ denotes the spin of a Molecular orbital, either α or β. (µλ|νη) is the abbreviation of four-center ERI. The conventional way is to evaluate this four-center ERI analytically and then form the HF exchange matrix. But the HF exchange matrix can also be calculated through the following semi-numerical scheme: where ϕ µ (r) is the value of a basis function at a given grid point r. w(r) is the numerical weight of the grid point. R σ ν (r) is defined as the kernel of the semi-numerical HF exchange matrix. In this scheme the numerical integration is performed with the standard DFT atom-centered grid scheme . To calculate the kernel R σ ν (r) we need to first combine the spin-resolved density matrix with basis set value ϕ(r) through a BLAS(Basic Linear Algebra Subprograms) level 3 matrix-matrix multiplication: The kernel of R σ ν (r) can then be expressed as: where v νη (r) is the two-center electrostatic potential (ESP) integral: In our implementation the code for the ESP integral generated with an integral code generator program we published earlier called CPPINTS . It is based on a combination of the Obara-Saika(OS) and Head-Gordon-Pople(HGP) algorithms, and applies a greedy search algorithm to produce the minimal number of temporary variables. The rate-determining step in the above scheme is to calculate the ESP integral at each grid point and involves all the effectively nonzero basis function pairs. In practice a basis function is usually a linear combination of primitive Gaussian functions, and thus the ESP integral calculation also involves a loop over the contraction for each pair of Gaussian basis functions. The total cost of ESP integrals then can be approximately estimated as O(N grids * N bf pairs * N 2 contraction )("bf stands for "basis function). Since the number of these basis function pairs increases linearly with respect to the molecular size, the overall cost of the above algorithm scales quadratically with respect to the molecular size with a given basis set. This scaling is the same as the conventional approach, and better than the cubic-scaling of the approximate resolution-of-identity (RI) scheme. On the other hand, each step in this scheme involves only two basis function indexes, and thus scales quadratically with respect to the basis set size for a given molecule, which is superior to the quadruple-scaling of the conventional approach. We have implemented of the above scheme and showed through the calculations of a series of alanine peptides on a normal multicore CPU that the semi-numerical scheme becomes competitive to the conventional analytical method for large basis sets and can be about six times faster for the aug-cc-pvtz basis. ## Implementation on the Phi Processor Our first try with the Phi processor is to compile the aforementioned implementation of the semi-numerical scheme with the Intel compiler that has the automated optimization feature for the Intel MIC architecture. As a result the test on the Phi processor KNL7250 is two times slower than running on one six-core E5-1650 CPU, significantly underperforms with respect to the potential of the Phi processor. Thus, the code needs to be rewritten for the latter. Here in this section we will discuss the special features of the Phi processor, and our experience on the implementation for it. The processing unit in each core of a Phi processor is divided into two parts, a redesigned Pentium/Atom core and two VPUs in one core in KNL processor. A VPU is a SIMD processing unit , and able to process 8 double precision floating point number simultaneously. As a comparison the floating-point unit(FPU) in a CPU can only handle one double precision floating point number per each instruction cycle. The standard instruction set for the VPU on the KNL processor is the avx-512 instruction set. Modern compilers like Intel compilers contain the so-called automatic vectorization function that can compile the code into this instruction set. However this function does not always work well, especially for quantum chemistry software where usually the most timeconsuming part is often a collection of MIMD operations. As a consequence, a quantum chemistry program running well on a CPU may still need to be efficiently vectorized manually to fit the SIMD structure of the Phi processor. As discussed in the above algorithm section the calculation of ESP integrals is the most timeconsuming part. Therefore it is imperative to rewrite the ESP integral code that transforms the original MIMD operations to efficient SIMD operations for the VPUs in a Phi processor. The computation of the ESP integral does not fit the VPU naturally because it is a collection of "MIMD" operations in the native formulas. The piece of code below is part of the innermost loop for computing the ESP integral for a (D, D) shell pair on the CPU: As one can see, all the operations are scalar operations and fit the FPU well for a CPU. Although the compilers can directly compile the above code into the avx-512 instruction set through the automatic vectorization function, the efficiency of the resulted code is very poor since every scalar has be to read in separately and each read operation costs a few precious processing cycles. To vectorize the code for the Phi processor one needs to reorganize the code into an aligned array structure. For the above example piece of code, we can arrange the variables into two C style structs: The C structs are aligned to the 64-bit cache line of the Phi processor to make the data access as efficient as possible. The above sample code then can be trivially rewritten using the structs. The advantage of this change is that all the data members of a struct are pulled into the VPU register with one read operation, and thus the execution of the computation requires far fewer processing cycles for data transmission between the register and the cache. Another important factor for an effective implementation on the Phi processor is the cache size. Compared with the traditional CPU architecture, the Phi processor has much smaller L1/L2 cache and register. The L1 data cache size for both is 32 KB per each core, and the L2 cache size per each core is 512KB. Because the work on each core can be shared by 2-4 working threads, the effective L1 and L2 cache sizes for practical usage also need to be divided by a factor of 2-4. In comparison, the total cache size of a recent generation of Xeon CPU is about 2MB per core. Here the semi-numerical scheme has a distinct advantage over the analytical four-center ERI approach. The ESP integrals in the semi-numerical scheme is a type of two-center integrals, it uses far fewer temporary variables for integral calculation than the traditional four-center ERI and thus a better fit for the Phi processor. For effectively loading the data from main memory to cache, it is important to utilize the data prefetching pragma to maximize the cache hits on the Phi processor. A simple example below shows how it can be used for assigning values to an array: for ( int i=0; i<N; i++) { #pragma prefetch p:1:16 #pragma prefetch p:0:6 for (int j=0; j<2*N; j++) { p[i*2*N+j] = -1; } } the syntax "p:1:16" means to prefetch the content of array p into the L2 cache for the next 16 cycles, and "p:0:6" means to prefetch the content of p array into L1 cache for the next 6 cycles. Because the cache size in the Phi processor is significantly smaller than the normal CPU, the automatic prefetching data operations by compilers on the Phi processor is designed to be much less aggressive than the CPU . Thus maximizing the use of cache is left to the programmer. Our experience shows that applying the prefetch pragma effectively can boost the performance of the whole implementation by 2-3 times. We have applied the use of the C structs and prefetching to all of the ESP integrals of various shell pair combinations up to (F,F) shell pairs. There are 15 integral files in total we have modified for adapting to the Phi processor. The aligned C structs is able to to convert the original scalar operations into array operations which fits the avx-512 instruction set, and the prefetch pragma is able to efficiently load the input and output data into the local cache of the Phi processor. The efficiency of such a change for the Phi processor is demonstrated in section 5. We have applied the use of the C structs and prefetching to all of the ESP integrals of various shell pair combinations up to (F,F) shell pairs. Each ESP integral file corresponds to a shell pair combination, and each file has multiple places where the data reorganization is needed for the computation of the fundamental integrals. Furthermore each integral file also has multiple loops that require data prefetching. There are 15 integral files in total we have modified for adapting to the Phi processor. The aligned C structs is able to to convert the original scalar operations into array operations which fits the avx-512 instruction set, and the prefetch pragma is able to efficiently load the input and output data into the local cache of the Phi processor. The efficiency of such a change for the Phi processor is demonstrated in section 5. ## Accuracy Assessment for the Semi-numerical HF Exchange Algorithm To access the accuracy of the semi-numerical implementation on the Phi processor, we performed a benchmark calculation over a series of fullerene molecules with standard basis sets of different sizes. The fullerene molecules include C60, C100, C180 and C240, all in the buckyball form. The basis sets chosen for the test are all commonly used basis sets , namely Pople Basis set 6-31G** and 6-311G(2df), and Dunning Basis set cc-pvdz and cc-pvtz. The implementation of the above semi-numerical scheme is based on the standard atom-centered grid scheme, which employs the Euler-Maclaurin formula for the radial part, and Lebedev grids for the angular part. Three different sets of pruned grid were used. One is the SG-1 grid with each carbon atom having about 3000 points. The other two are the "standard" and "fine" grids from the PQS program . The standard grid employs approximately 5000 points for each Carbon atom, and the fine grid about 9000. The errors of the semi-numerical scheme are calculated using the standard analytical four-center ERI method as the reference. All the HF SCF calculations employ the core Hamiltonian as the initial guess, which derives the initial density matrix through The diagonalization of the one-electron Hamiltonian matrix. The reference HF exchange matrix was computed with an integral threshold of 1.0 × 10 −10 . For the calculations with the semi-numerical scheme, both thresholds for the significant basis function pair and the ESP integral are set to 10 −12 . Tables 1, 2 and 3 list the errors of the semi-numerical scheme for the computation of the HF exchange matrix and the total HF energy in root-mean-square deviation (RMSD), maximum absolute deviation (MAD) and total energy difference per atom. As one can see the RMSD for the HF exchange matrix is at the level of 1.0 × 10 −6 , which is two orders of magnitude smaller than that of MAD. The accuracy for the HF exchange matrix improves as the grid becomes finer. But the fineness of the grid does not appear to affect the accuracy of the energy measured as the total energy per atom. The largest errors for the energy is between 150 to 200 µE h /atom, comparable to the error of the Resolution-of-Identity method for the HF exchange . It is not clear why the accuracy in energy does not respond to the fineness of the grid while the matrix does. We should note that these pruned grids are fine-tuned for the computation of DFT exchange-correlation energy. In our previous test , we observed that the error in energy with an unpruned grid set (128,302) (the radial part is 128 points and the angular part uses 302 points per each atom) is about 5 times smaller than the one with the SG-1 grid. More studies are needed to understand and reduce this discrepancy. The energy error per atom also grows as the system size become larger, as shown in Table 3. This can be understood by the formula of the semi-numerical scheme for the HF exchange matrix: The error in the HF exchange matrix calculation mainly comes from the numerical computation of the basis function pair ϕ µ (r)ϕ λ (r). Because the HF exchange is a long-range interaction, the ESP originated from the point r may still effectively interact with the numerical basis function pair to produce a numerically significant contribution to the HF exchange matrix even though the distance between the r and r is large. As the molecule grows larger, it can be expected that the number of ESP integrals grows accordingly, resulting in a multiplier effect for the error on the HF exchange matrix calculation. On the other hand, as the distance between r and r increases the first order reduced density matrix ρ(r, r ) actually decreases . For the HF exchange energy expressed in first order reduced density matrix form One can see that the decrease of the ρ(r, r ) makes the contribution from the r become smaller, which damps the contribution from basis function pairs at large distances. As the system becomes larger these two factors cancel each other out to an extent, one may expect that the increase of the error on the energy becomes slower. Such phenomenon can be observed in the Table 3. The damping effect of the density matrix can be more pronounced for a stretched system. To test this effect, we performed the HF SCF calculation on a series of alpha helix alanine peptides with a linear tertiary structure. The alanine series include peptides with 35, 40 and 45 alanines, respectively, and the calculations were performed with the SG-1 grid and Pople basis 6-311G(2df,2pd) . All of the other settings are the same as for the fullerene series. The total energy difference per atom is shown in Table 4. It is evident from the table that as the system is extended sufficiently the total energy difference per atom becomes stable. The reason is that the large end-to-end distance results in many effectively zero elements in the distance-decaying reduced first order density matrix ρ(r, r ). ## Efficiency Assessment for Semi-numerical Scheme The computational performance of the semi-numerical scheme for HF exchange implemented and optimized for the Phi processor is also benchmarked with the fullerene series. The Phi processor for the test is a KNL7250 with 68 cores and each core is at 1.40 GHz. The computing time on the Phi processor is compared with that on a six-core E5-1650 CPU at 3.2GHz. The code run on the CPU was from the earlier implementation of the semi-numerical scheme optimized for CPUs. The reference calculations with the analytical four-center ERIs were performed on the CPU only. Table 5 lists the processor timing data for the formation of HF exchange matrix for one SCF iteration. The numerical grid for the semi-numerical scheme is SG-1. The number listed in the parenthesis next to a CPU time is the ratio of the CPU time to the processing time on the Phi processor. All the calculations on the CPU were with multi-threads. Judging the performance of an implementation on the Phi Processor needs a reference. Intel has provided a series of showcases for the KNL7250 processor 1 . Top performers are able to achieve 1.0 to 1.5 times of speedup on KNL7250 processor compared with the 36-core dual E5-2697(each core is at 2.3GHz) CPU system. We estimate this CPU is about 4.3 times faster than the E5-1650 CPU used in our test( 2.3 × 36 3.2 × 6 ≈ 4.3). Therefore we consider a 4-6 time speedup to be good in our test. As shown in Table 5 our implementation of the semi-numerical HF exchange for the Phi processor is able to achieve approximately 3-7 times of speedup relative to the implementation for the CPU. This indicates the success of our implementation. Additionally it obtains 7-12 times of speedup in comparison with the traditional analytical integral calculation with large basis sets such as 6-311G(2df) and cc-pvtz. For moderate double-zeta basis sets 6-31G** and cc-pvdz, the speedup ratio is between 1.5-4 times. The difference on the speedup between the different basis sets can be explained by the semi-numerical algorithm itself, as it scales quadratic with respect to the basis set size, in contrast to the quadruple scaling with the four-center ERI method . For a given molecule the numerical grids is unchanged for different basis sets, therefore if the "density" of basis functions per molecular size increases, the semi-numerical scheme becomes more efficient per grid point. As a result the semi-numerical scheme is favored for larger basis sets such as 6-311G(2df) and cc-pvtz. ## Conclusion Efficient implementation of quantum chemistry methods to take the full advantage of new computer technologies is important for the large scale applications. In this work we extended our previous implementation of a semi-numerical method for HF exchange to the Intel Phi processor. The utilization of VPUs on the Phi processor, which operates in a SIMD fashion, imposes a major challenge for implementation since algorithms for the ERI naturally fits the MIMD-type. The other major technical obstacle is the relatively small cache size. These difficulties are overcome by organizing the data for computing various types of fundamental integrals to be aligned with the VPU register, and prefetching necessary data for effective use of the cache. The accuracy and efficiency of the implementation are evaluated with the computation of a series of buckyball molecules with basis sets of different sizes. The results showed that the accuracy of the semi-numerical implementation is similar to that of the RI method. The computational efficiency of the implementation matches the achievements of other types of software for the Phi processor. The advantage of the combination of the semi-numerical algorithm with the Phi processor can be seen through the 7-12 times of speedup relative to the computation with the traditional four-center ERI method on a six-core E5-1650 CPU with 6-311G(2df) and cc-pvtz basis sets. ## Acknowledgement This project is supported by Petroleum Research Funds (PRF 55229-DNI6) and National Science Foundation (CHE-1665344). F. Liu is grateful to the Colfax International and Intel Remote Access Program for free access to the Intel Phi processors. The authors thank Dr. Peter Pulay and Dr. Jon Baker for letting us use their pruned DFT grids.

chemsum

{"title": "An Efficient Implementation of Semi-numerical Computation of the Hartree-Fock Exchange on the Intel Phi Processor", "journal": "ChemRxiv"}

a_data-science_approach_to_predict_the_heat_capacity_of_nanoporous_materials

## Abstract: The heat capacity of a material is a fundamental property that is of significant practical importance. For example, in a carbon capture process, the heat required to regenerate a solid sorbent is directly related to the heat capacity of the material.However, for most materials suitable for carbon capture applications the heat capacity is not known, and thus the standard procedure is to assume the same value for all materials. In this work, we developed a machine-learning approach to accurately predict the heat capacity of these materials, i.e., zeolites, metal-organic frameworks, and covalent-organic frameworks. The accuracy of our prediction is confirmed with novel experimental data. Finally, for a temperature swing adsorption process that captures carbon from the flue gas of a coal-fired power plant, we show that for some materials the heat requirement is reduced by as much as a factor of two using the correct heat capacity. ## Introduction Metal-organic frameworks (MOFs) and related porous materials are promising candidates for a wide range of energy-related applications, including gas separation, gas storage, and catalysis. In recent years, considerable effort has been focused on the understanding and tuning of the adsorption 4,5 and catalytic 6,7 properties of these materials due to the significance of these applications in mitigating the negative impact of human related activities on the environment. For many applications, the thermal properties of the materials also play a role in their performance. For example, the energy penalty of a carbon capture process using a temperature swing adsorption process (TSA), where the adsorbent material is regenerated by heating the adsorption bed, is directly influenced by the heat capacity of the adsorbent material. 12 Often, it is assumed that the heat capacity is a constant for all materials in these energy penalty calculations. However, there is little foundation for this assumption, and it is rather a pragmatic simplification. Only few experimental studies on the heat capacity of these materials exist,, 16,17 and their reported values have been used throughout the literature ever since for the energy penalty calculations. Currently, we lack an understanding of how the heat capacity is related to the underlying crystal structure. Also, we do not have a methodology to accurately and efficiently evaluate the heat capacity at scale for the enormous number of available porous materials. Developing methods to accurately predict the heat capacity is essential to improve the accuracy of large-scale performance evaluation of porous materials in different chemical processes and to discover the best performing materials. In this work, we have used the state-of-the art quantum mechanical calculations to accurately predict the heat capacity of a set of representative MOFs. However, as these calculations scale with the number of atoms (N a ) as O(N 4 a ), they can only be carried out for a subset of MOFs for which the number of atoms is not too large. We show that we can take advantage of the fact that the heat capacity can be estimated as a local property, and hence use this data as a training set for a machine-learning methodology that enables fast, accurate, reliable prediction of the heat capacity of these materials, even if the number of atoms far exceeds the limit of our quantum calculations. We evaluate the accuracy of this approach using newly measured experimental data. In addition, we show that for carbon capture applications, the assumption of a constant heat capacity of these materials has resulted in a significant overestimation of the energy requirements. ## Theoretical aspects Theoretically, the heat capacity, which is a measure of how much energy it takes to raise the temperature of a solid, is related to the lattice vibrations or phonons. Lattice vibrations are the collective motions of atoms of a crystal, and the heat capacity is proportional to the change of the average energy of these collective vibrational modes. Within a harmonic approximation of lattice vibrations, the heat capacity is formulated as a summation over the contribution of these lattice vibrations via: where N is the number of vibrational modes (equal to three times the number of atoms), ω the vibrational frequency, T the temperature, and h and k B Planck's and Boltzmann constants, respectively. To evaluate the heat capacity, the lattice vibrational frequencies can be extracted from the Hessian matrix, that is, the second-order partial derivatives of the lattice energy with respect to atom displacements. For the temperature range of interest for gas separation applications (i.e., 260-400K), Kapil et al. 18 have shown that the contributions from gas molecules to the total heat capacity of the gas-framework system is a function of the gas molecule and proportional to the amount of adsorbed gas. This is due to the fact that the contribution of host-guest interactions to the heat capacity are negligible in this temperature range. As the contribution of gas molecules can be added to the heat capacity of the empty framework, in this study, we focus on predicting the heat capacity of an empty framework. To account for anharmonic and nuclear quantum effects, path integral molecular dynamics simulations can be used. 18,19 Nuclear quantum effects are of particular importance at very low temperatures. We use the harmonic approximation, as previous studies 18,19 have shown that it gives sufficient accuracy for the heat capacity of an empty framework for the temperature range of interest for gas separation applications. However, it is important to note that the harmonic approximation might fail to correctly describe flexible materials, and it will be interesting to investigate the extent of validity of this approximation for estimating the heat capacity of these materials in the future. The details of computational procedure to compute harmonic frequencies are explained in method section and SI. The harmonic formulation of the heat capacity follows the experimental observation of how the heat capacity of solids change with respect to the temperature. The heat capacity tends to zero at low temperatures, where only low-frequency modes are occupied, and then converges to a constant (3R) at high temperatures, where all vibrational modes are active. This behavior is shown for MOF-74 in Figure 1, where we show the density functional theory (DFT) computed heat capacity of the materials with two different metals, Cobalt and Zinc (Co-MOF-74 and Zn-MOF-74, respectively). In this figure, we also compare the DFT results with experimental data. For these experiments, it was essential to ensure a proper activation of the material. The details of the synthesis and experimental procedure can be found in the method section and SI. We observed a good agreement between computed values and the experimental measurements. These results provide confidence for our computational methodology and supports the conclusion of the previous work 18 that the harmonic approximation provides sufficiently reliable data on the heat capacity of MOF materials. ## Understanding structure-heat capacity relationships From a reticular chemistry point of view, a MOF structure is an assembly of metal nodes and organic linkers on a network topology. Therefore, it is important to understand how variation of each of these factors influences the heat capacity. We provide this understanding by analyzing the heat capacity of MOF-74 family and a set of zeolitic imidazolate framework (ZIF) structures. Isoreticular MOF-74 structures are an ideal case for understanding the role of metal on the heat capacity as it has been made with a wide range of metals, including Mg, Mn, Fe, Ni, Co, and Zn. 20,21 These materials have been studied extensively for carbon capture. 12,20,21 Figure 2a shows the computed heat capacity of these materials at room temperature, which can indeed vary considerably with the change of metal. For example, the Mg-MOF-74 shows the lowest molar heat capacity amongst these six metals. To explain this observation, we inspect the lattice vibrational frequencies of these systems. As the only difference between 5 these structures are their metal centers, we focus on the lattice vibrational frequencies of the metal centers. Figure 2d shows the histograms of the projected lattice vibrational frequencies of metal centers for Mg and Zn MOF-74, where we see that Zn has a shift in the vibrational frequencies to lower frequencies compared to Mg. Theoretically, different frequencies contribute to the heat capacity differently: At low and intermediate temperatures, the low-frequency modes are dominant in contributing to the heat capacity, and only at high temperatures the high-frequency modes also participate in the heat capacity (see Figure 2b for a quantitative picture). Substituting Mg with Zn shifts the vibrational frequencies to lower frequencies, and hence, leading to a higher heat capacity in Zn-MOF-74. One can explain the effect of changing the mass of the metal node if we envision this metal to be connected with harmonic springs to its neighbors. The vibrational frequency of this metal is proportional to k /m eff , where k is the spring stiffness and m eff is the effective mass. If change of the mass would be the only contribution, one would expect monotonically increasing heat capacity with increasing the molecular mass of the metal. Figure 2a shows that differences in mass only partly explains the data; when the masses are similar, the details of the metal-linker interactions are also important. It is important to note that in practice, the heat capacity is reported per gram of material. As the weight of the material increases more than the heat capacity, we typically see a decrease of the heat capacity if plotted per gram of material instead of per atom. A more affordable way to describe the interactions between the atoms of a MOF is to use a classical force field. For example, the universal force field (UFF) can predict the mechanical properties of a MOF with similar accuracy as DFT. 22,23 However, Figure 2a shows that this force field significantly underestimates the heat capacity. If we softened the spring constants between metals and linkers, we can obtain better agreement with the DFT results. However, one would need a metal-linker dependent correction factor to correctly capture the subtleties of the chemistry, which limits the usefulness of the UFF force field to predict the heat capacity of MOFs. Further analysis (see SI) shows that the deficiencies of the UFF force field is mainly due to the metal-linker interactions, whereas the linker contributions are well described. As MOF-74 has relatively high metal content, the errors are relatively large. Another important variation in MOF structures is the modification of linkers and the underlying network topology. To better understand how these modifications influence the heat capacity, we look at a set of zeolitic imidazolate framework (ZIF) structures with different organic linkers. These materials are ideal for this investigation as they can be synthesized with the same topology but varying linkers, as well as the same chemistry of linkers but varying topology. However, the number of atoms and the dimensions of the unit cell of these ZIF structures are too large for DFT calculations, and therefore we used the UFF force field. As the metal is not changing among the ZIFs, one can expect that we underestimate the experimental value, but this underestimation is equal for all ZIFs. We analyze the heat capacity of 200 hypothetical ZIF structures assembled using four unique linkers and 50 topologies. 24 Figure 2c shows the heat capacity of the ZIFs. We observe a large dependence of the heat capacity on the linker functionalization (dots with different colors) and a very small influence by the framework topology (dots with the same color) on the heat capacity of the materials. Comparing structures with the dichloroimidazolate (dcIM) with imidazolate (IM) linkers is instructive to further understand the role of linker functionalization, as the only difference between these structures is replacement of hydrogen atoms on the linker with chlorine. Clearly, replacing the hydrogens with heavier atoms (i.e., chlorine) increases the heat capacity of the material, similar to the simple spring-mass model described above, where the heavy chlorine atoms shift the vibrational frequencies to smaller values, leading to a higher heat capacity per atom. ## A machine learning approach As we have shown in the previous section, a force field approach gives interesting qualitative insight, but the DFT level of accuracy is needed to make quantitative predictions of the heat capacity. However, due to the cost of these DFT calculations, we can only compute the heat capacity for a small subset of materials. Therefore, a machine learning approach is an attractive alternative. A machine learning model exploits the similarities between structures in the training set to predict the target properties of new structures in the test set without performing the actual computations or experimental measurements on those. The similarity between different materials and structures can be explicitly encoded by chemically motivated descriptors or learned as part of a deep learning framework. 28,29 For example, machine learning models based on the chemical formula 26,30 or property-labeled materials fragment descriptor 25 were found to be effective in predicting heat capacity of solids and semiconductors. 31 However, these approaches are based on a global description of the entire material. Such an approach is of limited use for MOFs as their chemical diversity is so large that one needs an extremely large data set to be able to train a model based on such a global description. As we have to rely on DFT calculations, we can only generate data for a small subset of MOFs. In this work, we developed an alternative machine learning approach that is not based on such global descriptors. Recall that the heat capacity is given by the summation over atomic vibrations (equation 1). Hence, we can build a machine learning model to directly predict these atomic contributions. The importance of this step is that we can now make use of the locality approximation in our featurization. Especially, our simulations on ZIFs show that changes in the pore topology have a minor effect on the heat capacity, indicating that the relevant chemical environment is relatively short-ranged. Hence, our features only need to account for the local environment around each atom (see Figure 3b). Compared to the previous machine learning approaches, the main advantage of our methodology is that each MOF contains many different chemical environments for each atom. Hence, in our selection of our training set, we only need to ensure to generate data for a diverse set of atoms in different chemical environments. 32 The different chemical environments include different elements and coordination environments, as well as different sizes of crystals. As we can obtain such a diverse set using MOFs with a relatively small number of atoms or with high symmetry, we have a training set that we also can use to predict the heat capacities of those MOFs that are too large for DFT calculations. In essence, our machine learning approach is staying as close as possible to the physics of the problem. Once trained, it predicts the contribution of each atom to the heat capacity, depending on the local chemical environment. ## A dataset for machine learning the heat capacity We selected ∼230 structures with diverse chemical environments from the experimental structures in the computation-ready (CoRE-)MOF database, 33 experimental covalent organic frameworks (CURATED-COFs), 34 and the experimental all-silica zeolites in IZA database. 35 The DFT predicted heat capacity of these structures are shown in Figure 3a. The gravimetric heat capacity of the materials range from 0.4 to 1.1 J.g −1 .K −1 . Figure 3a shows that all silica zeolites have very close heat capacities, which justifies the use of constant heat capacity in performance evaluation of these materials in chemical processes. In contrast, MOFs and COFs show a wide range of heat capacities. The general trend is captured by the average atomic mass. However, to make quantitative predictions, we need to capture the very diverse chemistry of these materials in our machine learning approach. ## Machine learning for the heat capacity An important component in any machine learning model is the featurization. A featurization that captures the relevant physics will typically require less training data than a naive model. Our featurization includes the atom identity and descriptors to capture both chemical and geometric similarities of the local atomic environments (Figure 3b). To capture the local chemistry, we use Voronoi-tessellation based descriptors, 36,37 which give statistics of chemical heuristics for the neighboring atoms. Furthermore, we use symmetry functions 38 and AGNI fingerprints 39 to encode geometric similarities. These descriptors are expressive enough to provide enough flexibility for the machine learning model to capture the similarities between the chemical environments to predict heat capacity. We use 120 of the structures to train machine learning models to predict the heat ca- Ave. Atomic Mass training sets (see method section for details). This enables us using our machine learning models to predict the heat capacity of porous material databases with a controlled error and confidence. Figure 3c shows that indeed this approach effectively identifies those structures that the model has wrong predictions as uncertain. We can use our uncertainty-aware machine learning model to predict the heat capacity of any nanoporous materials. In Figure 4b and 4a, we show the predicted heat capacity at room temperature for the experimental structures reported in CoRE-MOF, 33 CURATED-COFs, 34 and IZA 35 zeolites databases. We note that for many of these materials, quantum calculations are not feasible due to the excessive computational costs. The materials exhibit a wide range of heat capacities, and similar to the previous section, we observe a clear relationship between heat capacity and atomic mass. It is interesting to compare our machine learning predictions with experimental measurements. In addition to the few experimental values reported in the literature, 40 we used Differential Scanning Calorimetry (DSC) to measure the heat capacity of some additional MOFs. Figure 4c shows how these experimental values compare with the machine learning predictions. Except for a few cases, we have a good match between the predicted and measured heat capacities. By carefully inspecting the experiment protocols, we realize that the materials with a large discrepancy appear to be not fully activated. Solvent molecules that remain in the pores of the material have a large effect on the heat capacity measurements. In our experimental procedure, we ensure that the MOFs are fully activated by sequential in situ activation and measurements of the heat capacity (see SI for further discussion). This may explain why some of the experimental data are considerably out of the expected range (Figure 4d). If we discard the materials that are not fully activated, these results show that the machine learning model can provide fairly accurate prediction of the heat capacity of fully activated MOFs. In Figure 2a, we presented the DFT values of the heat capacity for MOF-74 with different metals. An unexpected result is the heat capacity (per atom) of Mn-MOF-74, which is 13 Exp -Gravimetric C p (J.g 1 .k higher than the corresponding value for other similar mass MOFs, e.g., Ni-MOF-74. This is a puzzling result, as the only change is the metal, and if we increase the atomic mass we expect monotonic increase of the heat capacity due to a shift of frequencies to lower values. A feature analysis of the machine learning model shows that 60% of the heat capacity prediction can be explained by knowing the atomic mass of all atoms in the material (see Figure 3d and method section for details). The remaining 40% is the local geometry and chemistry. In the case of Mn-MOF-74, the feature analysis shows that the geometry is the decisive factor (see SI). Interestingly, if we take the Ni-MOF-74 geometry and simply replace the metal, the machine learning model gives similar values for all but Mn-MOF-74. For Mn-MOF-74, the feature analysis of the model shows that the higher predicted value of the heat capacity is caused by longer metal-linker bonds. These weaker bonds shift the frequencies to lower values. That machine learning gives us this explanation illustrates the importance of aligning the features of the model with the underlying physics. ## Discussion From a practical point of view, for new MOFs or COFs crystal structures, we can now provide an estimate of the heat capacity of the material, together with an estimate of the reliability of this prediction. One can expect that, for novel structures with chemical environments that are very different from the training set, our model will indicate that the predictions are unreliable. For these structures, we need to update the machine learning model by including additional DFT calculations. Our machine learning approach in not limited to porous materials and can be generalized to other classes of materials, provided one can generate the relevant training set. One of the practical motivation of our work is to understand the importance of accurate knowledge of the heat capacity in the use of porous materials in carbon capture processes. In particular, in a temperature swing adsorption process (TSA), one needs to heat the material to release the captured CO 2 . As the contributions from heating the adsorbed gases in the column are negligible for this application (see SI for discussion), this energy requirement depends on the heat capacity of the material. Figure 4 shows that for zeolites, it is indeed reasonable to assume that the heat capacity does not differ significantly among different zeolites. However, these results show that this assumption does not hold for MOFs. As our machine learning model gives the heat capacity with high accuracy, we can quantify the impact of the differences in the heat capacity on the ranking of materials in a TSA process. An objective for material design for a TSA process is to minimize the total heat required to regenerate the material per kilogram of recovered CO 2 . In Figure 5a, we show this heat requirement for a set of MOFs in a TSA process, for which, we compare the heat requirement using the actual values of the heat capacity with the assumption of a constant heat capacity (0.985 J.g −1 .K −1 ). 14 For the materials with a heat capacity close to the reference value (green dots), the heat requirement does not change, and hence, we do not need a correction. In contrast, for those materials with a low heat capacity (red and yellow dots), the heat requirement reduces by as much as 50% and make these the top performing materials. In fact, using the actual values of the heat capacity changes the ranking completely compared to using a constant value (see Figure 5b). Particularly for materials with a similar range of CO 2 uptake, the value of the heat capacity is prominent. For example, the ranking correlation is almost lost when we look at the materials with similar CO 2 uptake: the ones with lower heat capacity appear to be performing considerably better. As the heat capacity used in the previous studies 14 was significantly above the average, many of these studies have overestimated the energy requirements; for some materials as much as 50%. As these energy requirements contribute significantly to the cost of the capture process, our results can have significant practical impacts. ## Computations of heat capacity The lattice vibrational frequencies (phonons) were extracted from the Hessian matrix of the lattice energy within the harmonic approximation. To build the Hessian matrix, we use a finite difference approach where each atom of the crystal is displaced, and density functional theory (DFT) or a molecular mechanics force field is used to compute the force on all the atoms of the crystal upon this displacement. The eigenvalues and eigenvectors of this matrix correspond to the 3 × N vibrational frequencies and vibrational modes of the system, respectively. We use the projection of the eigenvectors on each atom of the crystal to compute the contribution of each atom to the heat capacity, that is used as labels for training the machine learning. The DFT calculations were performed within generalized gradient approximation (GGA) level of theory using Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional with DFT-D3(BJ) dispersion corrections. 41,42 We use GTH pseudopotentials, 43 and DZVP-MOLOPT-SR contracted Gaussian with an auxiliary plane wave basis set. Since the phonon calculation procedure relies on the assumption that the structures are at the minimum energy configurations that are consistent with the quantum or classical mechanical method used to describe the potential energy surface of the material. Therefore, in the DFT calculations, we use a tight optimization setting of that is described in SI to avoid negative frequencies. All lattice displacement and post-processing for vibration calculations were performed using Phonopy. 44 We use CP2K 45 for DFT calculations and LAMMPS, 46 LAMMPS-interface, 22 and phono-LAMMPS 47 for the molecular mechanics calculations. The DFT calculation recipe are adapted from our previous work (see SI for adaptations). 34 ## Measurements of heat capacity The MOFs were synthetized based on the reported procedures in the literature, outlined in detail in SI. To obtain a reliable measurement of the heat capacity, we consider MOFs with a robust solvothermal synthesis and stability of framework up to 200 • C. Furthermore, a well-defined state of sorption and clear stoichiometry at this temperature were requisites. Obtained crystalline phases were identified by matching powder x-ray diffraction (PXRD) patterns and thermogravimetric analysis (TGA) results to previous reports (See SI). The ground MOF samples were crimped in non-hermetically sealed aluminium pans. This asymmetry was assumed to be proportional to deviation of block temperature from ambient temperature, and corrected for accordingly. ## Machine learning The chemical environment of each atom is described using a feature vector comprised of elemental properties as well as descriptors for chemical and geometric similarities. For the elemental properties, we include atomic number and mass, row and column on the periodic table, covalent radii, and Pauling electronegativity. To capture geometric and chemical similarities, we use AGNI fingerprint and Gaussian symmetry functions, 38,39 in addition to the Voronoi tessellations based features on the statistics of atomic properties of neighbouring atoms. 36,37 These features were computed for each atom of the crystals using Matminer and Pymatgen. 48,49 We use gradient boosted decision trees (GBDT) machine learning model as implemented in XGBoost 50 to map the feature vectors to the heat capacity corresponding to each atom. To The final models, which are made available online, are train over all structures to exploit the full dataset. ## TSA process modelling and adsorption data To demonstrate the effect of the specific heat capacity of the adsorbent in the process per-

chemsum

{"title": "A Data-Science Approach to Predict the Heat Capacity of Nanoporous Materials", "journal": "ChemRxiv"}

efficient_discrimination_of_transplutonium_actinides_by_<i>in_vivo</i>_models

## Abstract: Transplutonium actinides are among the heaviest elements whose macroscale chemical properties can be experimentally tested. Being scarce and hazardous, their chemistry is rather unexplored, and they have traditionally been considered a rather homogeneous group, with most of their characteristics extrapolated from lanthanide surrogates. Newly emerged applications for these elements, combined with their persistent presence in nuclear waste, however, call for a better understanding of their behavior in complex living systems. In this work, we explored the biodistribution and excretion profiles of four transplutonium actinides ( 248 Cm, 249 Bk, 249 Cf and 253 Es) in a small animal model, and evaluated their in vivo sequestration and decorporation by two therapeutic chelators, diethylenetriamine pentaacetic acid and 3,4,3-LI(1,2-HOPO). Notably, the organ deposition patterns of those transplutonium actinides were element-dependent, particularly in the liver and skeleton, where lower atomic number radionuclides showed up to 7-fold larger liver/skeleton accumulation ratios. Nevertheless, the metal content in multiple organs was significantly decreased for all tested actinides, particularly in the liver, after administering the therapeutic agent 3,4,3-LI(1,2-HOPO) post-contamination. Lastly, the systematic comparison of the radionuclide biodistributions showed discernibly element-dependent organ depositions, which may provide insights into design rules for new bio-inspired chelating systems with high sequestration and separation performance. ## Introduction The transplutonium actinides Cm, Bk, Cf and Es are among the heaviest elements whose macroscale chemistry can be experimentally tested. 1,2 The physical and chemical properties of these radionuclides, however, have not been deeply characterized due to their scarcity, radioactivity, and challenging separations, which are among the most difficult ones within the periodic table. 3 Nevertheless, these elements are present in various human-driven activities, such as the generation of radioisotope thermoelectric generators, 1,4 neutron activation sources, 5 targets for super heavy element discovery, 6 and nuclear waste. 7 Past uncontrolled releases of radionuclides to the environment, either accidental or intentional, have demonstrated the need to better understand their behavior in vivo, including biodistribution and excretion, as well as to develop decorporation strategies to minimize adverse health effects in humans. 8,9 The biological outcomes from radiological contamination include both acute and chronic disorders, and their severity depends on multiple factors, such as quantity and duration of exposure. 10 Among the different types of radiological exposures, internal contamination is particularly dangerous since radionuclides can be deposited in tissues, producing long term radiological poisoning. 11 Over the last few decades, several chelating molecules have been developed to treat internal actinide contamination by forming highly stable complexes with the metals and enhancing their excretion. 12,13 For instance, the U.S. Food and Drug Administration (FDA) approved calcium and zinc salts of diethylenetriamine pentaacetic acid (Ca-DTPA and Zn-DTPA) to treat internal contamination with Pu, Am and Cm. 14 Even though DTPA is the frst drug approved to treat this type of radiological contamination, its therapeutic performance is hampered by multiple factors, including (1) a need to administer it in large quantities, 15 (2) its competition with biological ligands (e.g. transferrin and albumin) for the binding of the metal, 16 and (3) its inability to remove radionuclides deposited in organs. 17 In order to overcome the limitations of DTPA salts as decorporation agents, the octadentate hydroxypyridinonebased chelator 3,4,3-LI(1,2-HOPO) was developed. 18 This synthetic chelating molecule shows high binding affinity for actinides and lanthanides, 19,20 selectivity for f-block elements over biologically-relevant cations, 21,22 formation of excretable complexes with these elements, 23 and biocompatibility at therapeutic dosages. 23 In vivo studies have demonstrated the superior decorporation performance of 3,4,3-LI(1,2-HOPO) for multiple elements, including U(VI), Np(V), and Pu(IV), when compared to DTPA and other chelating agents. As a result, 3,4,3-LI(1,2-HOPO) has been approved for Phase 1 frst-inhuman clinical studies, as a decorporation agent for actinides. 18 The aforementioned in vivo studies, however, focused on lanthanides and earlier actinides, and recent spectroscopic and separation studies with Cm, Bk and Cf revealed coordination chemistry differences, 19,26,27 prompting us to question whether such effects could translate to different biodistribution patterns and if 3,4,3-LI(1,2-HOPO) would preserve its decorporation performance amongst heavier radionuclides. Here we present the internal accumulation and excretion profles in mice contaminated with 248 Cm, 249 Bk, 249 Cf and 253 Es. Although the main two deposition regions, skeleton and liver, were the same for the four metals, the total organ retention and evolution over time was element-dependent, with lower atomic number radionuclides showing higher liver and lower skeleton accumulations. Prompt post-contamination decorporation treatments with DTPA and 3,4,3-LI(1,2-HOPO) (Fig. 1) resulted in enhanced radionuclide clearance, with the latter ligand showing better therapeutic performance. Rather unexpectedly, these results also indicate that although transplutonium actinides present common characteristics as a series, their in vivo biodistribution and clearance, in the absence and presence of therapeutic chelators, are discernibly element-dependent. ## Biodistribution and preferential organ deposition The biodistribution of transplutonium elements was tested in young adult female Swiss-Webster mice. 248 Cm, 249 Bk, 249 Cf, and 253 Es were chosen for their availability and because their half-lives (3.4 10 5 years, 330 days, 351 years, and 20.5 days, respectively) were long enough to perform the experiments. The radionuclides were injected intravenously as citrate solutions with fnal administered activities for each actinide ranging between 0.23 and 0.93 kBq per mouse (Table S1 †). These activities were high enough to be traced but low enough to avoid acute radiation effects on the animals. 249 Cf and 249 Bk were injected together due to their availability in our laboratory. Nevertheless, it is highly unlikely they interfered with each other's biodistribution, considering the large excess of endogenous chelators in blood. While 249 Cf and 249 Bk concentrations were in the nanomolar and picomolar range in the bolus solutions (before injection and dilution in the mouse circulatory system), metal-binding proteins, such as hemoglobin, transferrin, and fetuin, are in the micromolar and millimolar range in mouse blood (Table S2 †). It is worth noting that the concentrations of the different species affect the binding equilibrium, and we had to use different injected masses for each actinide due to their distinct activities (Table S1 †). However, the large excess of binding proteins in blood compared to the injected radioisotopes likely minimized any actinide concentration effect in biodistribution by pulling the equilibrium towards the protein-actinide complexes through Le Chatelier's principle. The mice were euthanized at different time points (from 5 min to 24 h post contamination), and their tissues and excreta were collected and radioanalyzed. For decorporation efficacy tests, two mouse groups received a DTPA or 3,4,3-LI(1,2-HOPO) treatment (30 mmol kg 1 per mouse) through intraperitoneal injection 1 h after contamination, an administration time widely used to assess decorporation treatments, 18 and were euthanized 24 h after actinide exposure. All mice showed steady body weight, and no visible or palpable dermal infections, impaired mobility, or ascites were detected during the study, indicating a lack of acute toxicity. Fig. 2 shows the total percentage of actinide recovered dose (% RD) and their distribution in selected organs for each group. In the absence of treatment, the total body retention stabilized around 80% RD (Fig. 2a) for all four actinides after 1 h. Although the total body content remained fairly similar from 1 to 24 h, the organ distribution changed over time. One of the main accumulation regions was the skeleton, which showed distinct deposition behaviours for the four isotopes. 248 Cm accumulation in the bones was steady (between 28 AE 2 and 24 AE 4% RD) over 24 h, while 249 Bk, 249 Cf and 253 Es content increased from around 32% up to 42 AE 3, 50 AE 2, and 49 AE 2% RD, respectively. The different behaviours between the four isotopes were more pronounced in the liver, the organ with the second largest actinide content (Fig. 2c). 248 Cm and 249 Bk showed up to 40 AE 4 and 30 AE 3% RD accumulation, respectively, during the experiments, which contrasted with the lower liver deposition observed for 249 Cf (17 AE 2% RD) and 253 Es (12 AE 1% RD) after 24 h. The higher liver accumulation by earlier transplutonium actinides is consistent with a past study with trivalent 241 Am (0.9 kBq per mouse), which showed 49% RD liver content after 24 h. 18 The experimental protocol used in ref. 18 is the same as that used in our current study. Regarding other organs, kidneys (Fig. 2d) and soft tissues (Fig. S1 †) showed smaller actinide content (<5% RD after 24 h), which decreased over time for all tested radionuclides. The preference of transplutonium elements for skeleton and liver is in agreement with other biodistribution studies performed with lighter actinides. The distinct deposition profles for the different transplutonium actinides are clearly visible in Fig. 3a, which shows a decrease in the actinide liver/skeleton accumulation ratio with increasing atomic number. These results are consistent with liver and skeleton biodistribution trends across the lanthanide series, which are also element-dependent. 28 The deposition trend was more pronounced at 24 h post contamination (Fig. 3b), when the lighter transplutonium actinide studied here, 248 Cm, showed around 7-fold higher liver/skeleton accumulation ratio than 253 Es. Although mammals are not known to use actinides for essential biochemical processes, f-block elements may compete with Ca 2+ , Fe 3+ , Mg 2+ , and Mn 2+ for protein metal-binding sites, 29,30 and the observed differences in organ accumulation ratios likely stem from element discrimination at the molecular level. For instance, one identifed mammalian target of lanthanides and actinides after internal contamination is fetuin, a calcium-binding protein that participates in bone metabolism. Moreover, actinides also show high binding affinity for several proteins that participate in the shuttling of calcium to bone tissue. 34 Even though limited data comparing the interaction between bone-related proteins and trivalent actinides has been published, one study reported Cm 3+ having higher binding affinity for multiple bone glycoproteins than Am 3+ , 34 which may explain the larger bone deposition of trivalent actinides with higher atomic number. Regarding liver accumulation, actinides interact with multiple metal-binding proteins in the liver uptake pathway, including transferrin, ferritin, and calmodulin. 35,36 Transferrin is reported to be the main protein mediating actinide transport from blood to hepatic cells, and once actinides are internalized, they are transferred to other high molecular weight proteins. 35,36 A key step in this acquisition pathway is the interaction between the metal-transferrin complex and the cell surface receptor that mediates endocytosis of the complex. For instance, Pu 4+ shows high affinity for transferrin, but the resulting complexes (Pu 4+transferrin) are poorly recognized by the transferrin receptor, and only one isoform containing one Pu 4+ ion and one Fe 3+ ion is actively recognized and endocytosed. 37 Hence, Pu 4+ is only moderately internalized by hepatic cells compared to other metals. 38 Thus, the direct affinities between the metals and the transferrin may not be as important on defning the actinide liver uptakes, as the interactions between the metal complexes and the transferrin receptor are. There is not a systematic study that explores the binding between trivalent actinide-transferrin complexes and the receptor. However, a study with lanthanides indicated decreasing affinity of the receptor for the metaltransferrin complexes with increasing atomic number (La 3+ $ Nd 3+ > Gd 3+ > Yb 3+ ), 39 which, if trivalent actinides follow the same trend, may explain the larger liver deposition of early transplutonium elements that we observed. It is worth noting that the burden transition from a higher liver accumulation to a higher skeleton accumulation occurs between Cm 3+ and Bk 3+ Full body accounts for all the dose recovered from a mouse but the dose from the excreta. for the actinides (Fig. 3) and between Sm 3+ and Eu 3+ for the lanthanides, according to data compiled and reviewed by Leggett et al. 28 If the biochemical processes involved in the transport of these exogenous metals were exclusively driven by ionic interactions, the lanthanide transition would appear around Pm 3+ (with an ionic radius intermediate between those of Cm 3+ and Bk 3+ ) or the actinide transition would appear around Cf 3+ (with an ionic radius intermediate between those of Sm 3+ and Eu 3+ ). 40 This transition, however, could be explained based on differences on covalency between the lanthanide and actinide series. Over the past decade, a combination of experimental and theoretical studies has evidenced increasing covalent character in bonding across the late actinide series, contrasting with the trivalent lanthanides. In addition, small molecular transplutonium complexes have been shown to display strong energy degeneracy-driven covalency, distinct from the more traditional overlap-driven covalency typical of the early actinides. Those are very subtle differences but they could still affect several features, including bond lengths and electron density around the metal center, metal binding kinetics, or metal complex thermodynamic stability and could potentially explain variations in 4f-or 5f-element binding by different biological molecules, such as bone glycoproteins, small molecular ligands circulating in the blood or common metalloproteins. Lastly, the absorbed dose rates for each radionuclide in the liver and skeleton (the main two deposition regions) were calculated at 24 h to further confrm that the injected doses were low enough to avoid any acute radiation damage. The absorbed dose rates were sub-mGy/s for all actinides (Table S3 †), below the dose rates commonly used in mouse experiments. 47 ## Decorporation treatments against internal contamination Mice that were treated with chelating agents received the treatment 1 h after contamination, and were euthanized 24 h post-metal exposure. Administration of DTPA reduced the total retained 248 Cm, 249 Bk, 249 Cf, and 253 Es to 57 AE 3, 53 AE 6, 47 AE 5, and 48 AE 7% RD, respectively (Fig. 2). The bodily actinide content of the mice treated with 3,4,3-LI(1,2-HOPO) was below 30% RD for all four radionuclides, which corresponded to roughly 3 and 2-fold decreases relative to the control and DTPAtreated groups, respectively. Although actinide deposition in all organs notably decreased after 3,4,3-LI(1,2-HOPO) treatment, the changes in liver activity were the most signifcant ones, with reductions ranging from 6 to 23-fold compared to the control group. This is of particular interest since liver cancer is one of the main disorders associated with internal radionuclide contamination. 48 One of the reasons 3,4,3-LI(1,2-HOPO) significantly prevented actinide accumulation in liver despite the 1 h delayed treatment is the chelator's fast biodistribution into this organ, which occurs within 5 min after injection. 49 ## Excretion proles In addition to the bioaccumulation studies, the mice excreta were also collected and radioanalyzed. Fig. 4a displays cumulative actinide excretion over time in the absence of treatment. The four metals had similar elimination profles, where the largest portion of the radionuclides was excreted within the frst 30 min post-contamination. For the treatment groups (Fig. 4b), total excretion after 24 h was signifcantly higher than for the control group and inversely proportional to body accumulation. Treatment with DTPA promoted $ 50% RD clearance, which occurred primarily through the urinary pathway. 3,4,3-LI(1,2-HOPO), on the other hand, showed superior excretion rates with over 70% RD for all four actinides, and a signifcant increased fecal elimination component. The different excretion routes for DTPA and 3,4,3-LI(1,2-HOPO) complexes have been previously observed and explained based on the metal complex physico-chemical properties, including lipophilicity, solubility and ionization constants. 24 Noteworthy, we observed signifcant differences and opposite trends in ligand-promoted excretion as a function of the metal, following the order 248 Cm < 249 Bk < 249 Cf $ 253 Es for DTPA and 253 Es < 249 Cf $ 249 Bk < 248 Cm for 3,4,3-LI(1,2-HOPO). The trend seen with DTPA correlates with increasing stability constants for complexes formed from Cm to Es (Table S4 †), with higher metal affinities corresponding to higher excretion and decorporation power. In contrast, the effect seen with 3,4,3-LI(1,2-HOPO) is reversed from those observed for DTPA, and could be traced back to the decreasing amount of actinide initially deposited in the liver, when progressing from 248 Cm to 253 Es, as the liver is a direct pool for decorporation by 3,4,3-LI(1,2-HOPO), which is not the case for DTPA. 241 Am data was obtained from ref. 18, which followed the same experimental protocol as that used in this study. The injected dose of 241 Am was 0.9 kBq per mouse. ## Systematic comparisons of actinide biodistribution proles To further provide a thorough perspective of the different actinide bioaccumulation profles as well as their respective variations after chelator administration, we systematically compared deposition ratios of the tested radionuclides log % RD An 1 % RD An 2 in each organ and excreta tested, for the control, HOPO treatment, and DTPA treatment groups, 1 h post-contamination (Fig. 5). In this analysis, similar data previously collected and reported on 241 Am were included. 18 As mentioned above, lighter transplutonium actinides tended to accumulate to larger extent in the liver compared to heavier ones, which showed higher deposition in the skeleton. This trend is mostly reproduced, albeit sometimes attenuated, after administration of a chelating agent, even though a few features are worth noting. Although DTPA promoted the excretion of actinides primarily through urine, its administration yielded a large increase of 241 Am expulsion through the feces compared to all heavier radionuclides studied. This is best exemplifed by the brighter green "Feces" column on the top centre panel of Fig. 5, as well as the brighter red upper right corners of each of the remaining centre panels. In contrast, administration of 3,4,3-LI(1,2-HOPO) seems to accentuate the differences in excretion between 241 Am and 248 Cm, two actinides notoriously difficult to separate: while the urinary (fecal) output of 241 Am is smaller (larger, respectively) than that of 248 Cm in control contaminated animals, that difference is much increased after one 3,4,3-LI(1,2-HOPO) chelation treatment, as depicted by the much brighter colours in the upper right corner of the top right panel of Fig. 5, compared with the top left panel. This 3,4,3-LI(1,2-HOPO)induced excretion change was also concomitant with a reversed relative liver retention pattern. Thus, even though 3,4,3-LI(1,2-HOPO) showed the highest therapeutic performance as decorporating agent, both ligands had substantial effects on changing the relative biodistributions of the tested actinides, which could be of particular interest if one envisioned to exploit these different distribution ratios for purposed element separations in bio-inspired artifcial systems. Mostly motivated by the need to devise new efficient strategies for the mining and purifcation of rare earth metals as critical materials 50 or for the large-scale extraction of these elements from diluted environments such as seawater, 51 many biological and biologically-inspired molecules have recently emerged as potentially promising systems for f-element separations. Current state-of-the-art in this feld runs the gamut from biopolymers, organic acids, as well as small molecule metallophores, peptides and proteins produced by microbial species. 52 However, mammalian systems have not yet been explored in that context. The results presented here indicate that the molecular mechanisms involved in the mammalian transport and storage of actinide contaminants could eventually be decrypted, improved, and utilized to discriminate f-block metals. One may even envision the engineering of new bioreactors, such as modifed spheroid reservoir bioartifcial livers, 53 that would leverage these mechanisms. ## Conclusions In summary, we studied the biodistribution and excretion of the heaviest transplutonium elements available, namely 248 Cm, 249 Bk, 249 Cf and 253 Es, in a live mouse model. Their deposition in organs was observed within the frst fve minutes after contamination, and skeleton and liver were the main accumulation regions. Despite their similar charge and ionic radius, the distribution of radionuclides was element dependent, where lower atomic number transplutonium elements showed up to 7fold higher liver/skeleton deposition ratios compared to the heavier tested actinides. Treatment with both DTPA and 3,4,3-LI(1,2-HOPO) 1 h after contamination promoted higher radioisotope excretion, with the latter showing better performance (around 3-fold higher clearance compared to control). Although mice do not use actinides in their metabolism, transplutonium element interaction with endogenous chelators seems to be element dependent, resulting in different in vivo behaviour. Such discernible differences in a complex multi-component living system may provide future insights for developing new bio-inspired strategies for efficient sequestration and separation of transplutonium elements.

chemsum

{"title": "Efficient discrimination of transplutonium actinides by <i>in vivo</i> models", "journal": "Royal Society of Chemistry (RSC)"}

design,_synthesis_and_biological_activities_of_echinopsine_derivatives_containing_acylhydrazone_moie

## Abstract: Based on the broad-spectrum biological activities of echinopsine and acylhydrazones, a series of echinopsine derivatives containing acylhydrazone moieties have been designed, synthesized and their biological activities were evaluated for the first time. The bioassay results indicated that most of the compounds showed moderate to good antiviral activities against tobacco mosaic virus (TMV), among which echinopsine (I) (inactivation activity, 49.5 ± 4.4%; curative activity, 46.1 ± 1.5%; protection activity, 42.6 ± 2.3%) and its derivatives 1 (inactivation activity, 44.9 ± 4.6%; curative activity, 39.8 ± 2.6%; protection activity, 47.3 ± 4.3%), 3 (inactivation activity, 47.9 ± 0.9%; curative activity, 43.7 ± 3.1%; protection activity, 44.6 ± 3.3%), 7 (inactivation activity, 46.2 ± 1.6%; curative activity, 45.0 ± 3.7%; protection activity, 41.7 ± 0.9%) showed higher anti-TMV activity in vivo at 500 mg/L than commercial ribavirin (inactivation activity, 38.9 ± 1.4%; curative activity, 39.2 ± 1.8%; protection activity, 36.4 ± 3.4%). Some compounds exhibited insecticidal activities against Plutella xylostella, Mythimna separate and Spodoptera frugiperda. Especially, compounds 7 and 27 displayed excellent insecticidal activities against Plutella xylostell (mortality 67 ± 6% and 53 ± 6%) even at 0.1 mg/L. Additionally, most echinopsine derivatives exhibited high fungicidal activities against Physalospora piricola and Sclerotinia sclerotiorum. Plant virus diseases can be caused by more than 900 viruses, which reduce grain production and lead to huge economic losses all over the world . As a well-studied plant virus, tobacco mosaic virus (TMV) belongs to single-stranded RNA virus of the family togaviridae 4 and it can infect 268 species of plants in 38 families, such as tobacco, tomato, pepper, cucumber, causing their leaves to grow spots, wither and even leading to yield reduction . Although commercially available plant virus inhibitors ningnanmycin and ribavirin are widely used to control TMV, their inhibitory effects are lower than 60% 8 . Thus, the development of efficient alternative TMV inhibitors is still in great request. Natural products are an important source of plant virus inhibitor discovery. Compared with traditional synthetic plant virus inhibitor, plant virus inhibitor derived from natural products have many advantages, including low toxic, environmentally friendly, easy to decompose and specific to target species, etc 9,10 . Song et al. reported that the EC 50 value of purine nucleoside derivative for the inactivating activity against TMV was 48 mg/L, which was better than that of ningnanmycin (88 mg/L) 11 . Li et al. first found that phenanthroindolizidine alkaloid, (R)-antofine, exhibited a good inhibitory effect against TMV 12 . Wang et al. found some β-carboline analogues 7 , hemigossypol 13 , dehydrobufotenine derivatives 14 , pityriacitrin marine alkaloids 15 , pulmonarin alkaloids 16 and hamacanthin derivatives 17 exhibited higher anti-TMV activities than ningnanmycin. Many other natural alkaloids derivatives were also developed as potential TMV inhibitors . Although a variety of natural product derivatives have been found to exhibit high anti-TMV activity, few of them have been applied successfully in agriculture. Thus, it is necessary to discover novel natural TMV inhibitors with diverse structures. Echinopsine is a quinoline alkaloid isolated from Echinops sphaerocephalus L., the root of which was used as traditional Chinese medicine for treatment of deep-rooted breast carbuncles, ulcer, sodoku and breast milk stoppage. Although the bioactivity of Echinops sphaerocephalus L. extract has been widely studied 28 , the biological activity of echinopsine is still not clear. The anti-TMV activity of echinopsine has not been reported so far. However, a variety of natural alkaloids containing echinopsine moiety showed herbicidal, insecticidal, bactericidal, Biological assay. The anti-TMV, insecticidal and fungicidal activities of the synthesized compounds were tested using our previously reported methods 38,39 and the methods can also be found in the "Supporting Information SI". ## General synthesis. Ribavirin (Topscience Co., Ltd.), chlorothalonil (Bailing Agrochemical Co., Ltd.), carbendazim (Bailing Agrochemical Co., Ltd.) and other reagents were purchased from commercial sources and used as received. All anhydrous solvents were dried and purified according to standard techniques. The synthetic routes were given in Fig. 3. Echinopsine was prepared according to literature 40 . General procedure for the preparation of compounds 1-27. To a round bottomed flask (100 mL) were added methanol (50 mL), compound C (3 mmol), one benzaldehyde from D 1 -D 27 (3 mmol) and p-methylbenzene sulfonic acid (0.6 mmol). The reaction suspension was refluxed for 8 h. The reaction suspension was cooled to room temperature and partial methanol was evaporated under reduced pressure until a large amount of precipitation precipitated. The precipitate was filtered and washed several times with cool methanol to afford compounds 1-27. Data for compounds 1-27 can be found in the "Supporting Information SI". was used as solvent instead of DMF and the reaction was accomplished in 91.2% yield. Then product B reacted with hydrazine hydrate under reflux to afford hydrazine C, which can react subsequently with aldehyde D 1 -D 27 to give hydrazine 1-27 as products in 52.7-95.3% yields. During the synthesis of acylhydrazone 1-27, only trans isomers were obtained, which may due to the fact that trans isomers are more stable than cis isomers thermodynamically. Compounds 1-27 can precipitate from methanol, which made the purification of acylhydrazone derivatives easy and suitable for large-scale production. In vivo anti-TMV activity. The results of anti-TMV activities in vivo (inactivation, curative, and protection mode) of echinopsine and compounds 1-27 are listed in Table 1. In order to make the antiviral activity results more reliable, commercial plant virus inhibitor ribavirin was taken as control. In our previous work, the highly antiviral lead echinopsine was found, based on which a series of echinopsine derivatives containing acylhydrazone structure were synthesized in this work to study the influence of the variation of the functional groups on the antiviral activities of echinopsine. The antiviral results (Table 1) showed that some echinopsine acylhydrazone compounds exhibited moderate to good anti-TMV activity compared with ribavirin. Especially, the inactivation activity, curative activity, protection activity of compounds 1 (44.9 ± 4.6, 39.8 ± 2.6 and 47.3 ± 4.3%, 500 mg/L), 3 (47.9 ± 0.9, 43.7 ± 3.1, and 44.6 ± 3.3%, 500 mg/L), 7 (46.2 ± 1.6, 45.0 ± 3.7, and 41.7 ± 0.9%, 500 mg/L) were obviouly higher than that of commercialized anti-plant virus agent ribavirin (38.9 ± 1.4, 39.2 ± 1.8, and 36.4 ± 3.4%, 500 mg/L). For derivatives containing substituted phenyl (1-14), the electronic effect of the substituents on phenyl has an effect on the anti-TMV activities. The introduction of electron-withdrawing and electron-donating substituents led to the decrease of anti-TMV activities. For example, the structure-activity relationship shows the following: non-substituent (1) > p-hydroxyl (4) > p-phenoxy (8) > p-methylthio (9) > p-methoxy (5), non-substituent (1) > p-bromo substituent (13) > p-methylsulfonyl (10) > p-fluorosubstituent (11) > p-chloro substituent (12). However, there is no obvious linear relationship between anti-TMV activity and electron-donating and electronwithdrawing ability. For example, the structure-activity relationship shows the following: p-bromo substituent (13) > p-trifluoromethoxy substituent (14) > p-fluoro substituent (11) > p-chloro substituent (12), while the activity of compound 13 at 500 mg/L (inactivation activity, 42.9 ± 4.4%; curative activity, 31.1 ± 2.8%; protection activity, 35.8 ± 3.0%) is equivalent to that of ribavirin. The size of substituents also has an effect on the activities. For example, the activities of derivatives with a p-tert-butyl (3) and p-phenyl substituent (7) are higher than that with no substituents (1). Mono substitution or multi substitution on the benzene ring affected anti-TMV activity to a certain extent, for instance, compared with compounds 5 (inactivation, 20.6 ± 2.6%, 500 mg/L), the disubstituted compound 6 (inactivation, 32.3 ± 1.7%, 500 mg/L) exhibited higher activity. The anti-TMV activities of compounds 15-26 containing heterocyclic ring reduced obviously compared with that of compounds containing benzene ring (1). Compound 22, showed the highest activities at 500 mg/L www.nature.com/scientificreports/ (inactivation activity, 40.5 ± 3.5%; curative activity, 34.7 ± 4.0%; protection activity, 38.3 ± 4.0%), which was equivalent to that of ribavirin. However, the activity was greatly reduced when the benzene ring was changed to an anthracene ring, that is, the activities of compound 27 (inactivation, 38.9 ± 2.5%, 500 mg/L) was lower than that of compound 1 (inactivation, 44.9 ± 4.6%, 500 mg/L). Compound 3 showed the highest activities at 500 mg/L (inactivation activity, 47.9 ± 0.9%; curative activity, 43.7 ± 3.1%; protection activity, 44.6 ± 3.3%), which is significantly higher than that of ribavirin. Thus, this compound (3) can be selected as an anti-TMV candidate drug for further study. ## Insecticidal activities. The insecticidal activities of the target compounds 1-27 and echinopsine against Lepidoptera pests, such as diamondback moth (Plutella xylostella), cotton bollworm (Helicoverpa armigera), corn borer (Ostrinia nubilalis), oriental armyworm (Mythimna separata) and fall armyworm (Spodoptera frugiperda (J. E. Smith)) are listed in Tables 2 and 3, echinopsine was taken as control. The result showed that echinopsine and some derivatives showed broad spectrum insecticidal activities. Most of the compounds exhibited moderate to good larvicidal activities against P. xylostella. For derivatives containing substituted phenyl (1-14) and anthranyl (27), compounds 7, 14 and 27 exhibited 100 ± 0% mortality at 600 mg/L. In particular, compounds 7 and 27 still showed 67 ± 6% and 53 ± 6% mortality even at 0.1 mg/L. Compounds 15, 21, 23, 25 and 26 containing heterocyclic ring also showed 100 ± 0% mortality at 600 mg/L, which was better than echinopsine (90 ± 0% at 600 mg/L) (Table 2). At the same time, the insecticidal activities of compounds 15-26 containing heterocyclic ring against M. separata and S. frugiperda were higher than that of compounds 1-14 containing benzene ring. The compounds 5, 9, 14, 21, 24 and 25 exhibited higher activities (100 ± 0% at 200 mg/L) against M. separata than that of Fungicidal activity. The fungicidal results of compounds 1-27 and echinopsine are listed in Table 4. The commercial fungicide carbendazim and chlorothalonil were used as positive control. Overall, echinopsine and their derivatives exhibited broad-spectrum fungicidal activities against 14 kinds of phytopathogenic fungi. Most compounds showed relatively high fungicidal activities for Physalospora piricola and Sclerotinia sclerotiorum, among which the fungicidal activities of compounds 1-14 containing substituted phenyl were relatively higher than compounds 15-26 containing heterocyclic rings. Compound 13 and 14 showed more than 50% inhibitory rate against five and six fungi respectively. Compound 2 showed the widest spectrum of fungicidal activity, with more than 60% inhibitory rate against eight fungi. Compound 7 exhibits 89.0 ± 1.9% inhibitory rate against Rhizoctonia cerealis at 50 mg/L, higher than carbendazim and chlorothalonil. In summary, a series of novel echinopsine derivatives containing acylhydrazone moieties were designed, synthesized and their antiviral, insecticidal, and fungicidal activities were studied. The bioassays results showed that most compounds exhibited moderate to good anti-TMV activities in vivo, among which echinopsine (I) and its derivatives 1, 3, 7 showed higher anti-TMV activities than those of ribavirin, which can be used as lead structures for the development of anti-TMV drugs. Some compounds exhibited moderate to good insecticidal activity to P. xylostella, M. separata and S. frugiperda. In addition, most of these compounds exhibited good fungicidal activities against P. piricola and S. sclerotiorum. Further investigation on structural optimization and the mechanism of action are in progress in our laboratory.

chemsum

{"title": "Design, synthesis and biological activities of echinopsine derivatives containing acylhydrazone moiety", "journal": "Scientific Reports - Nature"}

novel_(2-amino-4-arylimidazolyl)propanoic_acids_and_pyrrolo[1,2-<i>c</i>]imidazoles_via_the_domino_r

## Abstract: The unexpectedly uncatalyzed reaction between 2-amino-4-arylimidazoles, aromatic aldehydes and Meldrum's acid has selectively led to the corresponding Knoevenagel-Michael adducts containing a free amino group in the imidazole fragment. The adducts derived from Meldrum's acid have been smoothly converted into 1,7-diaryl-3-amino-6,7-dihydro-5H-pyrrolo[1,2-c]imidazol-5ones and 3-(2-amino-4-aryl-1H-imidazol-5-yl)-3-arylpropanoic acids. The interaction of 2-amino-4-arylimidazoles with aromatic aldehydes or isatins and acyclic methylene active compounds has led to the formation of pyrrolo[1,2-c]imidazole-6-carbonitriles, pyrrolo[1,2-с]imidazole-6-carboxylates and spiro [indoline-3,7'-pyrrolo[1,2-c]imidazoles], which can be considered as the analogues of both 3,3'-spirooxindole and 2-aminoimidazole marine sponge alkaloids. ## Introduction Heterocyclic compounds of both natural and synthetic origin, containing in their structure pyrrole and imidazole rings, display a wide set of pharmacologically significant activities. The most important natural sources of such systems are marine sponges. Since the 70's of 20th century up to date more than 150 derivatives containing pyrrole and 2-aminoimidazole fragments in their structure were found among the metabolites of these marine organisms . This group of compounds is characterized by an exceptional molecular diversity. The main structural types of these substances are shown in Figure 1. The metabolites of Leucetta Sp. and Clathrina Sp. are presented by achiral imidazole alkaloids from the group of benzyl substituted 2-aminoimidazole (dorimidazole A (I), naamine A (II)), fused cyclic systems (2-amino-2-deoxykealiiquinone (III)) and spirolinked compounds ((−)-spirocalcaridine B (IV)) . Agelas Sp. are a source of alkaloids with core structures containing simultaneously pyrrole carboxamide and 2-aminoimidazole moieties such as the simple achiral compound oroidine (V) and spatially organized molecules in a complex manner with a large number of chiral centres like (−)-palau'amine (VI) . Oroidine (V) and other related vinyl 2-aminoimidazoles of this class are monomeric precursors of nagelamide A (VII), mauritiamine (VIII), sceptrin (IX), benzosceptrin A (X), axinellamines (XI) and stylissazole A (XII) alkaloids . Fused 2-aminoimidazole and azepinone derivatives XIII were isolated recently from an extract of Pseudoceratina Sp. . The variety of types of pharmacological activity revealed in these marine sponges' metabolites is not inferior to the chemodiversity of their structure. Many of them are reported to have properties such as α-adrenoreceptors and leukotriene B4 receptor antagonists , cyclin-dependent kinases GSK-3β, CK1 and nitric oxide synthase activity inhibitors , as well as antibacterial , antifungal , antihistamine and antitumor activities . Remarkable immunosuppressive properties are inherent to palau'amine (VI) . Ceratamines XIII are the disruptors of microtubule dynamics, therefore are of great interest in cancer drug discovery . Thereby, the stereocontrolled total synthesis of marine alkaloids such as axinellamines and the search of new 2-aminoimidazole and pyrrole containing compounds with a core structure that mimics metabolites of marine sponges with interesting biological properties has received considerable attention from both chemists and pharmacologists. In the middle of 2000s, the authors of the studies proposed a facile one-pot two-step procedure for the synthesis of diversely substituted 2-aminoimidazoles from α-bromocarbonyl compounds and substituted 2-aminopyrimidines. This methodology allowed the rapid synthesis of alkaloids of the isonaamine series and other polysubstituted 2-aminoimidazoles with moderate cytostatic activity and biofilm inhibitory activity against S. Typhimurium and P. Aeruginosa . We have used 4-aryl-substituted 2-aminoimidazoles described by the authors of the aforementioned works as polyfunctional building blocks for the formation of different fused and spirolinked heterocyclic systems. Last ones are able to act as precursors in the synthesis of the substances that mimic the core structure of marine alkaloids due to the presence of several reaction centres, which allow their further chemical modification. In the present work we disclose our results on the multicomponent reactions between 2-amino-4-arylimidazoles, aromatic aldehydes or isatins and cyclic or acyclic CH acids. As the last compounds we have used Meldrum's acid, malononitrile and ethyl 2-cyanoacetate. ## Results and Discussion In view of the structure of 2-amino-4-arylimidazoles containing four nonequivalent nucleophilic centres several pathways can be assumed for their reactions with carbonyl 1,3-bielectrophiles or their synthetic precursors in the case of three-component reactions between these amines, carbonyl compounds and CH acids. Previously, an unusual direction of the three-component reaction between 2-aminoimidazoles, aldehydes and 5,5dimethyl-1,3-cyclohexanedione has led to the formation of the Knoevenagel-Michael adducts (Figure 2) . By analogy with our results obtained with the use of other aminoazoles in the reactions with benzaldehydes and Meldrum's acid we expected the formation of one or several isomers of tetrahydroimidazopyrimidinone derivatives (Figure 2). ## Figure 2: The Knoevenagel-Michael adduct and expected products. However, a short time (3-5 min) and reflux of the equimolar amounts of amines 1, para-substituted benzaldehydes 2, and Meldrum's acid 3 in 2-propanol led to Knoevenagel-Michael adducts 4a-h (Table 1). Beside the short reaction times and mild conditions, this catalyst-free three-component condensation is characterized by a very facile performance since the solid products are formed as precipitates and are simply isolated in good yields without any additional purification (Table 1). In our synthetic practice this is the first example of the existence of stable β-adducts, which simultaneously contain Meldrum's acid and aminoazole fragments. In all earlier described experiments with participation of different α-aminoazoles as binucleophiles the reaction cascade readily accomplished by the formation of fused heterocyclic systems . An analogous three-component reaction involving indole or imidazo [1,2-a]pyridine derivatives instead of 2-aminoimidazoles is referred in the literature as Yonemitsu reaction or Yonemitsu-like reaction . The similar Michael-type adducts 6 were isolated from the reaction of imidazo[1,2-a]pyridine with aldehydes and Meldrum's acid in acetonitrile in the presence of a catalytic amount of proline (Scheme 1) and then they were successfully converted to the appropriate esters 7 and acids 8. In our case, we have isolated products 4a-h individually and characterized them by IR, 1 H, 13 C NMR, and mass-spectral methods. The 1 H NMR spectra of products 4 have two characteristic broad singlets that represent the exchangeable proton shifts of the crossed signals of NH and OH groups at a The isolated yields accounted on the quantities of the starting materials 1-3. ## Scheme 1: The three component condensation of imidazo[1,2-a]pyridine, aldehydes and Meldrum's acid described by Gerencsér at al. . 12.35-11.61 ppm and the NH 2 group of the aminoimidazole fragment at 7.47-7.26 ppm, as well as a singlet for the protons of two methyl groups. The existence of the dioxanedione cycle in enol form is proven by the presence of the singlet of a methyne proton near the saturated carbon atom at 5.42-5.56 ppm and the absence of the signal for the methyne proton of the dioxanedione cycle. With regard to the mass spectra, all compounds 4 exhibit similar behaviour in their fragmentation, showing the absence of the molecular ion peak and the presence of intense signals that occur due to cleavage of acetone and СО 2 molecules from the dioxanedione moieties. Their further transformation took place in the presence of the catalytic amounts of TFA in toluene under short time reflux (3 min) or addition of the catalytic amounts of TFA to the initial three-component mixture (Table 2). The reaction proceeded via 1,3-dioxanedione cycle cleavage followed by elimination of acetone and CO 2 to provide novel pyrrolo[1,2-c]imadazol-5-ones trifluoroacetates 9b,c,i precipitated from the reaction mixture (Table 2). The corresponding bicyclic amine 10a was obtained by the prolonged treatment of the product 4a with catalytic amounts of TFA in acetonitrile followed by the addition of an aqueous solution of NH 3 . The structures of cyclized products 9 and 10 were confirmed by spectral methods. The signals of NH, OH and methyl groups of the dioxanedione cycle are absent in the 1 H NMR spectra of trifluoroacetates 9. The broad signal of the NH 2 group shifts to the downfield signal of NH 3 + at 8.5-8.2 ppm. Protons of the СН 2 -СН fragment in the pyrrolidine cycle show the shifts of an ABX system for СН Х at 4.73-4.85, СН В at 3.81-3.84, СН А at 2.86-2.95 ppm. The same situation is observed for compound 10a, however, the signal of the free NH 2 group of the aminoimidazole moiety shifts to 6.36 ppm. The common feature of the mass spectra of salts 9 is the absence of the salt molecular ion peak and the presence of the intense signals that occur due to cleavage of the CF 3 COO − anion. Single crystal X-ray diffraction analysis of biphenyl compound 9i has finally proved the structures of the obtained products (Figure 3). Compound 9i exists as organic salt with trifluoroacetic acid in the crystal phase. The existence of the trifluoroacetic molecule as anion is confirmed by close values of the C-O bond lengths (1.229(2) and 1.238(2) , respectively) and the absence of the hydrogen atom at the carboxylic group. The analysis of the bond lengths in the imidazole ring has revealed that the C1-N1 and C1-N3 bonds are equal (1.320(3) and 1.320(2) , respectively) and the N1-C6 bond (1.414(6) ) is slightly elongated as compare to its mean value 1.376 . The hydrogen atoms at the N1 and N3 were located from the electron density difference maps. As a result we may describe the structure of the organic cation as superposition of two forms (Scheme 2). The prolonged reflux (10 h) of compounds 4b-g in acetonitrile in the presence of a catalytic amount of TFA leads to the opening of the pyrrolidone ring followed by the formation of acids 11b-g (Table 2). The process remarkably accelerates while adding water to the reaction mixture. The acids 11 can also be obtained from pyrrolo[1,2-c]imidazol-3-aminium trifluoroacetates 9 after prolonged reflux (12 h) in aqueous acetonitrile. The 1 H NMR spectra of acids 11 contain the signals of the protons of the aromatic system, the broad singlet for NH 2 group at 5.87-5.82 ppm and the signals of the ABX protons of the propionyl fragment -СН Х at 4.60-4.30 ppm and СН2 АВ at 3.00-2.60 ppm. Finally, the structure of acids 11 was confirmed by X-ray diffraction data of the sample compound 11b (Figure 4). Compound 11b was found to be a zwitterion and exists as monohydrate in the crystal phase. The absence of the hydrogen atom and equalization of the C6-O1 and C6-O2 bond lengths (1.254(2) and 1.259(2) , respectively) allow presuming the location of the negative charge at the deprotonated carboxylic group. The very close lengths of the bonds centred at the C1 atom (the N2-C1 bond length is 1.332(2) , the C1-N3 bond length is 1.337(3) and the N1-C1 bond length is 1.340(2) ) allows to describe the zwitterion as superposition of three forms with different location of the positive charge (Scheme 3). Literatur data concerning pyrrolo[1,2-c]imidazol-5-ones is quite limited and the known 6,7-dihydro analogs are represented only by several substances . Partially hydrogenated pyrrolo[1,2-c]imidazole is a part of (±)-axinellamines 11. 4-[(5R)-6,7-Dihydro-5H-pyrrolo[1,2-c]imidazol-5-yl]-3-fluorobenzonitrile (LCI-699, osilodrostat) is considered as an inhibitor of aldosterone synthase (CYP11B2) and 11β-hydroxylase (CYP11B1), which is responsible for cortisol production . This compound is under development for the treatment of Cushing's syndrome and pituitary ACTH hypersecretion . From this point of view the approach to pyrrolo[1,2-c]imidazole moiety by using acyclic methylene active compounds, that can lead to cyclic products, has a high potential for diversityoriented synthesis. In the three-component condensations of equimolar amounts of 2-amino-4-arylimidazoles 1, para-substituted benzaldehydes 2 and malononitrile (12) in 2-propanol the Knoevenagel-Michael adduct was not obtained. The reaction was complete to form a mixture of pyrrolo[1,2-c]imidazol-6-carbonitriles 13 and their azomethine derivatives 14 (Scheme 4). The use of a double excess of aromatic aldehydes 2 in this condensation prevented the formation of a mixture of substances and led to the formation of individual 5-amino-3-(arylidenamino)-1-aryl-7-aryl-7H-pyrrolo[1,2-c]imidazole-6-carbonitriles 14, as well as azomethines 16 in case of using ethyl 2-cyanoacetate 15 as the acyclic methylene active compound (Table 3). The isolated products 14a-f and 16a,b were characterized by IR, 1 H, 13 C NMR and mass-spectral methods. The mass spectra of compounds 14 and 16 show the similar type of fragmentation. They contain peaks of molecular ions, as well as signals corresponding to the loss of fragments From the comparison of these data with the results of elemental analysis, it follows that in the formation of condensed systems 14 with the participation of two molecules of aromatic aldehydes two molecules of water were cleaved. In the IR spectra, the most characteristic bands represent the absorption of NH 2 groups at 3420 and 3332 cm −1 and the nitryl group CN at 2250 cm −1 . In addition, there are characteristic bands at 1664-1668 cm Finally, the structure of azomethines 16 was confirmed by X-ray diffraction data of the sample compound 16a (Figure 5). All atoms of the bicyclic fragment lie in the plane within 0.01 . The analysis of the bond lengths has shown that the formally single exocyclic C1-N3 bond is shorter than the double endocyclic C6-C1 bond (1.336(6) and 1.354(9) , respectively).The C1 and C6 atoms are planar indicating their sp 2 hybridization. Such a distribution of electron density allows to discuss the zwitter-ionic form and to consider the structure of 16a as superposition of two resonance structures (Scheme 5). In the next step of our research we have involved isatin 18 as the compound bearing a carbonyl group, as well in this case the pyrolo[1,2-c]imidazole moiety will be spiro-fused with the oxindole moiety, and the resulting structures can be considered as analogues of 3,3'-spiroxindole alkaloids, such as spirotryprostatin B (17, Figure 6) . 4). The reduced reactivity of the carbonyl group of isatins compared with benzaldehydes, and the greater stability of their Knoevenagel adducts leads to the formation of individual spiro compounds, not to a mixture of substances. However, condensations with the use of N-unsubstituted isatins are accompanied by the resinification of the reaction mixture, which may be caused by competing reactions of heterocyclization of the mentioned Knoevenagel adducts. In similar reactions described in the literature , the authors recognized the importance of protecting the amide fragment of isatin, since it affects the reactivity and, in some cases, the enantioselectivity of processes. In order to prevent undesirable side reactions in the future, three-component condensations were carried out using N-substituted isatins. The isolated products 19a-h and 20a-c were characterized by IR, 1 H, 13 C NMR and mass-spectral methods. The Finally, the structures of spiroxindoles 19 and 20 were confirmed by X-ray diffraction data of the sample compound 19a (Figure 7). Compound 19a exists in the crystal phase as solvate with dimethylformamide and water in a 1:1:1 ratio. The spiro-joined bicyclic fragments are turned relatively to each other in such a way that the dihedral angle between mean planes of the bicycles is 84.5°. The analysis of the bond lengths has shown that the formally single exocyclic C6-N6 bond is significantly shorter than the double endocyclic C6-C5 bond (1.319(2) and 1.373(3) , respectively). The C1 and C6 atoms are planar indicating their sp 2 hybridization. Such a distribution of electron density allows discussing the zwitter-ionic form and considering the structure of 19a as superposition of two resonance structures similar to 16a (Scheme 6). ## Conclusion In the described three-component reactions with aldehydes or isatins and cyclic or acyclic CH acids the C 5 reaction centre in the 2-amino-4-arylimidazoles possesses higher nucleophilicity than both the exo-and endocyclic amino groups. Regarding the short reaction times of novel Yonemitsu-type reactions that has been achieved without application of any catalyst we assume that 2-amino-4-arylimidazoles are more reactive substrates for these syntheses leading to the stable Michael-type adducts with aldehydes and Meldrum's acid than the previously investigated indole and imidazo[1,2-a]pyridine. Moreover, as it has been shown that their further transformations may result in the formation of both unexplored heterocyclic systems containing a free amino group open for chemical modifications and the corresponding hetarylpropanoic acids providing useful templates for the synthesis of some marine alkaloids or their analogues. In domino reactions of the 2-amino-4-arylimidazoles with isatins and aliphatic CH acids stable Michael adducts have not been fixed. Cyclocondensation has readily led to the formation of 6'-substituted 3',5'-diamino-1-alkyl-2-oxo-1'-arylspiro[indolin-3,7'-pyrrolo [1,2-c]imidazoles], which can be considered as the analogues of alkaloids with both pyrrolo[1,2-c]imidazol and 3,3'-spiroxindole fragments in the core structure. ## Experimental Reagents and analytics: Starting materials were purchased from commercial suppliers. Melting points were determined on a Kofler apparatus and temperatures were not corrected. The IR spectra were recorded in KBr on a Specord M-82 spectrometer.

chemsum

{"title": "Novel (2-amino-4-arylimidazolyl)propanoic acids and pyrrolo[1,2-<i>c</i>]imidazoles via the domino reactions of 2-amino-4-arylimidazoles with carbonyl and methylene active compounds", "journal": "Beilstein"}

rapid_synthesis_of_layered_k_<i>x</i>mno₂_cathodes_from_metal–organic_framework

## Abstract: Layered transition metal oxides (LTMOs) are a kind of promising cathode materials for potassium-ion batteries because of their abundant raw materials and high theoretical capacities. However, their synthesis always involves long time calcination at a high temperature, leading to low synthesis efficiency and high energy consumption. Herein, an ultra-fast synthesis strategy of Mn-based LTMOs in minutes is developed directly from alkali-transition metal based-metal-organic frameworks (MOFs). The phase transformation from the MOF to LTMO is systematically investigated by thermogravimetric analysis, variable temperature optical microscopy and X-ray diffraction, and the results reveal that the uniform distribution of K and Mn ions in MOFs promotes fast phase transformation. As a cathode in potassiumion batteries, the fast-synthesized Mn-based LTMO demonstrates an excellent electrochemical performance with 119 mA h g À1 and good cycling stability, highlighting the high production efficiency of LTMOs for future large-scale manufacturing and application of potassium-ion batteries. ## Introduction Rechargeable potassium-ion batteries (KIBs) are considered as one of the most promising alternatives to lithium-ion batteries on account of the abundant raw materials and low redox potential of potassium for large-scale energy storage technologies. 1-3 However, there exist signifcant challenges in seeking suitable electrode materials to achieve sufficient electrochemical performance of KIBs because of the large size of K + , which causes huge volume changes and sluggish kinetics, and thus low specifc capacity, poor rate capability and fast performance degradation. In the past few years, numerous efforts have been devoted to the development and optimization of lowcost and high-performance electrode materials, particularly cathode materials. Recently, layered transition metal oxides (LTMOs) have been extensively investigated as cathode materials for KIBs. They have ionic diffusion paths with a large interlayer distance, benefting the migration of K + . Among them, layered potassium manganese oxide (K x MnO 2 ) has attracted great attention owing to their high theoretical capacities and abundance of raw materials. Generally, layered K x MnO 2 is synthesized through a long-time high-temperature calcination ($800 C), which typically takes more than 10 hours (Table S1 †). Apparently, this procedure takes time and consumes energy, leading to low production efficiency and high cost. It is well known that the atomic distance seriously impacts the diffusion process during pyrolysis treatment, and further influences the formation of a new phase. Aiming at rapid synthesis of LTMO, a short distance between alkali metal ions and transition metal ions as well as a uniform distribution is necessary. Metal-organic frameworks (MOFs) are a kind of porous crystalline material formed by metal ions and organic ligands through coordination bonds, which have a homogenous distribution of different elements at the atomic level. 26 With the merits of large surface area, ultra-high porosity, and structural diversity, they have attracted extensive attention in energy felds. 27,28 The investigation was mainly focused on MOF-derived anode materials but rarely on cathodes. A lot of efforts have been devoted to developing transition metal (Fe, Co, Ni, Mn, Cu, and Zn)-based MOF precursors for anode materials, while less on alkali-transition metal based-MOFs. Despite differences in the properties between alkali and transition metal ions, alkalitransition metal based-MOFs can be constructed and are suitable for rapid synthesis of LTMO cathodes due to their atomic level distribution of different elements. Herein, an ultra-fast synthesis method is developed for preparing a layered cathode material of K x MnO 2 directly from the K[Mn(HCOO) 3 ] MOF (KM-MOF). The formate anion (HCOO ) is the simplest and smallest carboxylate ligand and the atomic-level homogenous distributions of elements promises a short migration distance of metal ions, enabling a fast phase formation of layered K x MnO 2 in minutes (Fig. 1). The phase transformation process from MOF to layered metal oxide was systematically investigated by thermogravimetric analysis (TGA), variable temperature optical microscopy and X-ray diffraction (XRD). The fast-synthesized layered K x MnO 2 achieved excellent electrochemical performance with 119 mA h g 1 and good cycling stability, that surpasses most previously reported LTMO cathodes (Table S2 †), highlighting the high production efficiency of LTMOs for future large-scale manufacturing and application of KIBs. ## Materials All the reagents and solvents, including manganese chloride tetrahydrate (MnCl 2 $4H 2 O, Aladdin), potassium formate (HCOOK, Acros), formic acid (HCOOH, Tianjin Jiangtian Chemical Technology Co., Ltd) and methanol (CH 3 OH, Tianjin Jiangtian Chemical Technology Co., Ltd) were commercially obtained and used without further purifcation. Synthesis of K[Mn(HCOO) 3 ]. The K[Mn(HCOO) 3 ] precursor was prepared according to a previous report. 34 Solutions of MnCl 2 $4H 2 O (25 mM) in methanol (25 mL), HCOOK (20 mM) in methanol (25 mL) and 5 mL HCOOH were mixed in 100 mL glass vial. The resulting solution was kept at room temperature under the static condition for 24 h without any stirring, and then the crystals were obtained and washed with methanol and air-dried. Synthesis of KMO-F. The K[Mn(HCOO) 3 ] precursor was loaded in a porcelain boat and calcined in a tube furnace that was pre-heated to 1000 C for 8 min. The resulting product was immediately transferred and stored in a glove box flled with Ar. Synthesis of KMO-S. The K[Mn(HCOO) 3 ] precursor was loaded in a crucible and calcined in a tube furnace that was calcined at 400 C for 1 h, followed by calcination at 900 C for 15 h with a heating rate of 2 C min 1 in air. When it was cooled to 200 C, the resulting product was immediately transferred and stored in a glove box flled with Ar. ## Materials characterization Single-crystal X-ray diffraction data for the K[Mn(HCOO) 3 ] MOF were collected on a Rigaku XtalAB Pro MM007 DW at 298 K. Room temperature X-ray diffraction (XRD) patterns were recorded on a Rigaku MiniFlex 600 X-ray diffractometer at 40 kV and 15 mA with a Cu-target tube. Thermogravimetric analysis (TGA) was carried out on a Rigaku TG-DTA 8121 analyzer at a rate of 10 C min 1 from 25 to 900 C. Scanning electron microscopy (SEM) (JEOL JSM-7500F), transmission electron microscopy (TEM), and high-resolution transmission electron microscopy (HRTEM) (JEM-ARM200F) were used to investigate the material morphologies and structures. Variable temperature optical microscope photograph images were taken on a Linkam THMS600 with a thermal stage. XPS measurements were performed on a Thermo Scientifc ESCALAB 250. ICP-OES (Perki-nElmer 8300) was used to analyze the composition of elements. ## Electrochemical tests The working electrodes were prepared by casting the slurry of layered K x MnO 2 , Ketjen black (KB) and polyvinylidenefluoride (PVDF) at a weight ratio of 7 : 2 : 1 on aluminum foil. Coin-type cells were fabricated using potassium foil as the counter electrodes, 2.5 M potassium bis(fluorosulfonyl)imide (KFSI) in triethyl phosphate (TEP) as the electrolyte, and a glass fber membrane in a glove box flled with highly pure argon gas (O 2 and H 2 O levels < 0.1 ppm) as the separator. Galvanostatic discharge/charge tests were performed in a voltage range of 2.0-4.2 V (vs. K/K + ) on Neware battery testers. CV profles were collected on a Solartron 1470 Electrochemical Interface. For operando XRD, an electrolytic cell (Beijing Scistar Technology Co. Ltd) with one side beryllium (Be) window (20 mm in diameter) was employed and cycled at a current density of 50 mA g 1 between 2.0 and 4.2 V (vs. K + /K). The operando XRD patterns were recorded with a scan speed of 3 min 1 and a time step of 0.01 s. ## Results and discussion The KM-MOF was synthesized by a self-assembly process using precursors of manganese chloride tetrahydrate (MnCl 2 $4H 2 O), potassium formate (HCOOK) and formic acid (HCOOH) in methanol at room temperature. 34 X-ray crystallographic analysis reveals that the KM-MOF crystallizes in the monoclinic space group C2/c, with Mn(II) ions positioned at an inversion center. Each Mn(II) octahedron is coordinated with six Mn(II) octahedra through four equatorials syn-anti and two axial anti-anti formate bridges. The structure can be regarded as a distorted perovskite anionic framework with zig-zag chains of K ions in the channels (Fig. 2a and S2 †). The ultra-small distance (3.599 ) and uniform distributions of K and Mn ions in the KM-MOF would promote the phase transformation from MOF to LTMO in a short time. The XRD pattern of the as-synthesized KM-MOF is consistent with the simulated data, suggesting a successful synthesis of the KM-MOF (Fig. S1 †). A spindle-shaped crystal morphology of KM-MOF was simulated by using the Bravais-Friedel-Donnay-Harker (BFDH) method (Fig. 2b), which is verifed by the optical microscope photograph (Fig. S3 †). In general, the composition and structure of MOF precursors signifcantly affect the pyrolysis process and the corresponding derivative. Therefore, an insight into the KM-MOF to layered K x MnO 2 transformation process is required to improve the understanding of the rapid synthesis. To probe the transformation process, component evolution was monitored during the calcination process by TGA. TGA measurements were performed from 50 to 900 C with a heating rate of 10 C min 1 in air. As shown in Fig. 2c, a remarkable weight loss of 42.0% occurred from 250 to 410 C, which is assigned to the decomposition of the KM-MOF and the formation of metal oxide. Theoretically, the organic species were burned into CO 2 , leading to 57.6% weight loss according to the theoretical calculations, while the formation of metal oxide caused 14.0% weight gain upon the uptake of oxygen. The TGA measurement result is consistent with the theoretical values. A small weight loss of 3.0% was displayed from 700 to 800 C, which should be attributed to the sublimation of potassium because of its low boiling point of 759 C. The bond lengths of the KM-MOF obtained by using crystalline data are in the order of K-O > Mn-O > C-O (Table S3 †), indicating that the bond strengths are in the order of K-O < Mn-O < C-O. Therefore, during the thermal heating process, the K-O and Mn-O bonds would break prior to the C-O bonds. The morphology evolution of the KM-MOF is also monitored by variable temperature optical microscopy, and the optical microscope photographs at different temperatures are shown in Fig. 2d-g. At low temperatures below 300 C, there is no obvious change to the crystalline morphology (Fig. 2d and e). However, when the temperature increased to 350 C, the crystals became opaque (Fig. 2f), and molten at 400 C (Fig. 2g), suggesting damage of the crystalline structure. These results are well consistent with the TGA measurement results. To gain further insights into the phase conversion process from the KM-MOF to layered K x MnO 2 , variable temperature XRD was conducted (Fig. 2h). The diffraction peak of the (200) crystal plane that is based on K-O bonds disappeared at 250 C (Fig. S4 †). When the temperature increased to 300$500 C, the KM-MOF almost lost its crystallinity. From 600 C, new peaks appeared at 12.7 and 25.4 , which can be indexed to the planes of (001) and (002) of layered K x MnO 2 , respectively, indicative of the beginning of layered phase formation. These two peaks slightly shifted to lower diffraction angles of 12.6 for (001) and 25.2 for (002) planes from 800 to 900 C, indicating that the interlayer spacing became larger. This can be explained by the enhanced coulombic repulsion between the MO 6 slabs upon K + sublimation. The interlayer spacing is calculated to be 0.69 nm based on the diffraction peak of the (001) plane. The variable temperature XRD measurement results show that layered K x MnO 2 can be facilely synthesized from the KM-MOF through thermal sintering. Based on the aforementioned analysis, the transformation process from the KM-MOF to layered K x MnO 2 is schematically illustrated in Fig. 2i. During the heating process, K-O and Mn-O bonds successively break along with the uptake of oxygen to form K and Mn oxides from 300 to 400 C, and then the layered phase of K x MnO 2 is formed above 600 C. An ultra-fast synthesis method was developed to produce layered K x MnO 2 , by sintering the KM-MOF at 1000 C for only 8 minutes, and the resulting product is denoted as KMO-F. As a control, layered K x MnO 2 was also prepared using a conventional long-time sintering method from the same MOF precursor, denoted as KMO-S. Their layered crystal structures were revealed by the XRD patterns, and they are in good agreement with the previously reported results in the literature. 35 The pronounced peaks at 12.7/25.4 of KMO-S and 12.6/ 25.2 of KMO-F are assigned to the (001)/(002) planes (Fig. 3a and S5 †), and the interlayer spacings are calculated to be 0.69 and 0.68 nm, respectively. Rietveld refnement suggests that both KMO-F and KMO-S belong to the hexagonal symmetry with a space group of P6 3 /mmc. The lattice parameters of KMO-F phase are calculated to be a ¼ b ¼ 2.881 and c ¼ 14.132 , while they are a ¼ b ¼ 2.880 and c ¼ 14.062 for KMO-S. Therefore, they have the same crystalline structure, and the schematic crystal structure of the layered K x MnO 2 is shown in Fig. 3b. Mn ions are octahedrally coordinated with oxygen anions while K ions are located between the layers of MnO 6 octahedra. The overall K/Mn atomic ratios were measured by using inductively coupled plasma (ICP) to be 0.53 for KMO-F and 0.51 for KMO-S (Table S4 †), corresponding to the molecular formulas of K 0.53 MnO 2 and K 0.51 MnO 2 , respectively. The chemical properties of the layered compounds were investigated by X-ray photoelectron spectroscopy (XPS) (Fig. S6 †). The high-resolution Mn 3s spectrum of KMO-F shows a peak energy separation (DE) of 5.06 eV (Fig. S7 †), consistent with the previous reports. 35 According to the linear relationship between the chemical valence of Mn and the DE value, the average oxidation state of Mn in KMO-F was calculated to be 3.47 (Table S5 †). Correspondingly, the Mn 2p 1/2 and Mn 2p 3/2 peaks in the high-resolution Mn 2p spectrum can be deconvoluted into two peaks of Mn 3+ and Mn 4+ with equal integral areas (Fig. 3c), suggesting an average valence of 3.5, very close to the value of 3.47. Similar results were obtained for KMO-S. The morphologies and microstructures of the layered KMOs were characterized by scanning electron microscopy (SEM) and high-resolution transmission electron microscopy (HRTEM). The SEM images of KMO-F and KMO-S present similar morphologies with crystal particles in random shapes and sizes (Fig. 3d and f). The HRTEM images reveal highly crystalline structures with interlayer spacings of 0.69 nm for KMO-F and 0.68 nm for KMO-S (Fig. 3e and g), corresponding to the (001) plane. Uniform elemental distributions of K, Mn and O were measured over the whole particles by scanning transmission electron microscopy (STEM) (Fig. 3h and S8 †), suggesting a wellformed layered structure. Apparently, well-crystallized layered K x MnO 2 can be produced by the fast synthesis method, similar to the conventional long-time synthesis methods from the same MOF precursor, demonstrating the effectiveness and superiority of the new method. The electrochemical performances of KMO-F and KMO-S were investigated and compared under the same test conditions. Coin-type cells were fabricated using KMO-F or KMO-S cathodes and 2.5 M KFSI/TEP as the electrolyte. The cyclic voltammetry (CV) measurements were carried out at a scan rate of 0.1 mV s 1 in a voltage window of 2.0-4.2 V (vs. K + /K). Similar CV curves were displayed for the two cathodes with fve redox couples (Fig. 4a and b), suggesting similar electrochemical behaviors. These current peaks can be assigned to the reversible potassium insertion/extraction into/from layered K x MnO 2 with Mn 3+ /Mn 4+ redox reaction. Obviously, the well-overlapped CV curves after the frst cycle predict good cycling stability for the KMO cathodes. The galvanostatic charge/discharge profles at a current density of 50 mA g 1 are shown in Fig. 4c and d. The frst charge/discharge capacities of KMO-F and KMO-S are 119/ 133 and 84/111 mA h g 1 , respectively. Although KMO-F delivered a slightly higher initial capacity, similar cycling performance was obtained for them (Fig. 4e). KMO-F and KMO-S retained 80% and 85% of their capacity after 100 cycles at 50 mA g 1 , respectively, with high average coulombic efficiencies of about 99%. In addition, the KMO-F cathode delivers an initial discharge capacity of 81.14 mA h g 1 at 300 mA g 1 and retains 86% of its capacity after 120 cycles (Fig. S9 †). Good rate performance was obtained for both KMO-F and KMO-S (Fig. 4f). When the current density increased from 50 to 1000 mA g 1 , KMO-F maintained a specifc capacity of 65 mA h g 1 , compared with 73 mA h g 1 for KMO-S. The electrochemical performance once again manifests the feasibility of the fast synthesis method for high performance layered cathode materials. To further investigate the K-storage behavior for the KMO cathodes, CV measurements at various sweep rates from 0.1-2.1 mV s 1 were carried out (Fig. 4g and h). The capacity contributions from surface pseudo-capacitance and ion intercalation can be quantifed by separating the current response i at a fxed potential V into capacitive-controlled (k 1 v) and diffusion-controlled (k 2 v 1/2 ) processes, according to the following equation: 36 i(V) ¼ k 1 v 1 + k 2 v 1/2 . In our work, based on the integration of CV curves, 91.2% of the total charge storage capacity of the KMO-F cathode is capacitive at a sweep rate of 1.5 mV s 1 . Such high proportions of pseudocapacitive contribution result from fast kinetics, which should account for the good rate capability of the KMO-F cathode. The surface capacitive-controlled capacities and diffusion-controlled capacities for KMO-F and KMO-S at different scan rates are summarized in Fig. 4i. KMO-S shows slightly higher diffusioncontrolled capacities at various sweep rates. The K storage process of the layered cathodes was investigated by in situ XRD, and the data are shown in Fig. 5. By charging (extraction of K ions), the (001) and (002) peaks shifted to lower 2-theta degrees, indicating the increased interlayer spacing due to enhanced electrostatic repulsion between the adjacent oxygen layers upon K + extraction. During the discharge process, a reversible peak shift was observed, demonstrating good structural reversibility and cycling stability. ## Conclusions In this work, a fast synthesis method is developed by directly calcining an alkali-transition metal based-MOF of K [Mn(HCOO) 3 ] for preparing layered K x MnO 2 cathode materials thanks to the ultra-small distance (3.599 ) and uniform distributions of K and Mn ions in the KM-MOF. The new method saves time and energy compared with the traditional long-time high-temperature synthesis method. The fast phase transformation process was systematically investigated by multiple analysis techniques and theoretical simulations. The layered K x MnO 2 cathode demonstrated excellent potassium storage performance with high reversibility and cycling stability, which is similar to the cathode produced by a longtime-sintering method from the same precursor and is superior to most previously reported LTMO cathodes in KIBs. Our fndings provide a new strategy for synthesizing high performance layered cathode materials for KIBs.

chemsum

{"title": "Rapid synthesis of layered K_<i>x</i>MnO₂ cathodes from metal\u2013organic frameworks for potassium-ion batteries", "journal": "Royal Society of Chemistry (RSC)"}

co/ba/la2o3_catalyst_for_ammonia_synthesis_under_mild_reaction_conditions

## Abstract: Ruthenium catalysts may allow realization of renewable energy-based ammonia synthesis processes using mild reaction conditions (<400 °C, <10 MPa). However, ruthenium is relatively rare and therefore expensive. Here, we report a Co nanoparticle catalyst loaded on a basic Ba/La2O3 support and pre-reduced at 700 °C (Co/Ba/La2O3_700red) that showed higher ammonia synthesis activity at 350 °C and 1.0-3.0 MPa than two benchmark Ru catalysts, Cs + /Ru/MgO and Ru/CeO2. The synthesis rate of the catalyst at 350 °C and 1.0 MPa (19.3 mmol h −1 g −1 ) was 8.0 times that of Co/Ba/La2O3_500red and 6.9 times that of Co/La2O3_700red. The catalyst showed activity at temperatures down to 200 °C. High-temperature reduction induced formation of a BaO-La2O3 nano-fraction around the Co nanoparticles, which increased turnover frequency, inhibited Co nanoparticle sintering, and suppressed ammonia poisoning. These strategies may also be appliable to nickel catalysts. ## Introduction Ammonia (NH3) is an essential chemical feedstock in the modern chemical industry. More than 80% of the ammonia generated today is used as chemical fertilizer, and ammonia has made a huge contribution to solving the food crisis that resulted from the population explosion in the 20th century . Recently, ammonia has attracted attention as a hydrogen and energy carrier for greater utilization of renewable energy, and as a decarbonized fuel for use in power plants and ships . Ammonia, therefore, is considered an important material for realizing a sustainable society. Traditionally, ammonia has been produced via the Haber-Bosch process. The Haber-Bosch process uses very high pressures and temperatures (>450 °C and >20 MPa) and has been highly optimized from a process engineering standpoint. However, the process uses fossil fuels as its source of hydrogen, and therefore it emits large amounts of CO2 (1.9 ton-NH3 −1 ) to the atmosphere . If ammonia could be produced from hydrogen produced by renewable energy, the process could be harnessed to speed up decarbonization, slow down global warming, and increase food production, which are three important current global issues . A major milestone in the realization of green ammonia synthesis systems using H2 produced by renewable energy is the development of catalysts that have high ammonia activity under mild conditions (<400 °C, <10 MPa). The catalysts most commonly used in the Haber-Bosch process are Fe-based; however, these catalysts require a high temperature and pressure to dissociate the N≡N triple bond (945 kJ mol −1 ) and so are not suitable for use in conjunction with renewable energy . In contrast, Ru catalysts show unparalleled ammonia synthesis activity under mild conditions , but Ru is a rather rare element that is expensive to procure. Co is cheaper and more abundant than Ru, but neat Co is less active than both Ru and Fe because the N2 molecular adsorption energy of Co is lower than that of Ru and Fe . As a result, previously reported oxide-or carbon-supported Co catalysts, which have the advantage of being easy to prepare and handle, all show low ammonia synthesis activity under mild conditions . As part of efforts to address this issue, we previously reported that the addition of Ba to Co/MgO and pre-reduction at high temperature markedly improved the ammonia synthesis activity of the parent catalyst under mild reaction conditions (<400 °C, 1-3 MPa) . In fact, the resultant Co@BaO/MgO catalyst, where the Co core is encapsulated by a strongly basic BaO shell, showed an activity that was not only higher than that of other oxide-or carbon-supported Co-based catalysts, but also higher than that of active Ru catalysts such as Ru/La0.5Ce0.5O1.75 and Cs-Ru/MgO . Characterization of our catalyst revealed that the structure allows for exceptional electron donation from BaO via Co atoms to the antibonding π-orbital of N2 molecules, thereby promoting cleavage of the N≡N bond. Our findings prompted us to examine the use of other basic oxide supports, such as La2O3, which has a higher basicity than does MgO . Here, we report that Co/Ba/La2O3 pre-reduced at 700 °C showed high ammonia synthesis rates at a low reaction temperature of 350 °C: 19.3 mmol h −1 g −1 at 1.0 MPa and 35.7 mmol h −1 g −1 at 3.0 MPa. Investigations revealed that addition of Ba and increasing the pre-reduction temperature from 500 to 700 °C increased the ammonia synthesis activity by 6.9 and 8 times, respectively, via the formation of a core (Co) -shell (BaO-La2O3) structure. The presence of BaO retarded sintering of the Co nanoparticles during high-temperature reduction. The ammonia synthesis activitypromoting effects of the addition of Ba and of increasing the reduction temperature were observed also for Ni catalysts, for which the N2 adsorption energy of neat Ni is less than that of Co . To further understand how the addition of Ba and the increase of reduction temperature affected the ammonia synthesis rate, X-ray absorption fine structure spectroscopy (XAFS), spherical aberration-corrected scanning transmission electron microscopy (Cs-STEM), and energy electron loss spectroscopy (EELS) analyses were performed on pre-reduced catalyst without exposure to air. In the present study, both EELS and energy-dispersive X-ray (EDX) spectrometry were used; however, EELS can distinguish between Ba and La elements, whereas EDX spectrometry cannot because in that technique the excitation wavelength of Ba and La are so similar that the small peak attributable to Ba overlaps the large peak attributable to La. ## Catalyst preparation The Ba/La2O3 support was prepared by a precipitation and impregnation method as follows. First. a suspension of La hydroxides was formed by dropping an aqueous solution of La(NO3)3•6H2O (Wako Pure Chemical, Japan) into a 28 wt% solution of aqueous ammonia (Wako Pure Chemical). The La hydroxides were collected by filtration, washed with distilled water, and added to an aqueous solution containing Ba(OH)2•8H2O (Wako Pure Chemical). After stirring the suspension for 1 h, the aqueous solvent was removed by rotary evaporation. The resulting powder was calcined at 700 °C in static air and used as the catalyst support. Next, Co was loaded onto the support. Bis(2,4-pentanedionate)cobalt(II) dihydrate (Tokyo Chemical Industry, Japan) dissolved in tetrahydrofuran (Wako Pure Chemical) was used as the Co precursor. The support was added to the dissolved in the precursor and the suspension was stirred overnight. When the stirring was finished, the tetrahydrofuran was removed by rotary evaporation, leaving behind a powder that was then heated to 500 °C under an Ar flow. The Co loading was fixed at 20 wt% for each catalyst. Two benchmark Ru catalysts (Ru/CeO2 and Cs + /Ru/MgO) were also prepared as we reported previously . ## Ammonia synthesis activity test The rate of ammonia synthesis over the catalysts was measured by using 100 mg of catalyst and a conventional flow system with tubular reactor under either atmospheric pressure or high pressure, as reported previously . Research-grade gases (>99.99%) were supplied from high-pressure cylinders and purified with a gas purifier (Micro Torr MC50-904FV, SAES Pure Gas, US). The catalysts were pre-reduced in situ with pure H2 (60 mL min −1 ) at 500, 700, or 800 °C for 1 h at 0.1 MPa and then cooled at 300 °C in an Ar stream. The pressure was then adjusted to 0.1, 1.0, or 3.0 MPa. A mixture of N2 (30 mL min −1 ) and H2 (90 mL min −1 ) was then passed over the catalyst (space velocity = 72,000 mL h −1 g −1 ). The produced ammonia gas was trapped in an aqueous solution of H2SO4, and the rate of ammonia synthesis was calculated from the decrease in the electron conductivity of the H2SO4 aqueous solution, which was monitored with an electron conductivity detector (CM-30R, DKK-TOA, Japan). ## Kinetic analysis Reaction kinetics were analyzed as previously reported . The reaction orders with respect to N2, H2, and NH3 were calculated by measuring the N2, H2, and NH3 pressure dependence of the NH3 synthesis rate and by assuming that the rate of the reaction (r) could be described by the following expression: ## Characterization High-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) images, EDX elemental maps, and EELS spectra were obtained with an aberration-corrected electron microscope (JEM-ARM200CF, JEOL, Japan). The scanning transmission electron microscopy observations were conducted at 120 kV to reduce damage to the sample by the electron beam. Catalyst samples were pre-reduced at 500 or 700 °C under a H2 flow and then crushed and powdered. Samples of the powdered catalyst were then placed on transmission electron microscopy grids in a glovebox. The grids were then transferred by means of a special holder with a gas cell from the glovebox to the inside of the transmission electron microscopy column without being exposed to air. For the other observations, samples were dispersed in ethanol under ambient conditions, and samples of the dispersion were dropped onto a carboncoated copper grid and dried under a vacuum at ambient temperature for 24 h. The specific surface area of the catalysts was measured by using the Brunauer-Emmett-Teller method. Test samples were pretreated at 300 °C in a vacuum, and the amount of N2 adsorbed was measured with a BELSORP-mini gas adsorption instrument (BEL Japan, Inc., Japan). H2 chemisorption capacity was measured with a BELCAT-B apparatus (MicrotracBEL, Japan). H2 was fed to a 100-mg sample of catalyst at 60 mL min −1 , and the temperature was increased at a rate of 10 °C min -1 from room temperature to 500, 700, or 800 °C. The sample was maintained at the desired temperature for 60 min in the H2 flow; it was then purged with a stream of Ar (60 mL min −1 ) for 30 min, cooled to 35 °C, and flushed with Ar for 60 min. After pretreatment, H2 chemisorption measurement was carried out at 35 °C. XAFS measurements of the Co K-edges were performed on the B01B1 beamline at Spring-8(Hyogo, Japan) with the approval of the Japan Synchrotron Radiation Research Institute (Hyogo, Japan). X-ray diffraction (XRD) analysis was performed with a SmartLab X-ray diffractometer (Rigaku, Japan) equipped with a CuKα radiation source. The XRD patterns were analyzed by using the PDXL2 software (Rigaku) and three databases (International Centre for Diffraction Data database, Crystallography Open Database, and AtomWork database). Temperature-programmed reduction measurements were performed under a flow of 100% H2 using a BEL-CAT-II apparatus (MicrotracBEL). The flow rate of the gas was 60 mL min −1 . A 100mg sample of catalyst was heated from 25 to 1000 °C at a rate of 10 °C min −1 . The CH4, H2O, CO, and CO2 profiles were monitored by a quadruple mass spectrometer at m/e =16, 18, 28, and 44, respectively. ## Ammonia synthesis activity First, we examined the influence of reduction temperature (500-800 °C) on the NH3 synthesis rate of the Co/Ba/La2O3 catalyst (Fig. 1). The catalyst pre-reduced at 500 °C (designated Co/Ba/La2O3_500red) showed little NH3 synthesis activity at 300 °C, but the activity increased with increasing reaction temperature until a moderate activity was obtained at 450 °C. Compared with Co/Ba/La2O3_500red, Co/Ba/La2O3_600red and Co/Ba/La2O3_700red both showed greater NH3 synthesis activity across the whole reaction temperature range examined, and Co/Ba/La2O3_800red showed an activity between that of Co/Ba/La2O3_600red and Co/Ba/La2O3_700red. The apparent activation energy of the Co/Ba/La2O3 catalyst was found to decrease from 73.1 to 45.7 kJ mol −1 as the reduction temperature was increased from 500 to 700 °C (Fig. S1). Also, Co/Ba/La2O3_700red was found to have ammonia synthesis activity at reaction temperatures as low as 200 °C (synthesis rate, 0.3 mmol h −1 g −1 ; Fig. S2). Next, we examined the effect of adding Ba on the NH3 synthesis rate of different Co and Ni catalysts (Fig. 2). At a reaction temperature of 350 °C, the NH3 synthesis rate of Co/Ba/La2O3_700red was 19.3 mmol h −1 g −1 , which was 8.0 times that of Co/Ba/La2O3_500red and 6.9 times that of Co/La2O3_700red at the same temperature. Despite the activity's of Co/Ba/La2O3_700red being slightly lower than that of Co/Ba/MgO_700red, overall its activity was comparable with those reported for other Co catalysts, notwithstanding the different Co loadings, reaction pressures, and space velocities used in the previously reported studies (Table S1). For the Ni catalysts, reduction at high temperature and the addition of Ba were also found to have positive effects on NH3 synthesis activity: the fact that the NH3 synthesis rate of Ni/Ba/La2O3_700red was much higher than those of Ni/Ba/La2O3_500red and Ni/La2O3_700red at all temperatures suggested the possibility of activating Ni by using the same strategy used to activate Co. We also examined the effect of reaction pressure on the NH3 synthesis rate at 350 °C and compared the rates with Co/Ba/La2O3_700red to those with two benchmark Ru catalysts, Cs + /Ru/MgO and Ru/CeO2 (Fig. 3). Cs + /Ru/MgO is a well-known Ru catalyst with high NH3 synthesis activity . Ru/CeO2 is a candidate catalyst for use in ammonia synthesis processes that use renewable energy . At 0.1 MPa, the NH3 synthesis rate of Co/Ba/La2O3_700red was lower than those of Cs + /Ru/MgO_500red and Ru/CeO2_500red. However, when the reaction pressure was increased, the NH3 synthesis rate of Co/Ba/La2O3_700red increased drastically, whereas those of the two Ru catalysts increased only slightly. As a result, the NH3 synthesis rate of Co/Ba/La2O3_700red at 1.0 MPa and at 3.0 MPa exceeded the rates of the two benchmark catalysts (synthesis rate at 3.0 MPa: Co/Ba/La2O3_700red, 35.7 mmol h −1 g −1 ; Ru/CeO2_500red, 13.7 mmol h −1 g −1 ; and Cs + /Ru/MgO_500red, 11.4 mmol h −1 g −1 ). We also examined the NH3 synthesis rate of Co/La2O3_700red and found that it was much lower than the rates of the other catalysts at all pressures. To investigate the cause of the different pressure dependences, we performed a kinetic analysis at 350 °C and 0.1 MPa using the same four catalysts (Table 1 and Figs. S3 and S4). The fact that the reaction order with respect to N2 was almost unity for all of the catalysts indicated that the rate-determining step was dissociation of molecular N2. However, the fact that the reaction order with respect to H2 was −0.18 and −0.76 for Ru/CeO2_500red and Cs + /Ru/MgO _500red, respectively, indicated that H atoms strongly adsorbed onto the Ru surface and inhibited activation of molecular N2; this is referred to as hydrogen poisoning and is a typical drawback of Ru catalysts. In contrast, the positive reaction orders with respect to H2, +0.43 and +0.32, that were obtained for Co/Ba/La2O3_700red and Co/La2O3_700red, respectively, indicated that these catalysts were free from hydrogen poisoning and that N2 activation was promoted with increasing hydrogen pressure. A large negative reaction order with respect to NH3 (−0.51) was obtained for Co/La2O3_700red, but this value was reduced to −0.17 with the addition of Ba. These findings suggest that it is difficult for the ammonia yield of Co/La2O3_700red to approach the equilibrium value because adsorbed NH, NH2, and NH3 inhibit the reaction. However, the addition of Ba can be expected to promote the desorption of such adsorbates and thus accelerate the reaction, even near equilibrium. Together, these results indicate that Co catalysts may be a viable alternative to Ru catalysts. ## Effects of doping with Ba To understand more about the effects of Ba doping, we compared the physicochemical properties of Co/Ba/La2O3_700red and Co/La2O3_700red (Table 2). The specific surface areas of the two catalysts were comparable, but the mean Co particle size, as measured by STEM, was much smaller for Co/Ba/La2O3_700red than for Co/La2O3_700red (20 nm vs. 70 nm). The indication was that the addition of Ba inhibited sintering of the Co particles during reduction. Assuming the Co particles were cubic, this difference in particle size corresponds to 3.5-times greater Co dispersion for Co/Ba/La2O3_700red compared with Co dispersion for Co/La2O3_700red. On the other hand, the H2 chemisorption value (a measure of the Co dispersion) of Co/Ba/La2O3_700red was found to be only 1.6-times greater than that of Co/La2O3_700red. The indication was that the surface of the Co particles in Co/Ba/La2O3_700red was partly covered by the support material. Also, the turnover frequency (TOF) of Co/Ba/La2O3_700red was about 4.4 times that of Co/La2O3_700red. Together, these results indicate that the drastic increase in NH3 synthesis rate observed as a result of doping with Ba was the combined result of inhibition of Co particle sintering and an increase of TOF. To understand how doping with Ba affected the state of the Co, we subjected Co/Ba/La2O3_700red and Co/La2O3_700red to XANES analysis. The fact that the Co K-edge XANES spectra of the two catalysts were comparable with that of Co foil indicated that the Co atoms in the catalysts were fully reduced to a metallic state after reduction at 700 °C. Next, we investigated the morphology of Co/Ba/La2O3_700red by means of Cs-STEM and EELS. To avoid any unwanted structural or state changes, we used a special holder with a gas cell that allowed the sample to be transferred from the reactor to the STEM apparatus under an inert gas environment . HADF-STEM images revealed that the Co particles were encapsulated by a 2-3-mm-thick nano-fraction (Fig. 5a and Fig. S9). EELS mapping revealed that Ba and La elements were enriched in the nano-fraction(Figs. 5b, 5c, 4d, 5e). Although a low abundance of Ba was included overall in the catalyst (Ba:La molar ratio, 5:95), a high abundance of Ba was observed in the nano-fraction. In contrast, the fact that carbon element was not detected in the nano-fraction (Fig. 5f, 5g, 5h) indicated that the nano-fraction was an oxide or hydroxide of Ba or La. ## Effect of reduction temperature Finally, to understand the effects of reduction temperature on Co/Ba/La2O, we investigated the physicochemical properties of the catalyst after reduction at different temperatures. With the increase in reduction temperature from 500 to 700 to 800 °C, the specific surface area decreased from 37.5 to 24.9 to 10.1 m 2 g −1 , and the mean diameter of the Co particles increased from 10 to 20 to 34 nm. The indication was that greater sintering occurred at higher reduction temperature (Table 2). It must be noted that, the fact that the mean Co particle size was comparable both before and after exposure of Co/Ba/La2O3_700red to air indicated that exposure to air had no effect on Co particle size (see Figs. 5, S7, and S9). Also, the number of exposed Co particles decreased and the TOF drastically increased from 0.019 to 0.223 to 0.304 with increasing reduction temperature. To understand why the TOF after reduction at 700 °C was more than 12 times the TOF after reduction at 500 °C, we characterized the catalysts by using several methods. To investigate the state of Co, we measured Co-K edge XANES spectra in as-prepared Co/Ba/La2O3 and Co/Ba/La2O3 pre-reduced at different temperatures (Fig. 4). The fact that the spectra for Co/Ba/La2O3_500red and Co/Ba/La2O3_700red were comparable with that for Co foil indicated that inactive oxidic Co was reduced to a metallic state after reduction at ≥500 °C. To investigate the surface state of Co/Ba/La2O3_500red, we performed Cs-STEM and EELS observations without exposing the catalyst to the air (Figs. 6 and S10). In contrast to the findings for Co/Ba/La2O3_700red (Figs. 5 and S9), the Co particles in Co/Ba/La2O3_500red were not well crystalized and were partially surrounded by a cloud-like substance containing Ba, La, and carbon elements. The fact that carbon element was observed in parts of the cloud in Co/Ba/La2O3_500red indicated the presence of non-crystalized carbonate species of Ba and/or La, because crystalized carbonate species was not observed by XRD analysis (Fig. 6, Fig. S11). Recall that such carbon element was not observed for Co/Ba/La2O3_700red by EELS measurement (Figure 5 and S9). Carbonate species are acidic and as such decrease the electron-donating ability of the support material. Therefore, we concluded that one of the causes for the high TOF of Co/Ba/La2O3_700red was complete removal of La and Ba carbonate species during reduction at high temperature. It is likely that hydroxide species are also removed from the catalyst because the temperatureprogrammed reduction profile of fresh Co/Ba/La2O3 indicated that the formation of CH4, CO, CO2, and H2O was completed at a temperature below 700 °C (Fig. 7). Thus, these data indicated that the low crystalline nano-fraction encapsulating the Co nanoparticles in Co/Ba/La2O3_700red was composed of BaO and La2O3. Therefore, another of the causes of the high TOF of Co/Ba/La2O3_700red was ascribed to encapsulation of the Co particles by the low crystalline nano-fraction of BaO-La2O3. Previously, based on density functional theory calculations and Fourier-transform infrared spectroscopy measurements after adsorption of molecular N2 on the catalyst, we reported the high activity of Co nanoparticles encapsulated by BaO loaded on MgO (Co@BaO/MgO), which we ascribed to strong electron donation from BaO to N2 via Co . Although in the present study we were unable to fabricate a self-supporting disk of the Co/Ba/La2O3 catalyst that transmitted infrared light and therefore could not use Fourier-transform infrared spectroscopy, the high TOF of Co/Ba/La2O3_700red may also be due to such electron donation from BaO-La2O3 to the antibonding π-orbital of the N≡N bond of molecular N2. We also found that increasing the reduction temperature up to 800 °C further increased the TOF; however, the NH3 synthesis rate decreased due to sintering of the Co particles (Table 2). Co/Ba/La2O3_700red and Co/Ba/MgO_700red showed comparable TOFs, but the NH3 synthesis rate (350 °C, 1 MPa) of Co/Ba/MgO_700red was 1.3 times that of Co/Ba/La2O3_700red. These results indicated that fundamentally the core (Co) -shell (BaO-La2O3) structure enhanced the NH3 synthesis ability of surface Co, and that the difference of NH3 synthesis rate between these catalysts was due to the difference of the mean Co particle size (i.e., 10.6 vs. 20 nm for Co/Ba/MgO_700red and Co/Ba/La2O3_700red). The higher surface area of Co/Ba/MgO_700red (47.6 m 2 g −1 ) compared with that of Co/Ba/La2O3_700red likely contributed to the formation of fine Co nanoparticles. Thus, use of a basic support with a higher specific surface area is expected to afford catalysts with enhanced NH3 synthesis rates. ## Conclusions Here, we demonstrated that encapsulation Co nanoparticles by BaO-La2O3 on La2O3 support exhibited high NH3 synthesis activity under mild reaction conditions. The Co/Ba/La2O3 prereduced at 700 °C showed high NH3 synthesis activity with a synthesis rate of 19.3 mmol h −1 g −1 at 350 °C, which was 6.9 times that of nondoped parent catalyst. Moreover, the Co/Ba/La2O3 prereduced at 700 °C was active at temperatures down to 200 °C. We also found that addition of Ba Table 1. Results of a kinetic analysis over Cs + /Ru/MgO, Ru/CeO2, Co/Ba/La2O3, or Co/La2O3. [a] Reaction order with respect to N2. [b] Reaction order with respect to H2. [c] Reaction order with respect to NH3. ## Catalyst Order [a] n [a] h [b] a [c] Cs + /Ru/MgO_500red Table 2. Physicochemical properties and catalytic performances of supported co catalysts. [a] Specific surface area. [b] Measured using H2 chemisorption capacity. [c] Mean particle size of Co nanoparticles, as estimated by Cs-STEM (see Figs. S5-8). [d] At 350 ºC and 1.0 MPa. [e] Turnover frequency. Calculated from the H2 chemisorption value and the NH3 synthesis rate (see Table 1). Catalyst SSA [a] [m 2 gcat −1 ] H2 chemisorption [b] [μmol g −1 ] d [c] [nm] Rate [d] [mmol gcat −1 h −1 ] TOF [e] [s − ## Kinetic analysis Reaction kinetics were analyzed as previously reported . Reaction orders with respect to N2, H2, and NH3 were determined by measuring N2, H2, and NH3 pressure dependence for the ammonia synthesis rates on the assumption that the rate (r) was described by expression (1). Equations ( 2) to (5) were also used for this analysis: where r, w, y0, q, C, and (1 − m) denotes the ammonia synthesis rate, catalyst mass, ammonia mole fraction at the reactor outlet, flow rate, a constant, and a, respectively. Kinetic analyses were performed at 350 °C and 0.1 MPa. Other reaction conditions are described in Table S2. To avoid any contribution from the reverse reaction, kinetic measurements were carried out at a space velocity where the ammonia concentration at the reactor exit was far from the thermodynamic equilibrium concentration. ## Supplementary results Figure S1. Arrhenius plots for the ammonia synthesis reaction at 1.0 MPa.

chemsum

{"title": "Co/Ba/La2O3 catalyst for ammonia synthesis under mild reaction conditions", "journal": "ChemRxiv"}

evaluation_of_an_on-site_surface_enhanced_raman_scattering_sensor_for_benzotriazole

## Abstract: Benzotriazole (BtAH) has been used for decades as corrosion inhibitor and antifreeze. Since it is fairly soluble in water but very stable and can only be partly removed from wastewater treatment plants, it represents a threat to the environment and thus also to human health. therefore, it is of uttermost importance to have a detection method capable of monitoring the concentration of BtAH at trace level on-site. Here, we demonstrate that a sensor based on surface-enhanced Raman spectroscopy is capable of detecting trace-level concentrations of BtAH. We carefully studied the concentration dependency and the time dependent coverage. Moreover, we could not only perform the measurements with clean solution but also with real samples from a wastewater treatment plant, ensuring that our method proposed works in a complex environment. Benzotriazole (BTAH) is a versatile chemical compound. It has been known for seventy years as an effective corrosion inhibitor for copper and its alloys by preventing undesirable surface reactions 1 . Moreover, BTAH is used in dishwashing detergents as silver protection, as well as in anti-freeze, heating and cooling systems, hydraulic fluids and also in vapor phase inhibitors 2 , leading to a production of 1000 10000 t per year (only) in Europe. The most noteworthy characteristics are however, that BTAH is fairly soluble in water, not readily degradable and has a limited sorption tendency. Therefore, and due to the large variety of applications where BTAH is used, it is found ubiquitously in aquatic systems and can only be partly removed in wastewater treatment plants. In the greater Beijing area with 20 million people, for example, g l 1 0 / µ . were found in wastewater 3 . Furthermore, BTAH has been found in lakes and rivers in Switzerland in concentrations ranging between g l 0 1 / µ . and g l 6 3 / µ . respectively 4 . Hence it follows that, as an identified micro-pollutant, it is of a growing concern in the water resources being threatened in their biodiversity as well as human health . This is why an early detection of BTAH at trace-level concentration can critically avail the environmental monitoring. In order to detect BTAH, sophisticated mass spectroscopy is established at present. Since on-site analysis is becoming more and more important, another appropriate method is urgently needed. Surface-enhanced Raman scattering (SERS) is predestined for this issue, since it is a sensitive spectroscopic method enabling the detection of molecular analytes down to the attogram level 8 . This is particularly attractive because it combines high sensitivity with high information content for establishing molecular identity. The essential plasmonic nanostructures are generated by various procedures including the fabrication of plasmonic nanoparticle arrays assembled by a seed-mediated electroless plating method or single pulse UV-Laser treatment 9,10 . Fan et al. described SERS platforms using nanolithography methods in an overview, including electron-beam (e-beam) lithography and focused ion beam (FIB) as well as template-based methodologies to generate metallic nano-patterns 11 . Mosier-Boss reviewed the fabrication of the most common SERS-active substrates used. Three generic categories are classified: (1) metal nanoparticles in suspension; (2) metal nanoparticles immobilized on solid substrates; and (3) nanostructures fabricated directly on solid substrates by nanolithography and template based synthesis 12 . The potential of SERS as an analytical application has been intensively explored during the last three decades using different kind of substrates or nanoparticles 13 . Examples are the explosive detection for security applications and analyte detection in medical applications . This large interest has led to a commercialization of the technique leading to robust portable Raman spectrometers as well as commercially available SERS-substrates for on-site analysis. One of the first commercially available SERS-substrates was based on conventional optical SeRS-Substrates. The main SERS-substrates of this study (AMO C7) were prepared on silicon dioxide. In short, the SERS-active structures were prepared by soft-UV-nano-imprinting into Amonil resist on glass wafers 21,22 . The structures were then etched 700 nm into the SiO 2 -substrate using reactive ion-etching (RIE) with a CHF 3 -plasma. In this process the resist mask is virtually used up during the etching procedure, thus cleaning processes are obsolete. At the end, a gold layer of 200 nm was evaporated on top with a rate of 1.1/s at p = 1.87 × 10 −5 mbar using physical vapor deposition (Auto 306, BOC Edwards, UK). The resulting conical nanopillar structures of the so-called C7 substrates are shown in SEM-images in Fig. 1, where panel (a) shows the pillars in top view and (b) under a tilt angle of  15 . The samples have a rectangular pattern of the pillars with a pitch size of 375 nm. The pillar diameter is approximately 200 nm and their height is 337 nm, respectively. The active area of a single substrate is mm 7 11 2 ## × . Furthermore, we investigated the applicability of two commercially acquired SERS-substrates for the BTAH-detection: First, the substrates from AtoID (http://www.atoid.com, Lithuania) were used. They exist in two different versions, called "MatoS" (gold coating) and "RandaS" (silver coating). The active SERS area of × mm 5 3 2 was fabricated using ultra-short pulse laser ablation directly on the silver-and gold-coated soda-lime glass substrate, respectively. The resulting structure is stochastically nano-patterned with features between a few nm up to a micron in size 23 . Second, the "SERStrate" substrates from Silmeco (https://www.silmeco.com, Denmark), are made of silicon nanopillars coated with either gold or silver (vide supra). They exhibit an active area of mm 4 4 2 . A two step process is used to make these substrates. First, mask-less dry-etching is done to create the silicon nanopillars followed by electron beam evaporation of gold or silver to coat the silicon 24 . instrumentation. The Raman measurements were performed with a standard Raman system (Kaiser Optical Systems Inc., Ann Arbor, MI, USA) with an excitation wavelength of nm 785 λ= . The incident power of the laser emission was set to 12.5 mW at the probe head. Figure 2 shows thereby schematically the experimental setup. A second measurement run was performed using a portable Raman spectrometer i-RamanPro (B&W Tek, Inc. Model: BWS475-785S) with an operating wavelength of λ= nm 785 and a laser power of mW 12 5 . at the probe head. Analysis was performed using ORIGIN Pro software. ## Measurement routine. To ensure that BTAH is uniformly distributed over the active area of the substrates and to account for inhomogeneities of the substrate, a measurement routine involving point measurements on five different positions on the substrate was used. To achieve this, the substrate was positioned on a x,y,z-positioning stage with sub-µ m accuracy. The distance to the excitation (z-axis) was regulated on maximum signal intensity of the bare substrate. Each spectrum was taken with 10 s acquisition time and three accumulations www.nature.com/scientificreports www.nature.com/scientificreports/ to enhance signal-to-noise ratio. The acquired spectra of all five measurement points were then averaged to a single spectrum. For the concentration dependent measurements, the substrates were laid in . ml 0 8 of BTAH solution with a specific concentration for 15 min. After taking the sample out, we waited until the residual liquid evaporated (approximately 5 min) before starting the Raman measurements. The time-resolved measurements were performed on a single point. The acquisition time was reduced to s 3 with two accumulations leading to a total measurement time of s 30 for each time step. For this measurement, the substrate was placed in a glass beaker with 5 ml Milli-Q water. Then ml 0 8 . BTAH were added (concentration mg l 10 / ) to the water and a spectrum was taken every 30s. At the end, a negative measurement was performed using a flat gold mirror. For real wastewater samples, a concentration procedure was deemed necessary. For this, ml 20 of the sample were evaporated to ml 10 ( ml 5 ) achieving a doubling (quadruplicating) of the concentration. The measurements were then performed on the residue. ## Results and Discussion concentration dependent measurements. First, we performed a concentration dependent measurement series to determine the detection limit of the BTAH on the SERS-active substrates. For this we used our C7 substrates, which are described in the SERS-Substrates section. The measurement routine is as described in the experimental section. Figure 3 shows the experimental results for this substrate. The concentration series starts from the pure BTAH signal (a) reducing to the bare substrate (e), the latter giving the background signal for a better analysis. The spectra are shown in Fig. 3 (top). It is apparent that not all vibrational modes of BTAH visible in the pure sample can be taken for identification purposes, because of an overlap with bands of the substrate. Therefore, the analysis concentrates on the bands with a strong response at − cm 783 1 and at cm 1387 1 − , representing the benzene breathing mode and the triazole ring stretching mode, respectively 21 . Thereby, the benzene breathing mode has been used as reference peak for the determination of the detection limit. Decreasing the concentration, the peak height as well as its integrated area are decreasing. The latter value has been used to quantify the detectability, as it is more robust against noise or thermal drifts of the center frequency of the mode. The peak is thereby fitted with a Lorentzian profile after a background correction (ALS, asymmetric least squares), assuming no coupling between adjacent vibrational modes or between a mode and the background continuum, the latter justified because no Fano-like shape of the peak can be observed. The result of the analysis is shown in Fig. 3 (bottom). It is visible that an increase of the BTAH concentration above mg l 100 / does not lead to a further increase in peak area. In fact, there is a saturation behavior indicating that the BTAH layer is thick enough to mask any signal originating from the substrate. The detection limit determined with this method lies considerably beneath mg l 0 10 / . . The error margin of the analyzed concentrations lies in between 1 % and 2 %, which would result in error bars smaller than the symbols of the data points and therefore are not visible in Fig. 3 (bottom). With these results, the measurement routine was directly transferred to an examination of real wastewater samples. time-resolved measurements. After determining the detection limit for the pure BTAH in the previous section, the question still to be answered is how the BTAH is adsorbing on the SERS-substrates and how this process could be possibly described, i.e. Langmuir or Freundlich isothermal behavior. This is possible because SERS can detect the surface coverage at ultralow concentrations 20 . We, therefore, designed our experiment such that we could monitor the adsorption over a long period of time (60 min). The acquired spectra were again analyzed by fitting the width of the benzene breathing mode at cm 783 1 − . Figure 4 shows the resulting graph. One can see a monotonic increase of the peak intensity up to 48 min. Then the intensity equals a plateau indicating a saturated surface coverage masking again signals resulting from the SERS-substrates. From the experiment another critical time can be deduced. It is the minimal time necessary to be able to detect the BTAH. As shown exemplarily in the spectra of Fig. 4, the minimum exposure time lies at around 15 min, here the benzene breathing mode is clearly distinguishable from the background. From this point on an identification of the analyte on the SERS substrate becomes unambiguous and can be used as a guide for field applications on www.nature.com/scientificreports www.nature.com/scientificreports/ how long one has to wait before measuring the sample. In this particular case we achieved a signal enhancement of 75 % in waiting from 15 to 48 min. The recorded kinetic (temporal) behavior was fitted using a Hill-Langmuir description (Hill1, which is a modified Hill function with offset) 25 . The parameters used are as follows: START = Start area, END = End area, k = Michaelis constant, and n = Cooperative sites, respectively. The following applies for the bounds used: Lower Bounds: > k 0 and n 0 > , Upper bounds: none. This is an equivalent description, as used in enzyme kinetics, for a sigmoidal growth and therefore for the thermodynamic isotherms described by Langmuir or Freundlich. Our findings point, thereby, towards a non-layer-by-layer growth of the analyte at the substrate surface. Thus, most likely a Freundlich thermodynamic description can be used to understand the coverage process. Real wastewater samples. After determining the detection limit for the pure BTAH and understanding the coverage dynamics as described in the previous section, we performed a concentration dependent measurement series on a real wastewater sample. This is necessary for an evaluation of the substrates for environmental application use. Wastewater is thereby a complex matrix, which will mask the expected Raman peaks of the BTAH due to the coverage of the SERS-substrates with all kind of different analytes. The real wastewater measurements were conducted on all SERS-substrates and both equipments. In this case the triazole ring stretching mode at cm 1387 1 − has been used as reference peak for the determination of the detection limit because it is the more prominent one in the wastewater samples. Here, we show exemplary spectra revealing the detection potential of different samples and also of the portable process Raman instrument (Figs 5 and 6). The additional set of measured spectra is shown in Figs. S1-S4 in the supplementary information. Figures 5 and 6 show, that both tested substrates enable the detection of BTAH directly in the wastewater at a level of µ . g l 17 6 / when concentrating the BTAH in a pre-treatment step. The experiments showed that the portable process Raman spectrometer was also able to detect the BTAH after a pre-concentration step. The detection limits in this case resemble the results of the laboratory equipment. As can be seen from the recorded Raman spectra, there are differences in the detectability of BTAH on the different SERS-substrates. The results for the different substrates measured with the different spectrometers are summarized in Tables 1 and 2. One major finding can be extracted from the experiments. The positive results are all measured with substrates having a periodic nanosized structure (C7 and SilAg). This might be due to the deterministic structure, covering the whole substrate area giving rise to a collective enhancement, since each nanopillar enhances at its tip the same way as the neighboring ones. The RandaS substrates having a stochastic surface seem not to exhibit this effect. Furthermore, it seems that silver gives rise to a stronger enhancement factors compared to gold-surfaces, resulting in detectability of BTAH. Silver is known to be a stronger plasmonic material than gold, so this might be the reason for the better performance. Yet, this raises the question how chips will behave, if materials are adapted. Moreover, with more biopharmaceutical applications to be realized, can Ag-chips keep up with Au-surfaces in terms of biocompatibility or functionalization in the end? Finally, the results on the real wastewater samples emphasize the fact that the nanostructured samples perform superior. www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, we were interested in the amount of BTAH being detected in the laser spot itself to translate the given results into a more differentiated detection limit. Therefore, we calculated the surface area within the laser spot of m 125 µ in diameter, resulting in an area of mm 0 0460 2 . , acknowledging the pillar structure. We estimated the density ratio of a single BTAH molecule as a rectangle with the parameters as displayed in Fig. 1(c), resulting from the dimensions of the molecule. A monolayer of BTAH within the illuminated area would therefore translate to a mass of . pg 33 3 as detection limit. This amount of BTAH, which has been measured both with our sophisticated as well as with commercially available substrates is of comparable magnitude as reported elsewhere, e.g. by Altun et al. 20 , showing the possibility to use these SERS-active surfaces in environmental analysis. However, it is shown that BTAH is not limited to form strictly monolayers, see kinetic measurements, leaving the actual value to be possibly larger 1 . conclusion Benzotriazole (BTAH) is one of the best corrosion inhibitors and commonly used as an antifreeze. Since its problematical environmental effects and decent solubility qualities in water, it represents a threat to the environment and thus also to human health. It is therefore of uttermost importance to have a detection method capable of monitoring the concentration of BTAH at trace level on-site. Here, we have shown that a sensing method based on surface-enhanced Raman spectroscopy is capable of detecting trace-level concentrations of BTAH in actual wastewater samples. This has been possible with our self-made as well as with commercially available substrates and the use of a portable Raman spectrometer. There has only been a pre-treatment step necessary to achieve a detection limit of µ . g l 17 6 / . The conducted experiments demonstrate that it is possible to detect very small amounts of analyte substances other than very small , respectively. (e) Pure SilAg substrate. www.nature.com/scientificreports www.nature.com/scientificreports/ concentrations of that substance in solution. Enabling a direct detection of BTAH at a g µ and sub-µ g scale is still a challenging task for current research, yet with our results, the task seems achievable with an equipment already available and a small measuring time (below 48 min) rendering the technique interesting for on-site detection of this pollutant. We have calculated a theoretical detection limit of a monolayer of BTAH in the laser spot at the substrate surface of 33.3 pg, depending on the actual adsorption of BTAH. Furthermore, we have carefully studied the concentration behavior and modelled the concentration dependency and time-dependent coverage, showing that it can be described by Freundlich isotherms.

chemsum

{"title": "Evaluation of an on-site surface enhanced Raman scattering sensor for benzotriazole", "journal": "Scientific Reports - Nature"}

uncertainty-informed_deep_transfer_learning_of_pfas_toxicity

## Abstract: Perfluoroalkyl and polyfluoroalkyl substances (PFAS) pose a significant hazard because of their widespread industrial uses, environmental persistence, and bioaccumulativity. A growing, increasingly diverse inventory of PFAS, including 8,163 chemicals, has recently been updated by the U.S. Environmental Protection Agency. But, with the exception of a handful of wellstudied examples, little is known about their human toxicity potential because of the substantial resources required for in vivo toxicity experiments. We tackle the problem of expensive in vivo experiments by evaluating multiple machine learning (ML) methods including random forests, ## Introduction Perfluoroalkyl and polyfluoroalkyl substances (PFAS) encompass thousands of synthetic fluorinated aliphatic compounds 1,2 . PFAS pose a significant challenge of increasing concern because of their widespread presence, long-term persistence, extended biological half-lives (approaching nine years for some), and largely unknown toxicities. PFAS use has been identified at more than 400 U.S. military bases, and contamination has been found in the drinking-water systems of more than two dozen military sites. U.S. cleanup costs are estimated to be tens of billions of dollars, including $2 billion for the Department of Defense alone 3 . The U.S. Environmental Protection Agency (EPA)'s Distributed Structure-Searchable Toxicity (DSSTox) database of PFAS structures 4 , as recently updated, contains over 8,163 PFAS chemicals. PFAS compounds can be broadly classified into polymeric and non-polymeric families 2 . This study addresses non-polymeric PFAS, which have a higher propensity to be absorbed via the digestive system, creating an urgent need to understand their toxicities. Their toxicities will be important determinants of target cleanup levels and associated costs as well as identification of non-toxic substitutes for future consumer products. Traditional approaches for generating toxicity information (e.g., human epidemiological and experimental animal studies) are resource-intensive, and only limited studies have been conducted across this large set of compounds 5,6 . The exponential growth of chemical synthesis in recent decades necessitates scalable approaches for determination of PFAS toxicities. To reduce the expense and uncertainties inherent in animal experiments, it is crucial to perform high-throughput computational toxicity predictions. Here we explore a cheminformatics approach to predicting and understanding toxicity from chemical structure. The acceleration of computational toxicology in recent years can be attributed to 1) the development of large databases of chemical toxicities, 2) increased computing power with the advent of hardware such as Graphic Processing Units and other accelerators, and 3) advancement in machine learning (ML) that can take advantage of increased data and computational power . In particular, deep learning for prediction of chemical properties is becoming increasingly relevant 12,13 . Several studies have demonstrated that deep-learning models for chemical properties and toxicity prediction can outperform traditional Quantitative Structure-Activity Relationship (QSAR) approaches such as naive Bayes, support vector machines, and random forests (RFs) . However, a compound's toxicity is affected by multiple chemical and biological factors, adding complexity to the prediction of this crucial property 19 . following exposure of a living organism. This broad definition means that there are many considerations when characterizing acute toxicity. A common nonspecific method for gauging the relative toxicity of a set of compounds without any considerations of biological pathways involved is to compare median lethal doses (LD 50 ), the minimum dose of a substance shown to cause fatality in 50% of laboratory subjects within 24 hours after initial oral or dermal exposure. Oral rat LD 50 metrics are measured in test-substance quantity per unit mass of laboratory-rat body weight and are ranked by the EPA into four categories: I (high toxicity), II (moderate toxicity), III (low toxicity), and IV (very low toxicity). Acute oral toxicities and their respective EPA categories (defined in where oral rat LD 50 labels are available, before attempting predictions for PFAS compounds with unknown oral rat LD 50 . For this purpose, we review the uncertainty analysis derived from transferlearned target models to gain insights into the quality of predictions for the PFAS-like compounds. Finally, we temper toxicity predictions by implementing selective prediction through an abstention mechanism that forces our transfer learned target model to say "I cannot answer" when confidence in a prediction is low . When making predictions for compounds with unknown toxicity, it is extremely important to enforce an abstention mechanism as a precautionary measure against incorrectly classifying a highly toxic substance. We then apply the transfer-learned selective model to predict toxicity (or abstain) for 8,163 EPA PFAS compounds with no known oral rat LD 50 ; the details of these predictions are discussed in the results section. The added capability of a transferlearning model to abstain from prediction for some compounds opens up the possibility of creating a direct feedback loop to in vivo experiments, the details of which are further discussed in the conclusion section. We refer to the entire suite of computational toxicology tools developed as part of this study as "AI4PFAS" (Figure 1); details are discussed in the methods and results sections. ## Datasets The availability of in vivo acute oral toxicity measurements for PFAS is limited to a handful of wellstudied compounds in this family. To abate the lack of PFAS toxicity data, we constructed an expanded dataset, LDToxDB, of 13,329 unique compounds of any type with oral rat LD 50 measurements aggregated from the EPA Toxicity Estimation Software Tool (TEST), NIH Collaborative Acute Toxicity Modeling Suite (CATMoS), and National Toxicology Program datasets . LD 50 point estimates provided in mg/kg were converted to units of -log(mol/kg) to reflect per-molecule toxicity irrespective of molecular mass; the resulting histogram is shown in Figure 2. Most of these compounds were labeled as EPA toxicity class III, followed by a near-equal presence in II and IV, and lastly class I. SMILES were canonicalized using RDKit 27 and duplicate molecules were removed by querying each compound's hashed InChlKey. To broadly identify PFAS-like compounds within LDToxDB, molecules with two or more C -F bonds were identified by using an RDKit SMARTS query and tagged as a PFAS-like representative subset of the labeled LDToxDB compounds. Such compounds with 2 or more C-F bonds would be polyfluorinated, likely alkyl, but may not be designated as PFAS in various databases. The resulting 518 compounds, referred to as "LDToxDB-PFAS-like" and which include 58 compounds formally labeled as PFAS, served as an important validation group to confirm that models trained on LDToxDB were able to predict toxicity of PFAS and PFAS-like compounds via chemical-structure similarity. Finally, 8,163 PFAS compounds were extracted from the EPA DSSTox database 4 , most of which have no LD 50 ; referred to as "LDToxDB-PFAS"; and reserved for prediction. We note that 58 of these are also represented, with labels, in LDToxDB-PFAS-like. The dataset (composed of LDToxDB, LDToxDB-PFAS-like, and LDToxDB-PFAS) and the Python processing codes used to parse the data and construct the models are available at https://github.com/AI4PFAS/AI4PFAS. ## Chemical Featurization Chemical featurization is the process of translating chemical attributes associated with a compound into machine-readable numeric features. We computed features for all compounds in LDToxDB for use in supervised ML of acute oral LD 50 point estimates. Further, unsupervised ML can be applied to the chemical features in order to deduce chemical insights. Figure 1 shows the full set of featurizations and tools developed as a part of the AI4PFAS workflow; details are discussed below. Chemical featurization relies primarily on encoding structural features and atom identities within a molecule. Three types of chemical featurization were considered in this study: 1) Mordred descriptors 30 : We used the Mordred software package 30 to generate 1,800 unique molecular descriptors for each compound directly from RDKit molecules. Mordred provides quick featurization of a molecular dataset by generating a vast array of two-and three-dimensional descriptor characteristics from SMILES input. The full reference list of Mordred descriptors is available elsewhere 30 . We trimmed down the 1,800 descriptors to 300 by using Pearson correlation coefficient (PCC) analysis to remove redundant features. 2) Extended-connectivity fingerprinting (ECFP) 31 provides a mechanism for representing topological chemical space within a fixed-length bit string by iteratively measuring substructure connectivity at a provided radius around each atom. Numeric representations are created for each substructure identified in these iterations and then combined into a fixed-length bit string. Conventionally, an ECFP is described by its bit length and the maximum radius used for substructural querying: thus, for example, a 2048-bit ECFP4 has a length of 2048 bits and a maximum radius of four. Multiple bit lengths and radii were used for different purposes in this study. ECFPs are generated by using the open-source RDKit package for Python 27 . 3) Molecular graph encoding 32,33 improves on ECFP by representing molecules as graphs of arbitrary size with nodes representing atoms, and edges representing bonds. Each entity is given characteristic traits, which for nodes may include (but are not limited to) atomic identity, number of valence electrons, formal charge, and hybridization, and for edges, bond order and conjugation status. We have adopted the graph representation and the corresponding graph convolutional neural network from the MOLAN workflow 17 . 4) Non-negative matrix factorization (NMF) 34,35 is a dimensionality reduction technique that derives basis vectors under a non-negative constraint. We performed dimensionality reduction on ECFP to derive rich low-dimensional features. A 12-dimensional representation is found to be optimal. ## Supervised machine learning We used the following supervised ML methods to establish a baseline for the acute oral LD 50 prediction: 1) Random forest regressor 36 : This ensemble prediction method generates a specified number of decision trees, each based on randomly initialized conditional thresholds for filtering input values. RF models provide a consensus prediction from these decision trees. This is a shallow-learning strategy, since there is no propagation algorithm or loss function with which to adjust weights 37 . The RF regression was performed using Scikit-learn and independently featurized by ECFP, NMFreduced ECFP and Mordred descriptors 37 . 2) Gaussian process (GP) regression 38 : This method statistically models a prediction space by constructing a joint distribution from the multivariate normal distributions of input combination pairs. We used GP approximation as the basis for a predictive model where inputs were independently featurized by 2048-bit ECFP4 and Mordred descriptors. To reduce training cost, 200 important ECFP bits and 10 important Mordred descriptors were chosen from the RF Gini feature importance. The training was performed using the GPflow package. 39 3) Deep neural network 12,13 : Artificial neurons form the basis of a deep neural network (DNN). Composed of a linear unit and a non-linear activation function, neurons are stacked into sequential layers where each receives as input the output from all neurons in the preceding layer. Together, these layers form a multilayer perceptron (MLP The performance of all the supervised ML methods was evaluated by Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and coefficient of determination (R 2 ). We use two methods for partitioning our data into 80% training and 20% testing sets: (1) random sampling and (2) stratified sampling on binned LD 50 measurements. A five-fold cross-validation is employed with a random seed to ensure consistent data splits and minimize overfit bias in performance evaluations. Bayesian optimization is a powerful technique for finding optimal hyperparameters for black-box functions 41 . The hyperparameters of all supervised ML methods (DNN-Mordred, DNN-ECFP, GCN, and RF) are tuned using Bayesian optimization as implemented in the GPyOpt library 42 ; parameter bounds are in Table S1. ## Transfer learning In ML, repurposing knowledge from source domains for use within a target space is a powerful application of the transfer-learning concept. Low-dimensional knowledge is shared across domains and high-dimensional knowledge is trained from the basis of common understanding. This is done in practice by initially optimizing the MLP within the source domain. Prior to training on target data, the learning rates of upstream neurons are reduced relative to later ones in order to fix early neurons used for low-dimensional feature discrimination. In certain cases, no learning is allowed (i.e., the learning rate is set to zero), a process referred to as freezing. Downstream neurons may be reinitialized to random weights, and layers may be added. Training is then repeated within the target space, and success is indicated by positive transfer 20 . ## Uncertainty quantification Two approaches to uncertainty were examined. The first approach, deep ensemble, employs an ensemble of deep-learning models, each using a fixed neural network architecture with different randomly initialized layer weights (prior to training) to get multiple point estimates of prediction 43 . The variance derived from the point estimates serves as an approximation of uncertainty. The second method, a latent-space approach, relies on the distance of a prediction point to neighboring training points in the embedded space of the final hidden layer of the neural network 44 ## Learning with abstention Selective Prediction model : ML practitioners can use uncertainty associated with individual predictions to judge their quality. In particular, predictions with high uncertainty (i.e., low confidence) could be discounted by the human practitioner. On the other hand, a standard supervised ML approach always produces an answer, even for scenarios far outside of the training region, where such models are expected to perform poorly. Hence there is a need for an Artificial Intelligence (AI) that can replicate the human-like decision to say "I can't answer" for low-confidence/high-risk scenarios. Selective prediction is a ML paradigm where the goal is to learn a prediction model that knows when it does not know. A selective prediction model performs "learning with abstention" on its own. The selective prediction model is learned jointly as a pair (f, g), where f is a prediction function and g is a selection function which learns whether f should be allowed to predict or abstain, as described below: (1) 𝑖𝑓 𝑔(𝑥) ≥ 𝜏 𝐷𝑜𝑛 ′ 𝑡 𝑘𝑛𝑜𝑤, 𝑂𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 where , the input chemical feature space, and the tolerance . 𝑥 ∈ 𝑋 𝜏 ∈ (0,1) In particular, we use the SelectiveNet-based selective prediction model in this study 23 . This model architecture offers easy conversion of the main body block from a reference neural network into a corresponding network with a reject option, as illustrated in Figure 3. In a SelectiveNet, the representation (last) layer will be processed by three heads (as shown in Figure 3c): 1) A prediction head (f(x)) for LD 50, 2) a selection head (g(x)), a classifier that decides whether the model should abstain or not, and 3) an auxiliary head (h(x)) that enriches the representation layer. The joint loss for (1), given k labeled samples ( is written as 𝑆 𝑘 ), where is the hyperparameter that controls the coupling to the squared penalty function and c is the 𝜆 target coverage. The empirical coverage, , is computed as the mean of the selection function output 𝜙 for the k input samples. The empirical selective risk, r, is defined as where l(f(x i ), y i ) is the regressor loss for the prediction head. Finally, the overall loss for SelectiveNet is written as the combination of (2) and the auxiliary head loss ( , with α = 0.5 used in this work: An optimal selective prediction model is arrived at by optimizing the selective risk with respect to the coverage. This is done by converging the risk-coverage curve and selecting a coverage that results in minimal selective risk 45 . The choice of the hyperparameter and the neural network architecture as shown in Figure 3c are further discussed in the results. Hence we now transform the DNN-Mordred from a) as the main body into a SelectiveNet architecture. The SelectiveNet architecture adds two more output heads, corresponding to an auxiliary and a decision head, for selective prediction. In this workflow, the SelectiveNet is transfer-learned using the same method used in b), except that here, uncertainty per prediction for the PFAS with unknown toxicity can be automatically converted into a decision (i.e., predict or abstain) by learning with abstention. Results are organized into four subsections, each building towards the final objective of predicting the toxicity for 8,163 PFAS compounds with abstention. We first provide a review of current literature on oral rat LD 50 prediction, followed by results from our baseline ML benchmark models for LDToxDB. We then discuss transfer learning for the best-performing ML model on LDToxDB as the source task and LDToxDB-PFAS-like as the target task. We go on to show the benefits of uncertainty analysis for the target prediction task. Finally, we discuss the results of the SelectiveNet in predicting (or abstaining from) toxicity for 8,163 compounds from the EPA's PFAS structure list, most with no known LD 50 labels. ## Model Baselines Literature baselines for ML-based LD 50 predictions are presented in Table 2, with experiment sample size, methodology, and performance metrics for the top-performing model from each study. While variability in training datasets and testing protocols prevents a direct comparison, the bestperforming models use state-of-the-art ML based on DNN. In particular, Xu et al. 46 The benchmark results from this study for the prediction of LDToxDB are presented in Table 3. Random sampling was found to give better performance compared to stratified sampling (shown in Figure S1) and is used for each model. Reported metrics (R 2 , MAE, RMSE, and accuracy) represent average metrics computed across each testing fold for every model (i.e., five-fold cross-validation). Accuracies are provided as a supplemental metric, calculated by taking each compound's predicted LD 50 , converting it to mg/kg, and labeling with EPA toxicity categories. Since models used in this study are regressors, accuracies are expected to underperform in comparison to classification models present in the literature, and are hence not intended for a direct comparison with literature baselines. From Table 3, it is observed that the ML models evaluated in this study perform in the following order, evaluating each model by the reported MAE: DNN-Mordred < RF-Mordred < GP < GCN < DNN-ECFP < RF-ECFP < RF-NMF. These results suggest that DNN with Mordred descriptors input outperforms other models with an R 2 of 0.65. While variations in datasets prevent direct oneto-one comparison to Table 2, our DNN-Mordred model yields similar performance to that reported by Zhu et al. 50 and Gadaleta et al. 26 , justifying the evaluation of these models when further developed for the PFAS domain. ## Transfer Learning on LDToxDB-PFAS-like We next demonstrate how a DNN-Mordred model trained as the source task can be used to perform knowledge transfer within the PFAS domain. Since only 58 PFAS compounds are available with reported values for oral rat LD 50 , we use the broader 518 LDToxDB-PFAS-like subset as a measure of whether transfer learning has a beneficial outcome. The outcome of transfer learning is directly measured by the extent to which positive transfer occurred (i.e., no performance degradation after transferring knowledge from source to target task) 20 . For our comparison, we collect MAE and R 2 metrics from models within the target domain both before and after transfer learning. The transfer step involves freezing early layers trained within the source domain and reinitializing later layers to re-train within the target domain (illustrated in Figure 3b). We refer to this model setup as "Transfer-DNN-Mordred." Freezing all hidden layers of DNN-Mordred and retraining the output linear layer was found to be optimal (see Figure S2). The top panel of Figure 4 shows the performance effect of transfer learning on DNN-Mordred outcomes within LDToxDB-PFAS-like. Transfer-DNN-Mordred showed positive transfer, as seen by the marginal decrease in error and increase in R 2 when looking at regression predictions, affording greater stability in the target domain. The results are also further converted to EPA categories (this convention will be followed through the rest of the article for a direct comparison with EPA toxicity classes). While category IV accuracy is notably hampered, decreasing from 19.1% to 14.7%, the accuracy of category III compounds (the largest represented group in the LDToxDB) improves from 75.9% to 76.8% and overall predictive capacity improves, as seen from a stronger R 2 score. ## Uncertainty Quantification and Limitations With established evidence of positive transfer in "Transfer-DNN-Mordred" (Figure 3b), we turn our attention to calculating uncertainty per prediction. In practical settings where toxicity modeling provides consequential utility, uncertainty enables knowledgeable practitioners to discount spurious predictions. Uncertainty is evaluated here as the ability of the chosen metric to capture the model error; in other words, a suitable measure for evaluating the efficacy of an uncertainty metric is the correlation of uncertainty with model error. We evaluate two approximations for model uncertainty, 1) deep ensemble and 2) latent space distance, and analyze the best-performing mechanism within the context of our validation set. Literature on deep ensembles has shown that an ensemble model size as small as five is sufficient 51 . We evaluated the convergence of ensemble model size (Figure S3) and found that 10 DNN-Mordred models were sufficient for our purpose. To use latent space distances as a measure of uncertainty, we used a UMAP model on training-data latent space outcomes 28 . The Euclidean distance between the latent space of inference and the nearest training point was used. PCCs grouped by superclass are provided in 29 to provide granularity in assessing the PCC performance. Only superclasses with greater than 10 substituents are shown. The singleton subclasses column provides the number of single-member subchemical classes that are present in the corresponding superclass. 95% confidence intervals (CI) representing a probabilistic forecast of the true mean of Transfer-DNN-Mordred predictions for each compound. Note that when experimental values fall outside the 95% CI, it simply means that the variance across a sample of DNN models is not high enough to accurately capture confidence with respect to the true value. We observe from Figure 5 The 40.1% of validation compounds that fall outside the 95% CI of the population mean of DNN-Mordred predictions invoke the larger deep-learning problem of overconfidence 52 . The deepensemble uncertainty fails when multiple models share a similar incorrect explanation across input space. This is an active area of research with no universal solution 51 . Thus, for predicting unlabeled data with a probable shift from our training set (despite efforts to isolate and transfer-learn on "PFAS-like" chemicals), we turn to an alternative in the next section: selective prediction. Using the uncertainty quantification capacity that our model does have (demonstrated on 59.9% of validation compounds), we employ a model with the means of abstaining from prediction. In practice, this approach means more cautious predictions on unlabeled data and a prioritization framework for moving forward with in vivo experimental trials. ## Predicting PFAS compounds In this section, we discuss predicting toxicities for unlabeled PFAS chemicals. The ensemble approach discussed in the previous section works intuitively when a clear ensemble standard deviation threshold can be used to designate compounds with high uncertainty. The definition of such a domain-dependent threshold would require some human supervision. Further, the deepensemble predictions can become overconfident. The prediction of a larger, unlabeled set of PFAS chemicals introduces new considerations: Can we design an AI that can understand uncertainty per prediction (when labels are not available for comparisons) and decide whether it should predict or say "I cannot predict?" Can we include an in-built safety feature in a neural network so as to minimize or avoid a catastrophic scenario? (Such a catastrophic scenario may entail a model predicting a compound as belonging to EPA class IV whereas in reality it is a highly toxic compound belonging to EPA class I.) With these considerations for prediction of PFAS compounds, the SelectiveNet architecture was implemented with DNN-Mordred operating as the main body of the neural network (referred to as SelectiveNet Transfer DNN-Mordred, shortened to "SN-Mordred"; see Figure 3c). Transfer learning was performed as described earlier. The optimal SN-Mordred is arrived at by minimizing the risk by constraining the coverage 23 . Multiple models were trained, corresponding to coverage thresholds varying between C=0.5 and C=1.0. The selective heads of trained models were then calibrated within their respective validation sets, as recommended by Geifman et al. 23 , and the total empirical risk was calculated with respect to coverage (see Table S2 in SI). A coverage threshold of 0.6 was found optimal and used to calibrate the abstention mechanism for use on LDToxDB-PFAS. Featurization of LDToxDB-PFAS by Mordred descriptors was successful for 7,058 compounds. Figure 6a shows the distribution of selective-prediction outcomes. Since the SelectiveNet was trained with a coverage of 60%, we examine where SN-Mordred abstains by breaking down the EPA PFAS structure list by chemical superclass annotated by the ClassyFire server 29 (Figure 6a inset). After inference, 43 predictions are excluded as out-of-range (using water and the most toxic compound in LDToxDB as boundary limits). The dominant four superclasses within LDToxDB-PFAS (Organohalogens, benzenoids, organic acids, and benzenoids) are all predicted at a rate of approximately 75% with no trend of favorably pruning certain chemical superclasses. The most represented EPA class in LDToxDB is EPA class III (Figure 2). Consequently, it can be observed that SN-Mordred is most confident in predicting EPA class III (Figure 6b). SN-Mordred only predicted seven compounds to be in EPA class I, demonstrating considerable caution with respect to the most toxic EPA class. To underpin the results and decisions returned using AI, future efforts could include the development of deep learning or QSAR models using molecular descriptors strongly correlated with acute toxicity 49 or by building local QSAR models from closely similar structures 54 . Such efforts would provide a physical and mechanical basis grounded in molecular structure for interpreting toxicity estimates from AI. The derived relationships could further reduce the incidence of catastrophic decisions from AI predictions. ## Conclusions a rigorous ML-based computational toxicology workflow that we use to predict toxicity for ~8,163 PFAS compounds whose toxicities are poorly understood. We achieve this result by transfer learning on knowledge of all organic compounds with known values of oral rat LD 50 to predict on the PFAS compound space with informed uncertainties. Learning by abstention provides an automatic mechanism for converting uncertainty per prediction into model decisions. Organ-on-a-chip systems, now possible through advancements in microfluidic technologies, have allowed for the emulation of in vivo physiological conditions 55,56 . The selective prediction model can be used for deriving decisions on compounds whose toxicity values cannot be predicted reliably. The model decisions can be used to drive on-demand active learning of toxicology experiments using the organon-a-chip setup. In this age of big data, neural networks have been widely adopted for applications in the chemical sciences community. We anticipate that the AI4PFAS workflow can be used for predicting many other toxic endpoints. Some of the approaches presented in this study can be used to add a layer of safety to supervised ML predictions, especially in mission-critical applications such as computational toxicology. We hope that some of the good practices presented in this study are adopted and expanded on by the wider chemical sciences community.

chemsum

{"title": "Uncertainty-Informed Deep Transfer Learning of PFAS Toxicity", "journal": "ChemRxiv"}

drugex_v2:_de_novo_design_of_drug_molecule_by_paretobased_multi-objective_reinforcement_learning_in_

## Abstract: In polypharmacology, ideal drugs are required to bind to multiple specific targets to enhance efficacy or to reduce resistance formation. Although deep learning has achieved breakthrough in drug discovery, most of its applications only focus on a single drug target to generate drug-like active molecules in spite of the reality that drug molecules often interact with more than one target which can have desired (polypharmacology) or undesired (toxicity) effects. In a previous study we proposed a new method named DrugEx that integrates an exploration strategy into RNN-based reinforcement learning to improve the diversity of the generated molecules. Here, we extended our DrugEx algorithm with multi-objective optimization to generate drug molecules towards more than one specific target (two adenosine receptors, A1AR and A2AAR, and the potassium ion channel hERG in this study). In our model, we applied an RNN as the agent and machine learning predictors as the environment, both of which were pre-trained in advance and then interplayed under the reinforcement learning framework. The concept of evolutionary algorithms was merged into our method such that crossover and mutation operations were implemented by the same deep learning model as the agent. During the training loop, the agent generates a batch of SMILESbased molecules. Subsequently scores for all objectives provided by the environment are used for constructing Pareto ranks of the generated molecules with non-dominated sorting and Tanimoto-based crowding distance algorithms. Here, we adopted GPU acceleration to speed up the process of Pareto optimization. The final reward of each molecule is calculated based on the Pareto ranking with the ranking selection algorithm.The agent is trained under the guidance of the reward to make sure it can generate more desired molecules after convergence of the training process. All in all we demonstrate generation of compounds with a diverse predicted selectivity profile toward multiple targets, offering the potential of high efficacy and lower toxicity. ## Introduction The 'one drug, one target, one disease' paradigm, which has dominated the field of drug discovery for many years, has made great contributions to drug development and the understanding of their molecular mechanisms of action . However, this strategy is encountering problems due to the intrinsic promiscuity of drug molecules, i.e. recent studies showed that one drug molecule could interact with six protein targets on average . Side effects of drugs caused by binding to unexpected off-targets are one of the main reasons of clinical failure of drug candidates and even withdrawal of FDAapproved novel drugs . Up to now, more than 500 drugs have been withdrawn from the market due to fatal toxicity . Yet, disease often results from the perturbation of biological systems by multiple genetic and/or environmental factors, thus complex diseases are more likely to require treatment through modulating multiple targets simultaneously. Therefore, it is crucial to shift the drug discovery paradigm to "polypharmacology" for many complex diseases . In polypharmacology, ideal drugs are required to bind to multiple specific targets to enhance efficacy or to reduce resistance formation (in which case multiple targets can be multiple mutants of a single target) . It has been shown that partial inhibition of a small number of targets can be more efficient than the complete inhibition of a single target, especially for complex and multifactorial diseases . In parallel, common structural and functional similarity of proteins results in drugs binding to off-targets; therefore we also demand drugs to have a high target selectivity to avoid binding to unwanted target proteins. For example, the adenosine receptors (ARs) are a class of rhodopsin-like G protein-coupled receptors (GPCRs) having adenosine as the endogenous ligand. Adenosine and ARs are ubiquitously distributed throughout the human tissues, and their interactions trigger a wide spectrum of physiological and pathological functions. There are four subtypes of ARs, including A1, A2A, A2B and A3, each of which has a unique pharmacological profile, tissue distribution, and effector coupling . The complexity of adenosine signaling and the widespread distribution of ARs have always given rise to challenges in developing target-specific drugs . In addition to the similarity of the pharmacophores of some generic proteins (e.g. human Ether-à -go-go-Related Gene, hERG) should also be taken into consideration as they can be sensitive to binding exogenous ligands and cause side effects. hERG is the alpha subunit of a potassium ion channel and has an inclination to interact with drug molecules because of its larger inner vestibule as the ligand binding pocket . When hERG is inhibited this may cause long QT syndrome . In addition to visual recognition, natural language processing and gaming, deep learning has been increasingly applied in drug discovery . It does not only perform well in prediction models for virtual screening, but is also used to construct generative models for drug de novo design and/or drug optimization . For example, our group implemented a fully-connected deep neural network (DNN) to construct a proteochemometric model (PCM) with all high quality ChEMBL data for prediction of ligand bioactivity . Its performance was shown to be better than other shallow machine learning methods. Moreover, we also developed a generative model with recurrent neural networks (RNNs), named DrugEx for SMILES-based de novo drug design . It was shown that the generated molecules had large diversity and were similar to known ligands to some extent to make sure that reliable and diverse drug candidates can be designed. Since the first version of DrugEx (v1) demonstrated effectiveness for designing novel A2AAR ligands, we began to extend this method for drug design toward multiple targets. In this study, we updated DrugEx to the second version (v2) through merging crossover and mutation operations, which were derived from evolutionary algorithms, into the reinforcement learning (RL) framework. In order to evaluate the performance of our additions we tested our method into both multi-target and target-specific cases. For the multi-target case, desired molecules should have a high affinity towards both A1AR and A2AAR. In the target-specific case, on the other hand, we required molecules to have only high affinity towards the A2AAR but a low affinity to the A1AR for. In order to decrease toxicity and adverse events, molecules were additionally obliged to have a low affinity for hERG in both cases. It is worth noting that generated molecules should also be chemically diverse and have similar physico-chemical properties to known ligands. ## Data Source Drug like molecules represented as SMILES format were downloaded from the ChEMBL database (version 26). After data preprocessing, including recombining charges, removing metals and small fragments, we collected 1.7 million molecules and named it the ChEMBL set, used for SMILES syntax learning. This data preprocessing step was implemented in RDKit . Furthermore, 25,731 ligands were extracted from the ChEMBL database to construct the LIGAND set, which had bioactivity measurements towards the human A1AR, A2AAR, and hERG. The LIGAND set was used for constructing prediction models for each target and fine-tuning the generative models. The number of ligands and bioactivities for these three targets in the LIGAND set is represented in Table 1. Duplicate items were removed and if multiple measurements for the same ligands existed, the average pChEMBL value (pX, including pKi, pKd, pIC50, or pEC50) was calculated. To judge if a molecule is active or not, we defined the threshold of bioactivity as pX = 6.5. If the pX < 6.5, the compound was predicted as undesired (low affinity to the given target); otherwise, it was regarded as desired (having high affinity) . ## Prediction Model In order to predict the pX for each generated molecule for a given target, regression QSAR models were constructed with different machine learning algorithms. To increase the chemical diversity available for the QSAR model we included lower quality data without pChEMBL value, i.e. molecules that were labeled as "Not Active" or without a defined pX value. For these data points we defined a pX value of 3.99 (slightly smaller than 4.0) to eliminate the imbalance of the dataset and guarantee the model being able to predict the negative samples. During the training process, sample weights for low quality data were set as 0.1, while the data with exact pX were set as 1.0. This allowed us to particularly incorporate the chemical diversity, while avoiding degradation of model quality. Descriptors used as input were ECFP6 fingerprints with 2048 bits (2048 dimensions, or 2048D) calculated by the RDKit Morgan Fingerprint algorithm (using a three-bond radius). Moreover, the following 19D Before being input into the model, the value of input vectors were normalized to the range of by the MinMax method. Model output value is the probability whether a given chemical compound was active based on this vector. In the RF, the number of trees was set as 1000 and split criterion was "gini". In the SVM, a radial basis function (RBF) kernel was used and the parameter space of C and γ were set as [2 -5 , 2 15 ] and [2 -15 , 2 5 ], respectively. In the MT-DNN, the architecture contained three hidden layers activated by a rectified linear unit (ReLU) between input and output layers, and the number of neurons were 2048, 4000, 2000, 1000 and 3 in these subsequent layers. The training process consisted of 100 epochs with 20% of hidden neurons randomly dropped out between each layer. The mean squared error was used to construct the loss function and was optimized by the Adam algorithm with a learning rate of 10 -3 . ## Generative Model As in DrugEx v1, we organized the vocabulary for the SMILES construction. Each SMILES-format molecule in the ChEMBL and LIGAND sets was split into a series of tokens. Then all tokens existing in this dataset were collected to construct the SMILES vocabulary. The final vocabulary contained 85 tokens (Table S1) which were selected and arranged sequentially into valid SMILES sequences through correct grammar. The RNN model constructed for sequence generation contained six layers: one input layer, one embedding layer, three recurrent layers and one output layer. After being represented by a sequence of tokens, molecules can be received as categorical features by the input layer. In the embedding layer, vocabulary size, and embedding dimension were set to 85 and 128, meaning each token could be transformed into a 128 dimensional vector. For a recurrent layer, the long-short term memory (LSTM) was used as recurrent cell with 512 hidden neurons instead of the gated recurrent unit (GRU) which was employed only in DrugEx v1. The output at each position was the probability that determined which token in the vocabulary would be chosen to grow the SMILES string. During the training process we put a start token (GO) at the beginning of a batch of data as input and an end token (END) at the end of the same batch of data as output. This ensures that our generative network could choose correct tokens each time based on the sequence it had generated previously. A negative log likelihood function was used to construct the loss function to guarantee that the token in the output sequence had the largest probability to be chosen after being trained. In order to optimize the parameters of the model, the Adam algorithm was used for the optimization of the loss function. Here, the learning rate was set at 10 -3 , the batch size was 512, and training steps were set to 1000 epochs. ## Reinforcement Learning SMILES sequence construction under the RL framework can be viewed as a series of decision-making steps (Fig. 1). The generator (G) and the predictors (Q) are regarded as the policy and reward function, respectively. In this study we use multi-objective optimization (MOO), and each objective is a requirement to be achieved maximally for each scenario, albeit with differences in desirability. Our aim was defined by the following problem statement: Here, n equals the number of objectives (n = 3 in this study), and Ri, the score for each objective i, was calculated as follows: here the pXi (the range from 3.0 to 10.0) was the prediction score given by each predictor for the i th target, which was normalized to the interval as the reward score. If having no or low affinity for a target was required (off-target) this score would be subtracted from 1 (inverting it). In order to evaluate the performance of the generators, three coefficients are calculated with the generated molecules, including validity, desirability, and uniqueness which are defined as: where Ntotal is the total number of molecules, Nvalid is the number of the molecules parsed by the valid SMILES sequences, Nunique is the number of molecules which are different from others in the dataset, and Ndesired is the number of desired molecules. Here, we determine if generated molecules are desired based on the reward Ri if all of them are larger than the threshold (0.5 by default when pX = 6.5). In addition, we calculated SA score (from 1 to 10) for each molecule to measure the synthesizability of which larger value means more difficult to be synthesized. And we also computed QED (from 0 to 1) score to evaluate the drug-likeness of which larger value means more drug-like for each molecule. The calculation of both SA and QED scores were implemented by RDKit. To orchestrate and combine these different objectives, we compared two different reward schemes: the Pareto front (PF) scheme and the weighted sum (WS) scheme. These were defined as follows: (a) Weighted sum (WS) scheme: the weight for each function is not fixed but dynamic, and depends on the desired ratio for each objective, which is defined as: here for objective i the N s i and N l i are the number of generated molecules which have a score smaller or larger than the threshold. Moreover, the weight is normalized ratio defined as: and the final reward R * was calculated by Pareto front (PF) scheme: operates on the desirability score, which is defined as where ti is the threshold of the i th objective, and we set all of objectives had the same threshold as 0.5 as stated in the methods. Given two solutions m1 and m2 with their scores (x1, x2, ..., xn) and (y1, y2, …, yn), then m1 is said to Pareto dominate m2 if and only if: otherwise, m1 and m2 are non-dominated with each other. After the dominance between all pair of solutions being determined, the non-dominated scoring algorithm is exploited to obtain a rank of Pareto frontiers which consist of a set of solutions. The solutions in the top frontier are dominated by the other solutions in the bottom frontier, but the solutions in the same frontier are non-dominated with each other . In order to speed up the non-dominated sorting algorithm, we employed PyTorch to implement this procedure with GPU acceleration. After obtaining the frontiers ranking from dominated solutions to dominant solutions, the molecules were ranked based on the average of Tanimoto-distance instead of crowding distance with other molecules in the same frontier, and molecules with smaller distances were ranked on the top. The final reward R * is defined as: here the parameter k is the index of the solution in the Pareto rank, and rewards of undesired and desired solutions will be evenly distributed in (0, 0.5] and (0.5, 0.1], respectively. During the generation process, for each step, G determines the probability of each token from the vocabulary to be chosen based on the generated sequence in previous steps. Its parameters are updated by employing a policy gradient based on the expected end reward received from the predictor. The objective function is designated as follows: By maximizing this function, the parameters 𝜃 in G can be optimized to ensure that G can construct desired SMILES sequences which can obtain the highest reward scores judged by all the Qs. ## Algorithm extrapolation Evolutionary algorithms (EAs) are common methods used in drug discovery . For example, Molecule Evoluator is one of EAs, with mutation and crossover operations based on SMILES representation for drug de novo design. In addition, some groups also proposed other variations of EAs , e.g., estimation of distribution algorithm (EDA) which is a model-based method and replaces the mutation and crossover operations with probability distribution estimation and sampling of new individuals (Fig. 2) . Similar to EDA, DrugEx is a model-based method too, in which the deep learning model was employed to estimate the probability distribution of sequential decision making. However, we use a DL method to define model-based mutation and crossover operations. Moreover, we employed an RL method to replace the sample selection step for the update of model or population in EDA or EA, respectively. ## Molecular Diversity To measure molecular diversity, we adopted the metric proposed by Solow and Polasky in 1994 to estimate the diversity of a biological population in an eco-system . It has been shown to be an effective method to measure the diversity of drug molecules . The formula to calculate diversity was redefined to normalize the range of values from [1, m] to (0, m] as follows: where A is a set of drug molecules with a size of |A| equal to m, e is an m-vector of 1's and ] is a non-singular m × m distance matrix, in which f(dij) stands for the distance function of each pair of molecule provided as follows: here we defined the distance dij of molecules si and sj by using the Tanimoto-distance with ECFP6 fingerprints as follows: where | si ∩ sj | represents the number of common fingerprint bits, and | si ∪ sj | is the number of union fingerprint bits. ## Performance of Predictors All molecules in the LIGAND set were used to train the QSAR models, after being transformed into predefined descriptors, including 2048D ECFP6 fingerprints and 19D physicochemical properties. We then tested the performance of these different algorithms with five-fold cross validation and an independent test of which the performances are shown in Fig. 4AB. Here, the dataset was randomly split into five folds in the cross validation, while a temporal split with a cut-off at the year of 2015 was used for the independent test. In the cross validation test, the MT-DNN model achieved the highest value for R 2 and the lowest RMSE value for A1AR and A2AAR, but the RF model had the best performance for hERG based on R 2 and RMSE. However, for the independent test the RF model reached the highest R 2 and lowest RMSE across the board, although it was worse than the performance in the cross-validation test. A detailed performance overview of the RF model is shown in Fig. 4C-E. Because the generative model might create a large number of novel molecules, which would not be similar to the molecules in the training set, we took the robustness of the predictor into consideration. In this situation the temporal split has been shown to be more robust . Hence the RF algorithm was chosen for constructing our environment which provides the final reward to guide the training of the generator in RL. ## Model optimization As in our previous work in DrugEx v1, we firstly pre-trained and fine-tuned the generator with the ChEMBL and LIGAND set, respectively. When testing the different types of RNNs, we analyzed the performance of the pre-trained model with 10,000 SMILES generated, and found that LSTM generated more valid SMILES (97.5%) than GRU (93.1%) which had been adopted in our previous work. Moreover, for the finetuning process, we split the LIGAND set into two subsets: training set and validation set; the validation set was not involved in parameters updating but it was essential to avoid model overfitting and to improve uniqueness of generated molecules. Subsequently 10,000 SMILES were sampled for performance evaluation. We found that the percentage valid SMILES was 97.9% for LSTM, larger than GRU with 95.7% valid SMILES, a slight improvement compared to the pre-trained model. In the end, we employed the LSTM-based pre-trained/fine-tuned models for the following investigation. We employed the models for two cases (multi-target and target-specific) of multiobjective drug design towards three protein targets. During the training loop of DrugEx v2, the parameter of ε was set to different values: 10 -2 , 10 -3 , 10 -4 and we also tested it without mutation net, i.e. the value of ε was set to 0. Generators were trained by using a policy gradient with two different rewarding schemes. After the training process converged, 10,000 SMILES were generated for each model for performance evaluation. The percentage of valid, desired, unique desired SMILES and the diversity were calculated (Table 2). Furthermore, we also compared the chemical space of these generated molecules with known ligands in the LIGAND set. Here, we employed first two components of t-SNE on the ECFP6 descriptors of these molecules to represent the chemical space. ## Performance comparisons We compared the performance of DrugEx v2 with DrugEx v1 and two other DL-based de novo drug design methods: REINVENT and ORGANIC . In order to make a fair benchmark, we trained these four methods with the same environments to provide the unified predicted bioactivity scores for each of the generated molecules. It should be mentioned that these methods are all SMILES-based RNNs generators but trained under different RL frameworks. Therefore, these generators were constructed with the same RNN structures of and initialized with the same pre-trained/fine-tuned models. In the WS scheme we did not choose fixed weights for objectives but dynamic values which can be adjusted automatically during the training process. The reason for this is that if the fixed weights should be optimized as the hyperparameters, which would be more time consuming. Moreover, the distribution of scores for each objective was not comparable. If the affinity score was required to be higher, few of the molecules generated by the model with initial state were satisfactory, but if a lower affinity score was required, most of the generated molecules by the pre-trained/fine-tuned model met this need without further training of RL. Therefore, weights were set as dynamic parameters and determined by the ratio between desired and undesired molecules generated by the model at the current training step. This approach ensures that the objectives with lower scores would get more importance than others during the training loop to balance the different objectives and generate more desired molecules. The performance of the model with different ε is shown in Table S2. A higher ε generates molecules with larger diversity but low desirability compared to a lower ε in both multi-target and target-specific cases. In addition, an appropriate ε guarantees the model generates molecules which have a more similar distribution of important substructures with the desired ligands in the LIGAND set. With the WS scheme, the model generates molecules with a high desirability, but the diversity is lower than the desired ligands in the training set. On the contrary, the PF scheme helped the model generate molecules with a larger diversity than the ligands in the training set, but the desirability was not as high as in the WS rewarding scheme. Moreover, the generated molecules in the PF scheme have more similar distribution of substructures to the LIGAND set than in the WS scheme. In the multi-target case, these four methods with different rewarding schemes show similar performance, i.e. the WS scheme can help models improve the desirability while the PF scheme assists models to achieve better diversity and distribution of substructures (Table 2). Here, REINVENT with the PF scheme achieved the largest diversity, whereas DrugEx v1 had the most similar substructure distribution to the molecules in the LIGAND set, and DrugEx v2 achieved the best desirability with both PR and WS schemes compared to the three other algorithms. The diversity and distribution of substructures were also most similar to the best results. In addition, in the target-specific case results were similar to the multi-target case, (Table 3), and for the distribution of purine and furan rings, DrugEx v2 surpassed v1 to be most similar to the LIGAND set. When investigating the SA and QED scores, we observed that PF scheme helped all of generated molecules being more drug-like because of higher QED scores than WS scheme in both multi-target case (Fig. 6A-D) and target-specific case (Fig. 6E-H). In comparison of these methods, the molecules generated by REINVENT were supposedly easier to be synthesized and more drug-like than others, but the molecules of DrugEx v1 had more similar distributions with the molecules in the LIGAND set. With respect to chemical space, we employed t-SNE with the ECFP6 descriptors of all molecules for both multi-target (Fig. 6A-H) and target-specific cases (Fig. 6I-P). In the multi-target case, most of desired ligands in the LIGAND set were distributed in the margin and PR scheme could guide all of the generators to search more regions than WS scheme. In the target-specific case, the desired ligands in the LIGAND set were distributed more dispersed in both of the margin and the center regions. However, PF scheme was not shown the similar results as in the target-specific case to improve the coverage compared with WS scheme except for DrugEx v2. For both of these two cases, only part of the region occupied by desired ligands in the LIGAND set were overlapped with REINVENT and ORGANIC, but almost all of it is covered by DrugEx v1 and v2. Especially, in contrast to WS scheme DrugEx v2 had a significant improvement of chemical space coverage with PF scheme. A possible reason is that the molecules generated by DrugEx v1 and v2 offer a more similar distribution of substructures to desired ligands in the LIGAND set than REINVENT and ORGANIC. As an example, 16 possible antagonists (without ribose moiety and molecular weight < 500) generated by DrugEx v2 with PR scheme were selected as candidates for both multi-target cases and target specific case, respectively. These molecules were ordered by the selectivity which was calculated as the difference of pXs between two different protein targets. In the multi-target cases (Fig. 7A), because the desired ligands prefer A1AR and A2AAR to hERG, the row and column is the selectivity of A2AAR and A1AR against hERG, respectively, while the generated molecules are required to bind only A2AAR rather than A1AR and hERG in the target-specific case (Fig. 7B), selectivity of A2AAR against A1AR and hERG were represented as the row and column, respectively. In multi-target case (A), these molecules were ordered by the selectivity of A1AR and A2AAR against hERG as x-axis and y-axis, respectively. In target-specific case (B), these molecules were ordered by the selectivity of A2AAR against A1AR and hERG as x and y-axis, respectively. In order to prove the effectiveness of our proposed method, we tested it with 20 goaldirected molecule generation tasks on the GuacaMol benchmark platform . These tasks contain different requirements, including similarity, physicochemical properties, isomerism, scaffold matching, etc. The detailed description of these tasks is provided in ref and our results are shown in Table S3. We pre-trained our model with the dataset provided by the GuacaMol platform, in which all molecules from the ChEMBL database are included and similar molecules to the target ligands in the tasks were removed. Then we choose the top 1024 molecules in the training set to fine-tune our model for each task, before reinforcement learning was started. Our method scores the best in 12 out of 20 tasks compared with the baseline models provided by the GuacaMol platform, leading to an overall second place. Moreover, the performance between the LSTM benchmark method and our methods were similar in these tasks, possibly because they have similar architectures of neural networks. All in all, this benchmark demonstrated that our proposed method has improved generality for drug de novo design tasks. It is worth being mentioned that our method is not effective enough yet for some tasks of contradictory objectives in the narrow chemical space. The main reason is that our method emphasizes to obtain a large number of feasible molecules to occupy the diverse chemical space rather than small number of optimal molecules to achieve the highest score. For example, in the Sitagliptin MPO task, the aim is finding molecules which are dissimilar to sitagliptin but have a similar molecular formula to sitagliptin, and our method was not as good as Graph GA, which is a graph-based genetic algorithm. ## Conclusion and Future Prospects In this work, we proposed a Pareto-based multi-objective learning algorithm for drug de novo design towards multiple targets based on different requirements of affinity scores for multiple targets. We transferred the concept of an evolutionary algorithm (including mutation and crossover operations) into RL to update DrugEx for multiobjective optimization. In addition, Pareto ranking algorithms were also integrated into our model to handle the contradictory objectives common in drug discovery and enlarge the chemical diversity. In order to prove effectiveness, we tested the performance of DrugEx v2 in both multi-target and target-specific cases. We found that a large percentage of generated SMILES were valid and desired molecules without many duplications. Moreover, the generated molecules were also similar to known ligands and covered almost every corner of the chemical space that known ligands occupy, which could not be repeated by tested competing methods. In future work, we will try the generality of our proposed methods with different molecular representations, such as graphs or fragments . We will also integrate more objectives (e.g. stability, synthesizability), especially when these objectives are contradictory, such that the model allows user-defined weights for each objective to generate more reliable candidate ligands and better steer the generative process. ## Authors' Contributions XL and GJPvW conceived the study and performed the experimental work and analysis. KY, APIJ, ME and HWTvV provided feedback and critical input. All authors read, commented on and approved the final manuscript. ## Atoms Bonds Controls Common Atoms Aromatic Atoms --Rings Branchs On-Off Considering that the sterochemical information of molecules and ionic bonds were ignored, we removed 690 the "@", "", "/", ".".

chemsum

{"title": "DrugEx v2: De Novo Design of Drug Molecule by Paretobased Multi-Objective Reinforcement Learning in Polypharmacology", "journal": "ChemRxiv"}

open_software_platform_for_automated_analysis_of_paper-based_microfluidic_devices

## Abstract: Development of paper-based microfluidic devices that perform colorimetric measurements requires quantitative image analysis. Because the design geometries of paper-based microfluidic devices are not standardized, conventional methods for performing batch measurements of regularly spaced areas of signal intensity, such as those for well plates, cannot be used to quantify signal from most of these devices. To streamline the device development process, we have developed an open-source program called colorScan that can automatically recognize and measure signal-containing zones from images of devices, regardless of output zone geometry or spatial arrangement. This program, which measures color intensity with the same accuracy as standard manual approaches, can rapidly process scanned device images, simultaneously measure identified output zones, and effectively manage measurement results to eliminate requirements for time-consuming and user-dependent image processing procedures.Paper-based microfluidic devices enable measurement capabilities for a number of fields, from clinical diagnostics 1 to environmental management 2 and food quality monitoring, 3 by employing a variety of detection strategies with different signal output formats. These self-contained analytical systems are typically fabricated from paper patterned with hydrophobic barriers, made from materials such as wax, 4 photoresist, 5 glue, 6 or PDMS. 7 Patterned paper layers can be stacked 8 or folded 9 to create three-dimensional fluidic networks, which enable measurement of target analytes by automating complex liquid handling protocols. Depending on the selected signal formation strategy and analysis method, paper-based microfluidic devices can provide qualitative, semi-quantitative, or quantitative results 10. . Paper-based platforms that employ electrochemical, 11 fluorescence, 12 and chemiluminescence 13 detection strategies enable quantitative measurements, but generally require secondary equipment, such as a portable potentiostat 14 or handheld UV source. 15 To enable measurements without any requirements for specialized external equipment, many developers design devices using colorimetric detection strategies. Qualitative measurements (i.e., on/off sensors) may be interpreted by visual inspection, 8,16 and readout zones may be compared to printed read guides 1 or designed to provide distance-based outputs [17][18][19][20] to enable semi-quantitative measurements. Image analysis is used to characterize assay performance 21 during the device development process, but can also be performed at the point-of-use by smartphone applications 22 to provide quantitative measurements while reducing user training requirements. 23 These applications may operate using algorithms that are specific to the geometry of the device being analyzed or from unpublished code that is not available for modification, 24 requiring developers of paper-based devices to develop their own software tools or rely on manual image analysis protocols.As a critical component of paper-based assay development, especially for qualitative and semi-quantitative devices, image analysis facilitates investigation of device design criteria that determine how a user may interpret developed signal. During the prototyping process, device readout zones are typically imaged using a flatbed scanner or camera, and the acquired images are analyzed to inform device fabrication and assay conditions to provide sufficient analytical performance of the device. Images of device output zones are often measured using free and open-source tools (e.g., ImageJ 25 , Fiji 26 ) that support user-developed plugins 27,28 for application-specific analyses. Although numerous plugins exist, available options do not facilitate streamlined analysis of paper-based output zones of different colors and geometries. While manual approaches for image analysis of colorimetric signals have broad utility for general measurement needs across many fields, application of these tools for analysis of paper-based devices is labor-intensive, user dependent, and time consuming. Our program effectively packages the capabilities of existing general color analysis techniques and applies them towards solving the specific challenges facing the interpretation of paper-based assays (e.g., non-standardized zone numbers and geometries). For example, to analyze a paper-based output zone in ImageJ, a user must first define a region of interest (ROI) for analysis (e.g., using the "Oval" tool for a circular zone). This region is typically defined on a magnified view of a high-resolution (e.g., 600-800 dpi) 8,29 image. At high magnification, it can be difficult for a user to differentiate between the signal-containing area and surrounding material (e.g., hydrophobic wax barrier). To avoid introducing bias in measured signal intensities, this region must also be centered on the output zone so that the selected area does not contain any undesired surrounding material and adequately captures any nonuniform distribution of signal. When the ROI has been placed in the desired location, the output zone may be analyzed by a selected method (e.g., "RGB Measure") 30 . After a single measurement is complete, the ROI can be moved to or recreated on the next output zone so that the measurement process can be repeated. The area and placement of this region must be consistent throughout the analysis process so that measurements are consistent across output zones. Measured values can then be copied or exported for further statistical analysis. The reproducibility of these results may vary by software user, as size and placement of the ROI are both manually defined for individual measurements. Because the device development process typically requires analysis of replicate results across many conditions, potentially necessitating hundreds of devices, image analysis and data processing can be substantially labor-intensive and time-consuming for device developers. For complex devices, these requirements can inhibit broad screening of fabrication or use conditions (e.g., channel geometry, reagent storage, sample volume). Automated image processing can improve the time requirements and precision of measured results for colorimetric paper-based assays, but existing ImageJ plugins and available tools are not compatible with device-specific design geometries 31 and spatial arrangements of color localization in most paper-based devices. Existing ImageJ plugins, such as "ReadPlate" 32 , enable automated analysis of images of well plates with standard configurations (e.g., 96-well plates). This plugin streamlines analysis by allowing the user to define a grid of circular regions of interest that is superimposed on the well plate image. The grid is created by defining the number of rows and columns in the well plate, the pixel coordinates of bounding wells (e.g., wells A1 and H12), and the diameter of each analysis spot. Because paper-based devices are designed in custom, non-linear geometries 33 according to their intended performance and function, their output zones typically do not follow the spatial arrangement of commercial liquid handling tools. Other ImageJ plugins, including "Template Matching" 34 and "Template Matching and Slice Alignment" 35 , can perform automated recognition of desired image features based on a user-generated reference template or selection. These tools are designed to recognize the extent of agreement between a reference template and a larger image and may not be sufficient for recognizing multiple shapes or colors within a single image. Since the development of early paper-based devices, cellular phones have been used to enable quantitative analysis of colorimetric signal 21 . As smartphone technologies and quantitative measurement accessories 36,37 have advanced over time, many groups have written custom applications to quantify signal from paper-based devices at the point-of-care 22, . Smartphone image analysis applications are typically tailored to the geometries of individual paper-based assays 43 and cannot be used to measure output zones that differ from those of the original device. In many cases, the positions of test zones are detected using registration marks patterned within the paper device 22,44 . These recognition algorithms do not independently identify the positions of signal formation and instead analyze known areas of signal formation. Additionally, the source code for these applications is not always published with scientific manuscripts 24 , and the resulting lack of modifiable open-source options requires developers of paper-based devices to either (i) create their own analysis software or (ii) rely on existing inefficient options throughout the assay development process. To address this shortcoming, we have developed a free, open-source software called ColorScan that enables streamlined, automated analysis of paper-based microfluidic devices. This Python-based program automatically identifies and measures signal-containing zones of any geometry or color from images of paper-based devices. Our tool provides a variety of quantitative measurement options based on user-specified criteria and effectively manages data, even providing cropped images of output zones paired with measured results to facilitate figure creation. To verify the performance of this software, we compared the consistency and time requirements of our tool to manual measurements completed using ImageJ. Our software, which has the potential to simplify the time and labor-intensive process of quantitative image analysis for paper-based devices, is freely available 31,45,46 , as an easy-to-use Python program to facilitate widespread use and further improvement by other developers of colorimetric sensors. ## experimental design Identification of desired software features. We designed ColorScan to automate the workflow of our image analysis process, which consists of three main steps: (i) selection of a region of interest, (ii) color intensity measurement within the selected region, and (iii) management of measured results. When a paper-based device is imaged to facilitate analysis, arbitrary placement or rotation of the device on the scanner bed or within the camera's field of view can lead to variability of output zone position across replicate devices. Manual selection of analysis regions can be performed regardless of zone position, but is time consuming, while patterned registration marks for automated analysis programs place design constraints on device developers. We designed ColorScan to automatically recognize colored regions of an image based on hue (i.e., the color or shade of the output zone), saturation (i.e., amount of gray), and value (i.e., the brightness of the output zone). This recognition step is not dependent on the spatial location of colored pixels within the image file, enabling automated analysis of devices imaged in any orientation. Manual analysis protocols, in addition to being labor-intensive and user-dependent, usually only allow an image to be measured within a single color space. Device images are typically acquired in the RBG color space, requiring conversion to determine if another color space (e.g., HSV, CIELAB) 47,48 is more sensitive to the signal formed by a particular sensor or more intuitive for interpretation by visual inspection 49 . Standard image analysis approaches also require measurement results to be manually tracked and compiled into spreadsheets for processing. We developed ColorScan to not just automate analysis in multiple color spaces, but also effectively manage results by organizing them in a common location (i.e., a .csv file) for direct comparison during the assay development process. To streamline comparison of measurement results to their respective output zones, as well as process device images for presentation or publication, we configured ColorScan to crop and save an image of each measured output zone. The file name of each image is labeled so that it may be paired with its numerical measurement results. computational analysis approach. To make our software broadly accessible and readily modifiable, we chose to develop it using Python 50 , a programming language that is used in a wide range of fields, has a large assortment of libraries, and is compatible with a number of third-party modules. We used OpenCV 51 , an extensive open-source computer vision library, in conjunction with NumPy 52 , an array manipulation and numerical operation library, to perform image processing tasks. We also developed a graphical user interface, using the Python TkInter library 53 , to ensure that ColorScan would be accessible to a broad user base without requiring device developers to be experienced in coding. The source script for our Python-based program is available for download at our group's GitHub page 54 and we have included a detailed User Guide document, including instructions for downloading and using Python, as part of the Supplementary Information. The first step in our image analysis protocol (Fig. 1A) is identification of the colorimetric signal contained by the output zones of a paper-based device. We designed ColorScan to automatically recognize the color of output zones against the contrasting color of the patterned hydrophobic barrier, which is black wax in our devices. The software masks the area around the zones by temporarily converting the image to the HSV (i.e., Hue, Saturation, Value) color-space, and excluding pixels below minimum saturation and value thresholds defined by the user. This masking step produces a binarized image (Fig. 1B), in which color-containing pixels within the device output zones are identified for further refinement. To reduce the jaggedness of the masked output zone edges, ColorScan performs a box blur (Fig. 1C) based on a user-defined kernel size. Next, the software runs OpenCV's contour-identification algorithm, which is based on work by Suzuki and Abe 55 , on the binarized image to define the boundaries of the areas where color intensity measurements may be performed (Fig. 1D). In order to cut down on computation time, we set an arbitrary cutoff for contours containing fewer than five pixels in area. This threshold effectively determines the minimum feature size that ColorScan can detect. The open-source nature of the code, however, allows the user to modify this threshold to suit their needs. The maximum feature size is, in principle, the size of the image itself. At this point, the user selects a reference contour to facilitate identification of similar objects for analysis. ColorScan compares the sizes (i.e., pixel area contained within the border) and shapes, defined by the Hu image invariants 56 , of the reference contour and all of the identified contours based on thresholds set by the user in the graphical user interface. In this comparison, the OpenCV shape-matching function computes the sum of absolute differences between each of the seven Hu invariants for the two contours under comparison (i.e., the reference contour and another contour), and produces a score indicating how similar the two contours are. Contours with scores within the user-defined thresholds are selected for batch analysis (Fig. 1E). For device designs that use a colored hydrophobic barrier feature to provide contrast for visual interpretation of assay results 8,57 , the software may recognize these features as part of the output zone. To ensure that inclusion of barrier edge pixels does not bias measurement results, we developed a zone refinement tool that allows the user to geometrically constrain the reference contour to exclude undesired image features. This constraint needs only be manually performed on the reference contour and is applied to all similar contours before analysis. This feature also allows the user to select regular polygonal geometries, in addition to common circular output zones. Performing this step provides direct control of the exact size and position of each analysis region in a batch measurement (Fig. 1F). Once the user is satisfied that all of the output zones-and only the output zones-have been selected, the software will measure all of the pixels within each contour. Results can be presented as average color intensities in up to three color spaces, including RGB, HSV, and CIELAB, and also as histograms of RGB intensities, depending on user-defined preferences. Measurements are exported to a .csv file and may be compared to saved images of output zones, which are automatically cropped from the full image based on the bounding rectangle of each output zone contour and saved with their associated index identifier to facilitate data curation and presentation. Design of software and graphical user interface. We designed ColorScan to have a simple graphical user interface that would streamline the image analysis process and improve consistency across users. The first step in using ColorScan is selecting an image. Selecting an image using the "Select Image" button displays it in the main window of the software (Fig. 2A). The program supports most image file types, the full list of which can be found in the OpenCV documentation 58 . The image in the main ColorScan window updates as options in the Analysis Menu window (Fig. 2B) are adjusted. This window is opened with the "Analysis" button after image selection, and its features are spatially arranged, from top to bottom, in the sequence that they should be used to complete image analysis. At the beginning of the analysis process, only the masking controls are available in the Analysis Menu window. Additional features become available as each step in the analysis sequence is completed. First, the value slider is used to mask the area surrounding the output zones and highlights pixels above a selected brightness threshold. Next, the saturation slider is used to refine the masked areas by filtering unsaturated www.nature.com/scientificreports/ pixels (i.e., those close to grayscale) to show only pixels containing color. After identification of the output zones, the blur options can be used to smooth the mask edges and reduce the granularity of the areas to be analyzed. The "Dilate" and "Erode" buttons expand and shrink the masked and blurred areas, respectively, allowing for further refinement before selection of the output zone contours. Masking and blurring parameters can be saved into a named preset, which can be loaded during subsequent analyses to facilitate rapid, consistent analysis of assay replicates. Once masking parameters include all of the desired output zones, the user can find the contours of those zones using the "Find Contours" button. This will select all color-containing regions in the image, from which the user may select a reference contour by clicking on it. The "Find Similar Contours" button selects all regions that are similar in size and shape to the reference contour for analysis. The Size Tolerance and Shape Tolerance values should be adjusted so that only the device output zones are highlighted. After these zones are selected for analysis, clicking the Refine Zones button will open a separate window that enables geometric restriction of the analysis area within each output zone. This tool is useful for cases where undesired pixels from the hydrophobic barrier edge (e.g., black wax) or a patterned contrast feature (e.g., colored rings) are selected by the masking and contour finding steps but should not be included in the area that is being measured. To ensure that only signal from the paper-based assay is measured, the tools in the Refine Zone window allow the user to constrain the position, shape, and area of the analysis region within the reference contour. This analysis geometry is applied to all identified similar contours to provide consistent measurement area and shape across all analysis regions. It also allows for better handling of potential problem-cases where there is insufficient contrast between the color-containing regions of interest (i.e., the test zones) and the background color from the device (e.g., colored wax barriers or the paper itself), which By refining the zone area, the user may cut out regions potentially erroneously included in the analysis region. This process can be optimized for a specific device geometry and then saved as a preset to facilitate objective and reproducible measurements across multiple images. These techniques, in combination with judicious device design features (e.g., colored rings printed around output zones to improve contrast) make consistent contour identification possible. Importantly, the positional controls in the Refine Zone window (X Displacement, Y Displacement, Angle) can be used to define the analysis region at the end of a paper channel (Fig. 3). Many device designs 21,22,57 contain detection zones at the ends of paper channels used for fluid distribution. While the white or grey color of these channels can be difficult to distinguish from colorimetric signal during masking steps in ColorScan, our zone refinement approach allows these features to be analyzed in a controlled, reproducible manner. After zone refinement, pressing the "Analyze" button completes automated measurement of all analysis regions selected by the user-defined criteria and exports results and zone images to the same directory as the original image. ## Results and Discussion color intensity measurements. We compared the user experience and measurement results provided by ColorScan to those of our standard image analysis approach, completed using ImageJ, to evaluate the performance of our custom software. To complete this comparison, we analyzed four replicate paper-based microfluidic devices (Fig. 4A) containing six circular output zones each, using both ImageJ and ColorScan. These fourlayered devices, described in the Materials and Methods document as part of the Supplementary Information, contained internally stored and spatially separated reactants for six colorimetric reactions. When these devices were run with water, stored analytes and their reagents were rehydrated and delivered to output zones where colorimetric signal developed to indicate the presence of (i) cysteine, (ii) a neutral pH, (iii) sulfite, (iv) cobalt(II), (v) iron(II), or (vi) molybdate. Each of these zones was surrounded by a wax-printed contrast ring (Fig. 4B), which could be used to support signal interpretation by visual inspection. These devices were not designed to perform measurements for these analytes at the point-of-care, but the homogenous (e.g., sulfite zone) and heterogeneous (e.g., cobalt(II) zone) distributions of signal intensity in their output zones are demonstrative of signal formation patterns found in practical paper-based microfluidic devices. All devices were imaged using an Epson V600 Photo flatbed scanner at a resolution of 800 dpi. Scanning at lower resolutions is not expected to appreciably bias measurements acquired from a test zone-data from neighboring pixels within a zone that can be resolved in a high-resolution image are effectively accounted for through averaging in a low-resolution image. Practically, we suggest setting 300 dpi as a lower limit for images and scans. To complete our performance evaluation, we began by manually measuring the pixel intensity of each output zone in ImageJ using the "RGB Measure" tool 30 . We created a circular region of interest on the scanned device image and manually measured the output zones in the numerical order shown in Fig. 4B, then compiled measurement results in a Microsoft Excel spreadsheet. To acquire images of each output zone, as ColorScan does automatically, we manually cropped selected features from the scanned device image using Adobe Photoshop. In total, our manual measurement and image processing steps took approximately 24 min. ColorScan analysis of the same image took only 2 min and automatically provided an organized spreadsheet of results and cropped images of output zones. This is a reasonable analysis time for a trained ColorScan user, and we expect the time investment required for users to familiarize themselves with the software to be minimal. Unlike manual analysis performed using ImageJ, the time required to perform automated measurements in ColorScan does not significantly depend on the number of output zones being measured, offering substantial time savings over conventional methods. Measurement results obtained in the RGB color space using each software are shown in Fig. 5. Further For each set of output zones measured in the RGB color space, mean pixel intensities acquired using ColorScan were consistent with values obtained using ImageJ. On average, ColorScan and ImageJ values were 0.7% different, with a maximum difference of 1.8% for the blue channel intensity of the purple-colored signal in output spot 6 of each device (Table 1). These differences may be related to minor inconsistencies in analysis region position or size in each approach, or the computational methods used to average pixel intensity in each program. Additionally, the variance of mean pixel intensity values for replicate output zones is comparable for measurements performed using ColorScan and ImageJ (Table S3). These results indicate that ColorScan, in addition to providing a user-friendly approach for streamlined image analysis, performs pixel intensity measurements just as well as standard image analysis programs.

chemsum

{"title": "Open software platform for automated analysis of paper-based microfluidic devices", "journal": "Scientific Reports - Nature"}

interaction_of_synthetic_human_slurp-1_with_the_nicotinic_acetylcholine_receptors

## Abstract: Human SLURP-1 is a secreted protein of the Ly6/uPAR/three-finger neurotoxin family that co-localizes with nicotinic acetylcholine receptors (nAChRs) and modulates their functions. Conflicting biological activities of SLURP-1 at various nAChR subtypes have been based on heterologously produced SLURP-1 containing N-and/or C-terminal extensions. Here, we report the chemical synthesis of the 81 amino acid residue human SLURP-1 protein, characterization of its 3D structure by NMR, and its biological activity at nAChR subtypes. Radioligand assays indicated that synthetic SLURP-1 did not compete with [ 125 I]-α-bungarotoxin (α-Bgt) binding to human neuronal α7 and Torpedo californica muscle-type nAChRs, nor to mollusk acetylcholine binding proteins (AChBP). Inhibition of human α7-mediated currents only occurred in the presence of the allosteric modulator PNU120596. In contrast, we observed robust SLURP-1 mediated inhibition of human α3β4, α4β4, α3β2 nAChRs, as well as human and rat α9α10 nAChRs. SLURP-1 inhibition of α9α10 nAChRs was accentuated at higher ACh concentrations, indicating an allosteric binding mechanism. Our results are discussed in the context of recent studies on heterologously produced SLURP-1 and indicate that N-terminal extensions of SLURP-1 may affect its activity and selectivity on its targets. In this respect, synthetic SLURP-1 appears to be a better probe for structure-function studies.The three-finger fold is a protein domain structure comprising a disulfide-stabilized core from which three elongated loops (fingers) protrude (Fig. 1). It features prominently in two large protein families: snake venom neurotoxins and the Ly6 proteins, the latter first discovered in the mammalian immune system 1-4 . Besides their similar 3D structures, proteins with this fold also share a similar genetic organization and a conserved pattern and connectivity of cysteine residues that ultimately form the structure-stabilizing disulfides. These common features provide strong evidence that Ly6 proteins and snake venom neurotoxins are evolutionary related, however, despite the structural similarities the functional link between these two families has only emerged recently.Most Ly6 proteins are membrane-tethered by a covalently attached glycosyl phosphatidylinositol (GPI) anchor, such as for Ly6/neurotoxin 1 (Lynx1), but some are secreted proteins including SLURP-1 (secreted Ly6/urokinase-type plasminogen receptor-related protein), which was initially isolated from human blood and urine 5 . SLURP-1 is also expressed in keratinocytes and SLURP-1 mutations are implicated in the Mal de Meleda skin disease 4,6 . Additionally, SLURP-1 has been reported to regulate processes in the immune and nervous systems [7][8][9] .SLURP-1 (as well as other Ly6 proteins such as Lynx1 and SLURP-2) represents a functional link between the mammalian Ly6 proteins and snake neurotoxins. Many members from the latter group, which include the well characterized pharmacological agents α-bungarotoxin (α-Bgt) and α-cobratoxin (α-Cbt), are potent inhibitors of nicotinic acetylcholine receptors (nAChR). Co-localization studies and functional in vitro activity data have demonstrated that certain Ly6 proteins (Lynx1 and SLURP-1 and -2) also interact with nAChRs, suggesting that they might function as endogenous modulators of nAChR signaling in vivo 10,11 . Various recombinant versions of SLURP-1 have been shown to modulate nAChRs, mostly of the α7 subtype 7,8,12 , but with contradictory results due to the expressed SLURP-1 proteins containing additional C-or N-terminal fusion tags. For example, Chimienti and colleagues reported potentiation of α7 nAChR-mediated currents by a recombinant myc-His 6 -SLURP-1 fusion construct at low nanomolar concentrations 12 . In contrast, recombinant SLURP-1 expressed in E. coli, containing an additional methionine residue at the N-terminus (hereafter referred to as rSLURP-1), exhibited inhibitory activity at α7 nAChR (at micromolar concentrations) 13 . To resolve these conflicting data on SLURP-1 activities, we report here the chemical synthesis and biological activity of the 81 amino acid human SLURP-1 identical in amino acid sequence to the human serum-derived protein. Using a combination of solid phase peptide synthesis and native chemical ligation 14 , high purity protein was obtained in multi-milligram amounts sufficient for structural and functional studies. The synthetic protein was characterized by HPLC, MS, and NMR which confirmed the three-finger fold structure. Most importantly, our pharmacological data revealed for the first time the interaction of synthetic SLURP-1 (sSLURP-1) with several neuronal nAChR subtypes. ## Results Human SLURP-1 synthesis and NMR structural analysis. Human SLURP-1, with 81 amino acid residues and five disulfide bonds, is considerably larger than most three-finger proteins from snake venoms (up to 62 residues and four disulfides). It is reminiscent of the classical long-chain α-neurotoxins, which are typically composed of up to 75 residues with a 5 th disulfide bond in the central loop II. In contrast, in all Ly6 proteins, including SLURP-1, the 5 th disulfide resides in the N-terminal loop I 4 . Given the size of the target SLURP-1 molecule, we resorted to a peptide segment ligation approach to overcome the size limitation of traditional stepwise solid phase peptide synthesis (SPPS) (Figs 1 and 2) 15 . Accordingly, the SLURP-1 polypeptide chain was split into three segments, which were individually assembled by either Boc or Fmoc SPPS (see Materials and Methods). Thiazolidine-4-carboxylic acid (Thz) was used in place of Cys21 16,17 to prevent cyclisation and oligomerisation during the first chemical ligation of SLURP-1 [21-50] and SLURP-1 [51-81]. Following cleavage from the solid support and purification, the segments were joined via native chemical ligation in one-pot fashion as described previously 16,18 . The fully reduced SLURP-1 polypeptide was obtained in good yield (69% based on the limiting starting peptide segment SLURP-1[51-81]). Folding and disulfide formation of the synthetic molecule was achieved using protocols described recently for inclusion body refolding of rSLURP-1 produced in E. coli 13,19 . The folding kinetics and the overall HPLC folding profile were essentially identical to those reported for rSLURP-1 and allowed preparation of synthetic SLURP-1 in high purity and in multi-milligram quantities (Fig. 2D-F). High resolution MS analysis indicated a monoisotopic mass of 8837.1 ± 0.1 Da, in excellent agreement with the theoretical monoisotopic mass of 8837.02 Da demonstrating formation of five disulfide bonds (Fig. 2F). To verify the anticipated three-finger fold of the synthetic material, we performed NMR experiments under the same conditions as reported previously for rSLURP-1 (i.e., H 2 O/D 2 O (9:1), pH 4.8, 310 K) 19 . The natural abundance 1 H- 15 N HSQC spectrum of SLURP-1 showed good dispersion of amide proton and nitrogen resonances (Fig. 3A) suggesting that the molecule adopts a well-defined three-dimensional structure. Overall, the spectrum is highly similar to that of rSLURP-1 19 . Two-dimensional TOCSY and NOESY spectra were used to assign backbone amide and CαH protons. Comparison of the Hα chemical shift of each residue obtained from synthetic SLURP-1 spectra to the values available for rSLURP-1 (BMRB ID: 25225 and 25226; Fig. 3B) revealed excellent agreement, suggesting the proteins have highly similar three-dimensional structures. This suggestion is further supported by several key long-range NOEs observed in the NOESY spectra of synthetic SLURP-1, including Lys2Hα-Arg20Hα, Cys28Hα-Cys51Hα, Met29Hα-Cys71Hα, Thr30Hα-Arg49Hα and Leu76Hα-Tyr4Hδ/ε, all consistent with the proposed three-finger fold. Taken together, these data confirm that our synthetic SLURP-1 has a tertiary structure similar to that of rSLURP-1. Human SLURP-1 does not compete with α-Bgt at human neuronal α7, muscle-type nAChRs and AChBPs. The recombinant form of SLURP-1 was shown previously to displace bound α-Bgt from the muscle-type nAChR of Torpedo californica and the mollusk Lymnaea stagnalis AChBP 13 . However, in the present study, the synthetic version of SLURP-1 did not compete with α-Bgt for either proteins (Fig. 4). In addition, no competition with α-Bgt binding was observed at either human (h) α7 nAChR or Aplysia californica AChBP (Fig. 4). ## SLURP-1 inhibition of hα7 nAChR in the presence of the positive allosteric modulator PNU120596. In the [ 125 I]-α-Bgt binding assay, synthetic SLURP-1 showed no competitive antagonism at hα7 nAChR, consistent with the reported inactivity of rSLURP-1 in the same assay 13 . However, rSLURP-1 inhibited ACh-evoked currents at hα7 nAChR 13 and since the inhibition showed a direct relationship with the ACh concentration, we tested synthetic SLURP-1 at hα7 and rat (r) α7 nAChRs heterologously expressed in Xenopus laevis oocytes, under similar conditions (Fig. 5). At 10 μM, regardless of the ACh concentrations used (10, 100, 300, or 1000 µM), synthetic SLURP-1 did not antagonize ACh-evoked currents mediated by hα7 (Fig. 5A,B) and rα7 (Fig. 5A and C) nAChR subtypes. The activity of some ligands at α7 nAChR-mediated currents can be amplified in the presence of the α7 subtype specific positive allosteric modulator PNU120596 20 . Therefore, we investigated the activity of sSLURP-1 in the presence of 10 μM PNU120596 and indeed, inhibition of epibatidine (Epi)-induced Ca 2+ influx (59% inhibition by 5 μM SLURP-1 at 150 nM Epi) was observed in neuroblastoma Neuro2a cells expressing hα7 nAChR (Fig. 5D). sSLURP-1 was also tested on mouse muscle nAChRs expressed in Neuro2a cells but no significant inhibition of ACh-evoked Ca 2+ influx was detected (Suppl. Figure 1). Selective inhibition of heteromeric human neuronal nAChRs by SLURP-1. The activity of sSLURP-1 was also determined at respective ACh EC 50 currents of heteromeric human nAChRs expressed in X. laevis oocytes (Fig. 6). sSLURP-1 at 10 µM reversibly inhibited ACh-evoked current amplitude of hα3β4 nAChRs by ~60%, whereas ~30% inhibition was observed at hα3β2 and hα4β4 nAChRs, and no inhibition was observed at hα4β2 nAChR. sSLURP-1 inhibited ACh-evoked currents mediated by hα3β4 in a concentration-dependent manner with an IC 50 of 4.75 ± 0.78 µM (Fig. 6D). At hα3β4 nAChR in the presence of <300 µM ACh (EC 50 ), sSLUPR-1 inhibition was enhanced (~80% with both 30 and 100 µM ACh), whereas with 1 mM ACh, sSLURP-1 inhibitory effect was comparable to that observed with 300 µM ACh. Interestingly, although the hα9α10 subtype was not inhibited by sSLURP-1 in the presence of 6 μM ACh (EC 50 ), sSLURP-1 inhibition became clearly manifested at 100 and 300 μM ACh (~25% and ~40% inhibition of ACh-evoked current amplitude, respectively) (Fig. 6A and B). Furthermore, we also demonstrated the sensitivity of rα9α10 nAChR to sSLURP-1 inhibition which strongly correlated with the increased concentration of agonist (Fig. 6A and C). ## Discussion Human SLURP-1 has the canonical three-finger folded structure of the snake α-neurotoxins, which are known as potent antagonists of nAChRs 4 . SLURP-1 has been shown to participate in a number of cellular regulation pathways, supposedly by acting on the homomeric α7 nAChR subtype 21,22 . However, the mechanism of interaction between SLURP-1 and α7 nAChR remains unclear due to the disparities in the activities of the recombinant human SLURP-1 constructs used. Recently, it was demonstrated that rSLURP-1 inhibited α7 nAChR-mediated currents 13 , whereas potentiation was reported previously for the myc-tagged fusion protein 12 . In contrast, binding of rSLURP-1 at the orthosteric and allosteric sites of different targets was registered 13 , which is in agreement with the similar activity of ws-Lynx1, thus supporting the proposed binding models 23,24 . We suspected these discrepancies potentially originate from the different chemical structures of the various recombinant forms of SLURP-1 used. Hence to resolve the conflicting data, we chemically synthesized the SLURP-1 protein (sSLURP-1) identical in amino acid sequence to the naturally-occurring human molecule 5 , and determined its structural and biological properties. Three-finger proteins have been obtained previously by stepwise chemical solid phase peptide synthesis (SPPS) 25 . However, to the best of our knowledge, this approach has been limited to selected short-chain α-neurotoxins typically comprising ~60 amino acids and four disulfide bonds, with the longest being the synthetic non-conventional neurotoxin, built of 66 residues with five disulfides 26 . Larger proteins, including SLURP-1, are generally difficult to synthesize by stepwise Fmoc or Boc SPPS alone 15 . Chemical synthesis of SLURP-1 was achieved using a convergent approach whereby the polypeptide was divided initially into three shorter peptide segments, each ranging in size of about 20-30 amino acids. The segments were prepared by Boc or Fmoc SPPS in good yield and purity, purified individually by HPLC and chemo-selectively linked together using the recently established one-pot native chemical ligation protocol 16 . Folding and disulfide bond formation of the synthetic 81-mer was achieved over 72 h using a glutathione redox shuffling system. To our knowledge, the successful chemical synthesis of SLURP-1 reported here, is the first example of a long chain three-finger protein of the Ly6 family obtained solely through chemical synthesis. To unambiguously confirm the anticipated 3D structure, we performed NMR experiments that allowed near-complete assignment of backbone NH and CαH protons (as well as partial side chain proton assignments). This analysis and comparison with data previously obtained for rSLURP-1 19 , established that the synthetic molecule is structurally highly comparable to the recombinant protein. Contrary to earlier reports of various recombinant SLURP-1 versions interacting with AChBPs and nAChRs 19,21 , we did not observe competition of synthetic SLURP-1 with radio-iodinated α-Bgt for binding to AChBPs of L. stagnalis and A. californica, nor to muscle-type T. californica nAChR (Fig. 4). Interestingly, although neither synthetic SLURP-1 and rSLURP-1 did compete with α-Bgt binding to hα7 nAChR, only rSLURP-1 inhibited ACh-evoked currents at hα7 nAChR with an IC 50 of ~1 μM and the inhibition was enhanced with increasing ACh concentrations 13 . However, synthetic SLURP-1 did not inhibit either hα7 or rα7 nAChRs in the presence of low and high concentrations of ACh (Fig. 5A-C). The antagonistic effect of sSLURP-1 was only observed under the influence of the α7 nAChR positive allosteric modulator PNU120596, where substantial inhibition of hα7-mediated epibatidine-induced Ca 2+ influx (Fig. 5D) was observed. Screening of sSLURP-1 at 10 μM against a number of heteromeric human neuronal nAChR subtypes, demonstrated preferential inhibition of nAChRs co-expressing α3 and β4 subunits (Fig. 6A and B), with hα3β4 nAChR being more sensitive to inhibition by sSLURP-1 compared to hα3β2 and α4β4 nAChRs. This finding suggests that the binding site of sSLURP-1 might be located at the interface of α3 and β4 subunits and sSLURP-1 behaved as a competitive antagonist of the hα3β4 subtype. On the other hand, both hα9α10 (Fig. 6A and B) and rα9α10 (Fig. 6A and C) nAChRs showed ACh-dependent sSLURP-1 inhibition, with sSLURP-1 exerting its action at relatively high ACh concentrations. Ws-Lynx1 (a recombinant version of Lynx1 lacking the GPI anchor) was also shown to profoundly inhibit hα7, α3β2, and α4β2 nAChRs 23 , and the chimeric α7/glycine (Gly) receptor 24 in similar fashion. Overall, sSLURP-1 inhibition at the human αβ (except hα4β2) and α9α10 nAChRs subtypes can be registered at low ([ACh] ~EC 50 ) and high ACh concentrations, respectively, conditions that are more "physiologically relevant". In contrast, the action of sSLURP-1 on the hα7 nAChR is observed only in the presence of the artificial potentiator PNU120596. Taken together, the results obtained for α7 and α9α10 nAChRs suggest that sSLURP-1 behaves as a 'silent' negative allosteric modulator, exerting its inhibitory effect at the nAChRs only when the receptor channels are in a stable open state. A similar unmasking effect on ligand activity was also reported for α-conotoxin MrIC 27 and the marine sponge-derived 6-bromohypaphorine 28 , where both behave as an agonist by eliciting concentration-dependent increases in [Ca 2+ ] i via PNU120596-modified hα7 nAChR. As for α9α10, the mode of action of sSLURP-1 on this nAChR subtype mirrors the proposed negative allosteric mechanism of rSLURP-1 13 and ws-Lynx1 23,24 where both are more potent inhibitors of α7 nAChR in the presence of high ACh concentrations. Despite the structural similarities in the NMR-Hα chemical shift profiles of sSLURP-1 and rSLURP-1 (Fig. 3B), these proteins clearly behave differently at their targets. Using the published NMR structure of rSLURP-1 (PDB ID: 2MUO), we built a model for sSLURP-1 (Fig. 7A) to investigate in more detail the structural differences between the two proteins. Molecular dynamic simulation of both structures indicates that the additional N-terminal methionine (Met0) residue in the rSLURP-1 is tightly packed inside the disulfide-rich core of the molecule (known to be important for stabilizing the overall conformation), whereas the absence of this residue in sSLURP-1 may allow Arg20 to protrude from the protein surface, possibly allowing it to participate in receptor interactions (Fig. 7B and C). Furthermore, Lys2 and Asp75 form a stable salt-bridge, which is absent in rSLURP-1 (Fig. 7C, blue arrow). Such structural changes may account for the different biological actions observed for sSLURP-1 and rSLURP-1. Recently, another member of the Ly6 protein superfamily, SLURP-2, was recombinantly expressed in E. coli with an additional N-terminal methionine residue, similar to rSLURP-1 29 . In previous publications utilizing different fusion forms, SLURP-2 was claimed to act selectively on α3-containing nAChRs 30 . However, rSLURP-2 is functionally more similar to ws-Lynx1 than to rSLURP-1. At micromolar concentrations, rSLURP-2 inhibited α4β2, α3β2 and α7 nAChRs, whereas at lower concentrations it potentiated the α7 nAChR-mediated currents 29 . This study also provided the NMR structure for rSLURP-2 (PDB ID: 2N99) revealing a considerable conformational mobility, comparable to that earlier observed for rSLURP-1. Although we did not perform a direct comparison of the various reported recombinant forms of SLURP-1, our results suggest that even one additional methionine residue at the N-terminus, probably by affecting the spatial structure, can produce marked changes in the functional activity of Ly6 proteins. structures showing the differences in the "head" region where Met0 is located in rSLURP-1. The N-terminal amino groups are indicated by red arrows. In the sSLURP-1 structure, a salt-bridge between residues Lys2 and Asp75 is present (blue arrow). Some residues have been omitted for clarity. ## Conclusions We wish to emphasize that our work on the synthetic protein identical in amino acid sequence to the naturally-occurring human SLURP-1 does not undermine the previous work on Ly6 analogs produced in E. coli. The activities reported with them may open new avenues to diagnostics and drug development. However, our results clearly show that unraveling physiologically-relevant mechanisms for endogenous regulators requires the study of compounds which should be as close as possible to the native proteins. In this context, total chemical synthesis has been particularly instrumental in the past for small proteins, and recent advances in peptide chemistry 14,31 have allowed this concept to be extended to larger proteins. Furthermore, our work highlights the need for strict compound characterization standards if results are to be reproducible and transferable. With the advancements in modern analytical techniques that are customary in synthetic organic chemistry (high-resolution mass spectrometry, NMR and X-ray structure determination) it has become necessary to apply the same rigorous standards to proteins and other biologics produced by recombinant expression. Our study revealed, for the first time, human sSLURP-1 interactions with several neuronal nAChR subtypes (hα3β4, hα3β2, hα4β4 and hα9α10). We expect these findings will be important for understanding the in vivo function of SLURP-1 in human health and disease in the future. ## Methods Peptide synthesis. All amino acids used were of the L-configuration. The SLURP-1[51-81] (CSSSCVATDPDSIGAAHLIFCCFRDLCNSEL) segment was synthesized by automated Fmoc SPPS using standard protocols. The peptide was assembled on 2-chlorotrityl chloride resin using the following side chain protecting groups: Cys(Trt), Asp(tBu), Glu(tBu), His(Trt), Asn(Trt), Arg(Pbf), Ser(tBu) and Thr(tBu). Resin cleavage and side-chain deprotection were carried out by suspending the dried peptide-resin in cleavage cocktail (trifluoroacetic acid (TFA):triisopropylsilane: H 2 O;95:2.5:2.5) (v/v/v)). After stirring for 1.5 h at room temperature, majority of the TFA was evaporated under vacuum and the peptide was precipitated with ice-cold diethyl ether. The peptide was dissolved in 50% acetonitrile (ACN)/water containing 0.05% TFA and lyophilized. Peptide α-thioalkylesters corresponding to SLURP-1[1-20]-α -thioester (LKCYTCKEPMTSASCRTITR-[COS]-Ser) and SLURP-1[21-50]-α-thioester (Thz-KPEDTACMTTLVTVEAEYPFNQSPVVTRS-[COS]-Lys) were assembled by manual in situ neutralization Boc chemistry as described previously 32,33 . The following standard side chain protection groups were used: Cys(4-MeBzl), Arg(Tos), Asp(OcHx), Asn(Xan), Glu(OcHx), Gln(Xan), Lys(2Cl-Z), Ser(Bzl), Thr(Bzl), Tyr(Br-Z). Following chain assembly, peptides were side chain-deprotected and simultaneously cleaved from the resin by treatment with anhydrous HF containing 10% (v/v) p-cresol for 1 h at 0 °C. HF was evaporated under reduced pressure. The crude product was precipitated and washed with chilled diethyl ether, then dissolved in 50% (v/v) aqueous ACN containing 0.1% TFA (v/v) and lyophilized. Peptides were purified by reversed-phase high-pressure liquid chromatography (RP-HPLC) using a preparative Vydac C18 (22 × 250 mm) column on a Shimadzu Prominence platform. Crude peptides were dissolved in a 10% (v/v) ACN-water mixture containing 0.05% (v/v) TFA, before being loaded onto the column pre-equilibrated with 10% of solvent B (ACN: H 2 O:TFA; 89.5:10:0.05) in solvent A (H 2 O:TFA; 99.5:0.05). Peptides were eluted using linear gradients of solvent B in solvent A, and fractions were collected across the expected elution time. Peptide purity and identity were assessed by ESI-MS on API-2000 mass spectrometer (Applied Biosystems) and by analytical scale uHPLC on a Shimadzu Nexera system equipped with an Agilent Zorbax C18 column (1.8 μm, 2.1 × 100 mm). Fractions containing the desired product were pooled, lyophilized and stored at −20 °C. One-pot native chemical ligation. Initially, 76 mg of SLURP-1[51-81] (MW: 3278.7, 23.2 µmol) and 90 mg of SLURP-1 [21-50] (MW: 3545.9, 25.4 µmol) were dissolved in 15 mL of ligation buffer (6 M GdmHCl, 200 mM Na-phosphate, 50 mM mercaptophenylacetic acid, 40 mM TCEP, pH 7.0). The mixture was stirred under an argon atmosphere for 12 h after which LC-MS analysis indicated near quantitative product formation, yielding SLURP-1[21-81] (Cys21Thz) with an observed mass 6590.2 ± 0.6 Da, calculated mass: 6590.4 Da (average isotope composition). Methoxyamine HCl was added to a final concentration of 250 mM and the pH was adjusted to 4.0-4.1 with concentrated HCl. The reaction was left for 8 h and stirred under an argon atmosphere. The pH was adjusted to 7.0 by adding 4 M NaOH and 71 mg of SLURP-1 [1-20] (MW: 2467.8, 28.8 μmol) were subsequently added. The pH was re-adjusted again to 6.9-7.0, and the mixture was stirred at room temperature for 10 h under an argon atmosphere. A fresh portion of TCEP (20 mM final concentration) was added and the mixture was stirred for another 20 min. The product was then filtered and purified by HPLC on a Phenomenex C18 column (22 × 250, 5 μm, 300 ). It yielded 142 mg (16 μmol, 69%) of the fully reduced 81-mer polypeptide (>95% purity). In vitro protein folding and disulfide formation. For in vitro folding and disulfide bond formation, 25 mg (2.8 μmol) of the purified and fully reduced peptide were dissolved in 5 mL of 6 M GdmHCl to give a concentration of 5 mg/mL. Folding was carried out at 4 °C and initiated by rapid 1:40 dilution of the peptide solution with folding buffer (100 mM Tris, 2.0 M urea, 0.5 M arginine, 4 mM reduced glutathione, 1 mM oxidized glutathione, adjusted to pH 8.0 at 4 °C with conc. HCl). The reaction was left at 4 °C and stirred for 3 days after which the mixture was acidified with TFA to give a pH of ~4, filtered and purified by HPLC on a Zorbax C3 column (10 × 250, 3 μm, 300 ). The mass of synthetic SLURP-1 was determined by high-resolution ESI-MS on an AB SCIEX 5600 Triple-TOF mass spectrometer equipped with a nanoelectrospray ionization source. SLURP-1 observed mass 8837.1 ± 0.1 Da; calculated mass 8837.02 Da (monoisotopic mass). Isolated yield: 11.4 mg (1.3 μmol, 46%). NMR analysis. Synthetic SLURP-1(4 mg) was dissolved in 500 μL of 90% H 2 O/10% D 2 O solution and adjusted to pH 4.8 by adding 1 M NaOH. 2D 1 H-1 H TOCSY, NOESY as well as 1 H-15 N HSQC spectra were collected at 310 K using a 600 MHz Bruker spectrometer equipped with a cryogenically cooled probe. All spectra were recorded with an interscan delay of 1.0 s. The NOESY and TOCSY mixing times were 200 ms and 80 ms, respectively. Standard Bruker pulse sequences were used with WATERGATE for solvent suppression. NMR data were processed using Topspin (Bruker) and analyzed by CCPNMR 34 . Molecular modeling. SLURP-1 was modeled by removing the N-terminal methionine residue from the PDB 2MUO structure using UCSF Chimera 35 . Both sSLURP-1 and rSLURP-1 structures were subjected to consequent equilibration 100 ns NVT (constant number of particles, volume and temperature) and 100 ns NPT (constant number of particles, pressure and temperature) ensemble simulations with constrained heavy atoms followed by 50 ns unconstrained molecular dynamics simulations using the GROMACS-5.0.4 package (reference temperature 310 K, 100 mM NaCl). three individual frames from each of these two simulations were used as starting structures for 5 ns unconstrained molecular dynamics simulations to confirm reproducibility of results. ## Electrophysiology. In vitro cRNA synthesis. Plasmid pMXT construct of human nAChR α7 and plasmid pSP64 construct of human nAChR α4 were linearized with BamHI, and plasmid pT7TS constructs of human nAChR α3, α9, α10, β2, and β4 were linearized with XbaI restriction enzymes (NEB, Ipswich, MA, USA). Plasmid pcDNA3.1/Hygro(+) construct of rat nAChR α7 was linearized using XbaI, and plasmid pSGEM constructs of rat nAChR α9 and α10 were linearized using NheI restriction enzymes (Promega, Madison, WI, USA). All linearized plasmid constructs were subjected to in vitro cRNA transcription using SP6 (human nAChR α7 and α4) and T7 (human nAChR α3, α9, α10, β2, and β4, and rat nAChR α7, α9 and α10) mMessage mMachine ® transcription kits (AMBION, Foster City, CA, USA). Oocyte preparation and microinjection. Stage V-VI oocytes (Dumont's classification; 1200-1300 μm in diameter) were obtained from Xenopus laevis, defolliculated with 1.5 mg/mL collagenase Type II (Worthington Biochemical Corp., Lakewood, NJ, USA) at room temperature (21-24 °C) for 1-2 h in OR-2 solution containing (in mM) 82.5 NaCl, 2 KCl, 1 MgCl 2 and 5 HEPES at pH 7.4. Oocytes were injected with 5 ng of human nAChR α3β2, α3β4, α4β2, α4β4 or α7 cRNAs, 35 ng of human nAChR α9α10 cRNA or 9 ng of rat nAChR α7 and α9α10 cRNA (concentration confirmed spectrophotometrically and by gel electrophoresis) using glass pipettes pulled from glass capillaries (3-000-203 GX, Drummond Scientific Co., Broomall, PA, USA). Oocytes were incubated at 18 °C in sterile ND96 solution composed of (in mM) 96 NaCl, 2 KCl, 1 CaCl 2 , 1 MgCl 2 and 5 HEPES at pH 7.4, supplemented with 5% fetal bovine serum (FBS), 50 mg/L gentamicin (GIBCO, Grand Island, NY, USA) and 10000 U/mL penicillin-streptomycin (GIBCO, Grand Island, NY, USA). All procedures were approved by the University of Sydney Animal Ethics Committee and were performed in accordance with the Australian code of practice for the care and use of animals for scientific purposes (8 th edition, 2013). Oocyte two-electrode voltage clamp recording and data analysis. Electrophysiological recordings were carried out 2-7 days post cRNA microinjection. Two-electrode voltage clamp recording of X. laevis oocytes expressing human nAChRs was performed at room temperature (21-24 °C) using a GeneClamp 500B amplifier and pClamp9 software interface (Molecular Devices, Sunnyvale, CA, USA) at a holding potential −80 mV. For rat nAChRs, electrophysiological recordings were made using turbo TEC-03X amplifier (NPI Electronic, Germany) and WinWCP recording software (University of Strathclyde, UK), at a holding potential −60 mV. Voltage-recording and current-injecting electrodes were pulled from GC150T-7.5 borosilicate glass (Harvard Apparatus, Holliston, MA, USA) and filled with 3 M KCl, giving resistances of 0.3-1 MΩ. Oocytes expressing human nAChR α9α10 were incubated with 100 μM BAPTA-AM (Sigma-Aldrich, St. Louis, MO, USA) at 18 °C for ~3 h before recording and perfused with ND115 solution containing (in mM): 115 NaCl, 2.5 KCl, 1.8 CaCl 2 , and 10 HEPES at pH 7.4. Oocytes expressing rat nAChRs were perfused with Ba 2+ Ringer's solution containing (in mM) (115 NaCl, 2.5 KCl, 1.8 BaCl 2 , 10 HEPES at pH 7.2), whereas other human nAChR-expressing oocytes were perfused with ND96 solution. All oocytes were perfused at a rate of 2 mL/min in an OPC-1 perfusion chamber of < 20 µL volume (Automate Scientific, Berkeley, CA, USA). Initially, oocytes were briefly washed with bath solution (ND96/ND115/Ba 2+ Ringer's solution) followed by 3 applications of ACh using a HPLC injector with a 50 µL sample loop. Washout with bath solution was done for 3 min between ACh applications. Oocytes were incubated with sSLURP-1 for 5 min with the perfusion system turned off, followed by co-application of ACh and sSLURP-1 with flowing bath solution. All sSLURP-1 solutions were prepared in ND96/ND115 + 0.1% bovine serum albumin (BSA), except for sSLURP-1 in Ba 2+ Ringer's solution. Peak current amplitudes before (ACh alone) and after (ACh + sSLURP-1) sSLURP-1 incubation were measured using Clampfit 10.7 software (Molecular Devices, Sunnyvale, CA, USA) or WinWCP software (University of Strathclyde, UK), where the ratio of ACh + sSLURP-1-evoked current amplitude to ACh alone-evoked current amplitude was used to assess the activity of sSLURP-1 at nAChRs. All electrophysiological data were pooled (n = 3 to 14) and represent means ± standard error of the mean (SEM). Data analysis was performed using GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA). Data sets were compared using unpaired two-tailed Student's t-test. Differences were regarded statistically significant when p < 0.05. The IC 50 was determined from concentration-response curve fitted to a non-linear regression function and reported with error of the fit. Calcium imaging of SLURP-1 interaction with α7 nAChR. Mouse neuroblastoma Neuro2a cells were cultured in Dulbecco's modified Eagle's medium (DMEM, PanEco, Russia) supplemented with 10% FBS (PAA Laboratories, Austria). Cells were sub-cultured 24 h before transfection and were plated at density of 10,000 cells per well (black 96-well plate, Corning, USA), followed by Lipofectamine (Invitrogen, USA) -mediated transient co-transfection of hα7 nAChR-pCEP4, fluorescent calcium sensor pCase12-cyto (Evrogen, Russia) and chaperone Ric3-pCMV6-XL5 or NACHO TMEM35-pCMV6-XL5 plasmid constructs (OriGene, USA). Mouse muscle α1, β1, δ and ε nAChR-pRBG4 plasmid constructs were expressed similarly, but without a chaperone. Transfected Neuro2a cells were grown at 37 °C in 5% CO 2 -incubator for 48-72 h, then medium was removed and the cells were washed with external buffer containing (in mM) 140 NaCl, 2 CaCl 2 , 2.8 KCl, 4 MgCl 2 , 20 HEPES, 10 glucose at pH 7.4. Cells were pre-incubated with 5 μM sSLURP-1 for 20 min at room temperature before agonist addition (ACh or epibatidine (Tocris, UK)). To potentiate α7 nAChR response, PNU120596 (10 μM) was added to the pre-incubation solution. Cells were excited at 485 nm and emitted fluorescence was detected at 535 ± 10 nm, using a multimodal microplate reader Hidex Sense (Hidex, Turku, Finland). Fluorescence was recorded every 2 s for 3 min following agonist addition. Responses were measured as peak intensity minus basal fluorescence level, and are expressed as a percentage of a maximal response obtained to agonist. Data files were analyzed using HidexSence software (Hidex, Turku, Finland) and OriginPro 7.5 software (OriginLab, MA, USA, for statistical analysis). Negative controls were run in the presence of 4 μM α-Cbt. Radioligand assay of sSLURP-1 binding to AChBPs and nAChRs. In competition experiments with [ 125 I]-α-Bgt, sSLURP-1 (1-100 μM) was pre-incubated 3 h at room temperature with AChBPs (L. stagnalis AChBP or A. californica AChBP at final concentrations of 2.4 nM, and 140 nM, respectively) or nAChRs (hα7 nAChR-expressing GH 4 C 1 cells or T. californica electric organ membranes at a final concentration of toxin-binding sites of 0.4 nM and 1.25 nM, respectively (measured using [ 125 I]-α-Bgt)), in 50 μL buffer consisting of 20 mM Tris-HCl and 1 mg/mL BSA, pH 8.0 (binding buffer). Radioiodinated α-Bgt was added to a final concentration of 0.2 nM, and the mixture was incubated for 5 min. Binding was stopped by rapid filtration on double DE-81 filters (Whatman, Maidstone, UK) pre-soaked in binding buffer (for AChBPs) or GF/C filters (Whatman, Maidstone, UK) pre-soaked in 0.25% polyethylenimine (for GH 4 C 1 cells and T. californica electric organ membranes), unbound radioactivity was removed from the filters by washout (3 × 3 mL) with the binding buffer. Non-specific binding was determined in all cases using 3 h pre-incubation with 10 μM α-Cbt.

chemsum

{"title": "Interaction of Synthetic Human SLURP-1 with the Nicotinic Acetylcholine Receptors", "journal": "Scientific Reports - Nature"}

kinetic_study_of_disulfonimide_catalyzed_cyanosilylation_of_aldehyde_using_a_method_of_progress_rate

## Abstract: Kinetic study of organic reactions, especially multistep catalytic reactions, is crucial to in-depth understanding of reaction mechanisms. Here we report our kinetic study of the chiral disulfonimide catalyzed cyanosilylation of aldehyde, which reveals that two molecules of TMSCN are involved in the rate-determining C-C bond forming step. In addition, the apparent activation energy, enthalpy of activation and entropy of activation were deduced through the study of temperature dependence of the reaction rates. More importantly, a novel and efficient method which makes use of the progress rates was developed to treat the kinetic data obtained from continuous monitoring of the reaction progress with in situ FT-IR. Scheme 1 Chiral disulfonimides catalyzed asymmetric cyanosilylation of 2-naphthaldehyde 2. Experimental kinetic studies were carried out by monitoring of reaction progress with in situ FT-IR (see Supporting Information for details). To determine the reaction orders of all components, several reactions were carried out under identical conditions, only varying the initial concentrations of reactants 2 and 3, and the catalyst 1a, respectively. From the data obtained by in situ FT-IR measurements, concentration of aldehyde vs time profiles with different initial concentrations of TMSCN 0 were obtained as shown in Figure 1a. Profiles of vs time with different initial concentrations of aldehyde 0 and catalyst loading [1a]0 were also obtained (Figure S13, S16, Supporting Information). Reaction progress kinetic analysis (RPKA) is a methodology developed and formalized by Blackmond and co-workers. 9 Compared with the classical kinetic approach (method of pseudozero-order) where the concentration of one substrate is artificially fixed at a pseu-do-constant high value (usually tenfold), RPKA allows reactions to be carried out at synthetically relevant conditions which are closer to standard reaction conditions and more reasonable. One of the key points of RPKA is to determine the reaction orders in "a trial and error procedure" by constructing "graphical rate equations" and seeking "overlay" of them by dividing the rate curves by the concentration of the substrate under study taken to the power of the reaction order. We first evaluated our kinetic data using the method of RPKA. The concentration of aldehyde vs time profiles (Figure 1a) with different initial concentrations of TMSCN 0 were converted to rate vs profiles (Figure 1b) which clearly indicate a positive reaction order in as the rate significantly increased upon increasing the concentration of 3. Then the rate vs profiles (Figure 1b) were converted to rate/ vs profiles (Figure S8, Supporting Information), however no sign of "overlay" between these "graphical rate equations" was observed. When the rate vs profiles (Figure 1b) were further converted to rate/ 2 vs profiles as shown in Figure 1c, the "graphical rate equations" got much closer to each other especially in the middle range of the reaction progress ( = 0.075 -0.15 M). While it is still difficult to judge whether these "graphical rate equations" overlay or not. Although RPKA has been proved to be a powerful method 2,9 to deduce the reaction orders (mainly integer numbers such as 0 and 1) of the components participated in the reaction and to determine whether there is catalyst activation or deactivation, substrate or product inhibition or acceleration, it doesn't work quite well when the reaction mechanism is more complex and the reaction equation is more complicated (the orders could be noninteger even negative if the overall rate expression for the reaction is written in power-law form). As already mentioned by Blackmond,9a in some reactions "it may be found that none of the plots of graphical rate equations result in all the curves falling on top of one another". In 2016, Burés developed a very simple and practical graphical method using a normalized time scale, to determine the order in catalyst. 10 However, this method is limited to the order in the catalyst concentration which is not a thermodynamic driving force of the reaction. Besides it, the orders of the reactants also are supposed to be determined in most kinetic studies. Seeking to make full use of the kinetic data obtained from the steady state catalytic cycles of the entire reaction progress, and deduce the reaction orders of all components in a more efficient and convenient way, we developed a novel method to treat the kinetic data. Taking the determination of the order of TMSCN as an example, the detailed procedures of this method are illustrated as follows (see Supporting Information for details): Step 1. Obtain the rate vs [reactant A] profiles. Several reactions were carried out under identical conditions only varying the initial concentrations of reactant B, whose order is to be determined (TMSCN 3 in this case). The profiles of rate vs aldehyde concentration Figure 1b (rate vs ) were obtained from the vs time profiles, which were deduced from the data sets of in situ FT-IR measurements. Step 2. Fit the rate vs [reactant A] profiles and get the functions. An accurate function (such as high order polynomial function) was used to fit the curves of each reaction in the rate vs profiles (Figure 2a, blue lines). Step 3. Obtain the data sets of (rate, [reactant B]). A series of concentrations of 2 (reactant A) with a fixed interval (here 0.01 M) and within a selected range (here 0.08-0.16 M), were used to calculate the instant progress rates from the fitting functions and the instant concentrations of 3 (reactant B) corresponding to each . The obtained data sets of (rate, ) were shown as red squares in Figure 2a. Step 5. Plot the profile of (order of [reactant B]) vs [reactant A]. The profile of (order of ) vs was plotted (Figure 2c), which not only shows an approximate value for (order of ) but also indicate changes in reaction order as a function of changing substrate concentrations. This method may look complex when described in steps, while actually all fitting and data processing can easily be done using standard office software such as Excel and Origin. The average value for the order of was calculated to be 1.94 (nearly second order). Thus, the apparent rate order of 3 in powerlaw form reaction rate equation was obtained. The reaction order of approximately 2 in TMSCN suggests that two molecules of this substrate are involved in a step which has a significant influence on the rate. With the method described above, the average reaction order of aldehyde 2 was determined to be only 0.17 which nearly is zero order. The average value for the order of catalyst 1a was calculated to be 1.23 which is close to first order (see Supporting Information for details). Taking the average reaction orders determined by our method, the power-law form of the rate equation, which reflects the molecular-level behavior of the reaction as an empirical approximation, can be stated as shown in equation ( 1). The low reaction order in 2 suggests that the step involving its activation is significantly faster than the addition of cyanide and a catalyst species associated with aldehyde 2 is possibly involved in the rate-limiting step. The temperature dependence of the reaction rates was studied in the range of 273.15 -303.15 K under otherwise identical conditions. The apparent activation energy of the reaction was deduced to be 41 kJ mol -1 (9.9 kcal mol -1 ) according to the Arrhenius equation by plotting ln k vs 1/T (Figure S20,21, Supporting Information), which implies that the reaction is relatively sensitive to temperature. The enthalpy of activation ΔH ‡ was deduced to be 39 kJ mol -1 (9.3 kcal mol -1 ) and the entropy of activation ΔS ‡ was deduced to be -81 J mol -1 K -1 (-19 cal mol -1 K -1 ) according to the Eyring equation by plotting ln (k/T) vs 1/T. The Gibbs energy of activation ΔG ‡ was calculated to be 61 kJ mol -1 (15 kcal mol -1 ) at 273.15 K (Figure S22-24, Supporting Information). Based on these studies, we propose a catalytic cycle for the disulfonimide catalyzed asymmetric cyanosilylation of aldehydes (Scheme 2). After a long dormant period 8 which is up to a few hours, the pre-catalytic cycle ends. Then the Brønsted acid precatalyst 1a reacts with TMSCN to generate the catalytically active Lewis acid organocatalyst 1a-TMS which interacts with the aldehyde 2 to generate activated species 5. The low reaction order in 2 suggests possible saturation kinetics in , indicating that the formation of 5 is significantly faster than the reverse reaction and the addition of cyanide. Subsequently, two molecules of TMSCN interact with 5, possibly forming the new C-C bond via aggregated cyclic transition state as shown in 6, to produce species 7 and regenerate one molecule of TMSCN, which is proposed to be the rate-determining step. This is similar to the well-known Grignard reaction, which proceeds through an aggregated six-membered ring transition state bridged by a dimeric di-cation of Grignard reagent. 11 A study of the relationship between the enantiomeric excess of the product and the enantiomeric excess of the catalyst revealed that there is no nonlinear effect 12 in this reaction (Figure S27, Supporting Information), which is consistent with the involvement of only a single catalyst molecule in the stereo-determining step. At last, product 4 is quickly released from 7 and active catalyst 1a-TMS is regenerated. Scheme 2 Proposed catalytic cycle. Two molecules of TMSCN are possibly involved in the rate-determining C-C bond forming step. In summary, kinetic investigation of disulfonimide catalyzed cyanosilylation of aldehyde was conducted and the orders of the reactants and catalyst for a power-law form of rate equation were obtained. An aggregated cyclic transition state with two molecules of TMSCN involved, was proposed. A novel and efficient method which makes use of the progress rates to deduce orders of both reactants and catalyst, was developed to treat the kinetic data obtained from continuous monitoring of the reaction progress, and is expected to attract widespread attention. We predict that these studies may not only facilitate in-depth understanding of the reaction mechanism, but also benefit the future design and application of powerful organocatalysts.

chemsum

{"title": "Kinetic Study of Disulfonimide Catalyzed Cyanosilylation of Aldehyde Using a Method of Progress Rates", "journal": "ChemRxiv"}

smart_molecular/mos2_heterostructures_featuring_light_and_thermally-induced_strain_driven_by_spin_sw

## Abstract: In this work we exploit the ability of spin-crossover molecules to switch between two spin states, upon the application of external stimuli, to prepare smart molecular/2D heterostructures.Through the chemical design of the hybrid interface, that involves a covalent grafting between the two components, we obtain a hybrid heterostructure formed by spin-crossover nanoparticles anchored on chemically functionalized monolayers of semiconducting MoS2. In the resulting hybrid, the strain generated by the molecular system over the MoS2 layer, as a consequence of a thermal or light-induced spin switching, results in a dramatic and reversible change of its electrical and optical properties. This novel class of smart molecular/2D heterostructures could open the way towards a novel generation of hybrid multifunctional materials and devices of direct application in highly topical fields like electronics, spintronics or molecular sensing. The research on graphene and other two-dimensional (2D) materials has been propelled by the possibility of studying and exploiting the properties of matter in the 2D limit. 1 Nowadays, this topic is moving towards the assembly of monolayers of different types to afford van der Waals heterostructures. In the 2D area, apart from the chemical functionalization of 2D materials, 5 the use of molecules as precursors, constituents or functional components of novel 2D systems and heterostructures, has been scarcely investigated. 6,7 An interesting possibility in this context deals with the fabrication of mixed molecular/2D heterostructures, in which the properties of the "all surface" 2D layer can be tuned by the hybrid interface, i.e., by the interactions established between the molecules and the 2D material. 8,9 A versatile family of 2D layered materials in which this molecular/2D concept can be exploited, is that formed by transition metal dichalcogenides (TMDCs) of formula MX2 (M = metals traditionally restricted to groups IV-VII; X = S, Se, Te). The members of this family display a wide range of electronic properties as a function of their composition and structures, including insulating, semiconducting, metallic and superconducting properties. Among them, the most deeply studied system is MoS2, owing to its chemical stability, and electronic properties. The weak interaction between stacked layers makes it feasible to isolate monolayers of this material, create van der Waals heterostructures and exploit them to design hybrid structures and electronic devices displaying new low-dimensional physics and unique functionalities. 13 Of special interest is the correlation between the electronic structure and the structural arrangement of S atoms in the monolayer. In fact, by a slight gliding of the S atoms, a switching between a trigonal prismatic symmetry around the Mo atom (2H-phase) and an octahedral one (1T-phase) occurs, giving rise to an electronic modulation of the 2D material from semiconducting to metallic. 14 This phase transition from 2H to 1T can be induced by different external stimuli such as chemical modification 15,16 or electron beam radiation, 17 while the inverse process can be achieved by thermal treatment, 18 infrared radiation, 19 or aging of the material. 20 An attractive possibility in this context is to induce the phase transition by applying strain. This has been theoretically predicted 21,22 and experimentally demonstrated in ultrathin MoS2 layers in which a band gap narrowing under low tensile strain has been achieved, which enables the reversible tuning of their optical and electronic properties. 23,24 Sophisticated devices have been reported to prove the strain modulation of both the optical and electronic band gaps, through photoluminescence (PL), Raman spectroscopy, and transport measurements. 27,28 In addition, an intrinsic negative piezoresistivity has been observed 24 and exploited to develop a piezoresistive composite in which the resistance decreases when the strain increases (negative Gauge factor). 29 In this scenario, we have envisioned the possibility of preparing MoS2-based heterostructures where the second component is an active molecular system that can directly and reversibly tune the strain applied on the 2D material and, therefore, its electronic structure and electric conductivity. We have chosen switchable molecular-based spin-crossover (SCO) materials for this purpose. These materials are, in most cases, Fe(II) compounds that undergo spin transition between low spin (LS with S=0) and high spin (HS with S=2) configurations upon the application of an external stimulus such as light, temperature, pressure or chemical activation. This spin transition comes along with an intrinsic change in volume (up to 11.5%) 30 and a variation of their mechanical, 31 magnetic, 32 electrical 33 and optical properties. 34 Interesting devices 35 and composites based on SCO materials embedded in different polymers have been already reported where a change in volume upon spin transition induces strain effects on a second component modifying their electrical and/or mechanical properties. However, in the field of 2D materials, the inducement of strain by phase transition materials has not been demonstrated so far. The first example of SCO/2D heterostructure was based on the deposition of SCO nanoparticles on graphene. In this case, the thermal spin transition of the nanoparticles resulted in significant changes in the transport properties of graphene. However, the origin of these electrical changes was electronic rather than mechanical. In fact, it was ascribed to the interaction of the graphene electrons with the phonons of the SCO nanoparticles. 39 Just very recently, an inverse approach has been followed, consisting in the deposition of a mechanically-exfoliated graphene layer on a large SCO crystal with a polymeric spacer in between. Nevertheless, in this transistor geometry, the mechanical influence of the SCO crystal over the properties of the 2D layer is attenuated by the presence of the spacer, thus resulting in a small modification of graphene conductivity, mainly attributed to changes in the induced electrostatic potential. 40 In the present work, we focus on SCO nanoparticles as phase transition material to induce a strain on the 2D material and thus modulate its electrical and optical properties. These nanoparticles are based on the [Fe(Htrz)2(trz)](BF4) coordination polymer (Htrz = 1,2,4-triazole and trz = triazolate) covered with a silica shell (from now on SCO-NPs). 41,42 These core/shell nanoparticles have shown to undergo a cooperative spin transition near room temperature and, depending on the purpose, their sizes and composition can be tuned. 43,44 In addition, the silica shell not only preserves the chemical stability of the SCO system, but also provides an anchoring point for their further chemical functionalization, giving rise to robust heterostructures. 41 ## Results Chemical design of a hybrid SCO/MoS2 heterostructure. In order to prepare the hybrid SCO/MoS2 heterostructures, a straightforward solution grafting protocol that coats MoS2 flakes with pre-synthesized SCO-NPs has been developed. This involves the chemical functionalization with 3-iodopropyl(trimethoxysilane) (IPTS) of ultrathin chemically exfoliated MoS2 layers (CE-MoS2), followed by the anchoring of the SCO-NPs to these functionalized layers via a covalent bond between the trimethoxysilane group and the silica shell 41,45 (Fig. 1). In more detail, this protocol starts with the preparation of CE-MoS2 layers by reacting bulk 2H-MoS2 with n-butyllithium (n-BuLi) used as reducing species. 46 During this process an electron transfer to MoS2 occurs, giving rise to a partial structural reorientation and an electronic band structure modification, that triggers the transition from the semiconducting 2H to the metallic 1T-phase, 14,47 as confirmed by X-ray photoelectron (XPS) and Raman spectroscopies (Fig. 2 and Table S1). After re-dispersion, the CE-MoS2 flakes retain an excess of negative charge on their surfaces (Fig. S1), 48 providing high colloidal stability and facilitating their ulterior covalent functionalization. On the other hand, core-shell SCO-NPs are synthesized following the protocol reported by some of us 44 (see Methods). In order to functionalize the MoS2 layers, a CE-MoS2 suspension is mixed with a 3iodopropyl(trimethoxysilane) (IPTS) solution. The excess of negative charge accumulated in CE-MoS2 facilitates the nucleophilic attack to IPTS, resulting in the displacement of I and the formation of new covalent S-C bonds. 49 This gives rise to MoS2 flakes decorated with propyltrimethoxysilane groups (PTS-MoS2) (Fig. 1, step i). The successful anchoring of the IPTS onto the MoS2 surface is evidenced by a clear decrease in the measured x-potential value (from -30 mV to -4 mV, Fig. S1). 49 XPS confirms that functionalization takes place without modification of Mo and S oxidation states (Fig. 2a and b). However, during the process, CE-MoS2 S2p, and Mo3d peaks blue shift ~1 eV (Table S1), suggesting a phase transition from the metallic 1T into the semiconducting 2H-phase. This conversion is confirmed in PTS-MoS2 Raman spectrum where J peaks at 152, 232 and 326 cm -1 (fingerprint of the metallic phase) 50 , decrease substantially at the expense of E 1 2g and A1g peaks (Fig. 2c). Moreover, PL signal is restored, further supporting the conversion to the semiconducting phase 50,51 (Fig. 2d). These results contrast with those reported by Chhowalla and coworkers where 1T-phase was preserved after functionalization but with unusual semiconducting properties and intense photoluminescence recovery. 49 Such a difference may be related with our higher degree of functionalization of the MoS2 units (~30 % according to thermogravimetric analysis; Fig. S2), which could be originated in our higher ratio of 1T-phase in the starting CE-MoS2 (~85 %). Spectroscopic studies are indicative of the formation of a covalent bond between the MoS2 and the IPTS. C-S vibration is expected at ~690 cm -1 in the FTIR spectra. Although a signal is observed in this region, unfortunately, it overlaps with ethanol and IPTS vibrations, hindering a definitive assignation of this peak. Nevertheless, the appearance of an additional contribution at ~164 eV in the XPS S2p signal that modifies the deconvolution analysis, points out to functionalization of sulphur. 49 Intensity / a.u. Intensity / a.u. ## Photon energy / eV from energy-dispersive X-ray spectroscopy (EDAX) analysis performed on the thinnest flakes by a high resolution transmission electron microscope (HR-TEM). These results show the presence of silicon and the absence of iodine, while confirming that the integrity of the flakes is maintained after functionalization (Fig. S3). In order to anchor the SCO-NPs to the functionalized layers forming the final heterostructure (SCO/MoS2), the methoxysilane groups decorating PTS-MoS2 are used to graft the SCO-NPs silica shell by mixing both suspensions 41,45 (Fig. 1, step ii). In this way, we obtain MoS2 functionalized with SCO-NPs bearing two aspect ratios (70 x 40 nm or 40 x 40 nm, Fig. S4, 44 SCO/MoS2-1 and SCO/MoS2-2a respectively, Fig. S5) and three coverage degrees, accounted by the Fe:Mo ratios (See Methods, Fig. S6). The coexistence of the two components in the heterostructure is confirmed by HR-TEM (Fig. 3a and S5-S6) and XPS spectra, which exhibit the characteristic Fe2p3/2 and Fe2p1/2 peaks at ~709 eV and ~722 eV, respectively, coming from the SCO-NPs (Fig. S7) and similar Mo and S spectra to those recorded for PTS-MoS2 (Fig. 2a, b). Finally, the integrity of the SCO-NPs in the hybrid is demonstrated by the magnetic data, which show a SCO thermal hysteresis very similar to the bulk, with transition temperatures at ~380 K ( T1/2 up ) and ~340 K ( T1/2 down ). Fig. 3b and S8. ## Influence of the SCO-NPs on the properties of MoS2 in the hybrid SCO/MoS2 heterostructure. To study the potential of SCO-NPs to modify the electronic structure of MoS2 layers, we rely on electrical transport and PL measurements. In most of the SCO/MoS2 samples, we observe thermal hysteresis in the conductivities that follow well the SCO transitions observed from SQUID measurements. The strongest effects are obtained for heterostructures formed with SCO-NPs of 70 nm and a Fe:Mo ratio of 2, named SCO/MoS2-1. As reported in Fig. 4a, this sample shows a sharp increase of ca.900% in the conductance upon heating, coincident with the spin transition from LS to HS ( T1/2 up ≈ 370 K), and a sharp drop in conductance during the reversal cooling down process, corresponding to the transition from HS to LS ( T1/2 down ≈ 340 K) (Fig. 4a). Remarkably, it presents an opposite behavior compared to the one observed for SCO-NPs (Fig. 4b), where the nanoparticles are less conductive in the LS state than in the HS. Keeping the Fe:Mo ratio equal to 2 but using smaller SCO-NPs of 40 nm (SCO/MoS2-2a), the observed thermal variation in the electrical response is qualitatively similar, but with a smaller switch of ca.30% (Fig. S9ef). This result points out that, when the relative quantity of SCO-compound: MoS2 is maintained, there is a more relevant effect induced on the 2D material by the size and shape of the nanoparticles (whose axial elongation increases as their size does), than by their number. Moreover, in SCO/MoS2 hybrids the conductivity is up to 5 orders of magnitude larger than that of pure SCO-NPs (Fig. S9), overcoming the typical insulator character of the SCO compounds. This feature is also very important when dealing with the sample stability. In fact, a general problem encountered when measuring the transport in pure SCO-NPs is the high ## HS LS voltages required (100 V), leading to a fast sample degradation. 52,53 On the contrary, the higher conductivities of the heteroestructures allow to measure at much lower voltages, thus guaranteeing sample integrity and the switching properties over several thermal cycles. Additionally, we investigated the influence of the SCO-NPs/MoS2 ratio on the transport properties, while keeping the shape and size of the particles unmodified. When the Fe:Mo ratio is further decreased from 2 to 0.4, SCO/MoS2-2b, sample conductivity increases but the hysteretic effects coming from the SCO-NPs are completely lost, likely due to the low concentration of nanoparticles, Fig. S9g. Accordingly, sample conductivity decreases when the Fe:Mo ratio increases, and starting from a ratio of 5 (SCO/MoS2-2c) sample behavior is reversed and approaches the one observed in assemblies of pure SCO-NPs, where the LS state becomes more conductive than the HS state 54,55 (Fig. 4b and Fig. S9h). From this reversal behavior in the conductivity of SCO/MoS2 hybrids, we can conclude that for high concentrations of SCO-NPs, the charge transport is dominated by the nanoparticles. In contrast, when the concentration of the nanoparticles is decreased, the transport mainly occurs through the MoS2 flakes. These results suggest that the changes observed in the MoS2 flakes come from the strain generated by the SCO-NPs. In this mechanism, the volume change of the nanoparticles induced by the spin transition is expected to strain the flakes, resulting in an intrinsic modulation of the band structure of the layers, and thus modifying their conductivity (for further details and discussion see Supplementary pages 11-16). 29 Furthermore, because of the direct gap semiconductor nature of the MoS2 monolayers (2Hphase), it is possible to gain direct information on their band gap energy through PL measurements. 51 In fact, it is well known that when a tensive strain is applied to a MoS2 layer, its PL redshifts and weakens its intensity as consequence of a narrowing in the band gap and a transition in the semiconductor from direct to indirect band gap behavior occurs. 23,56 This is exactly the effect we have observed by Raman spectroscopy performed in the SCO/MoS2-1 hybrid in the two spin states (Fig. 5a). When the SCO-NPs are in the LS state (room temperature), a maximum in the PL signal (A peak) is observed at ~1.88 eV, which redshifts ~30-60 meV and decreases in intensity when the spin transition occurs (T > ~370). This variation can be monitored by measuring PL as a function of temperature, Figure S12. Complete cycles of heating and cooling are shown in Fig. 5b. We observe that the energy of the A peak presents a clear hysteretic behavior that completely resembles that of the spin state of the SCO-NPs. In the temperature range in which the spin state of the particles can be LS or HS, the A peak differs in intensity and energy at each temperature. This points out to a clear effect of the nanoparticles spin state on the MoS2 band structure, additional to the thermal one. 57 To further prove that this effect is induced by the SCO-NPs, we have performed a blank experiment on CE-MoS2 transformed into the semiconducting 2H-phase by thermal treatment (CE-MoS2/2H). 18 As can be observed in Fig. S13, the hysteretic behavior is absent in this case, being the A peak at each temperature completely independent of the heating or cooling path. Based on previous works, 56 the redshift observed in the SCO/MoS2-1 hybrid inside the temperature interval where the hysteresis loop is observed (i.e. ~ 40 meV at 355 K) corresponds to a ~0.6% of tensile strain, whereas the observed decrease of the PL intensity can be attributed to the increase in indirect band gap behaviour of the MoS2 as a consequence of the strain. Moreover, since the spin transition of the nanoparticles can be triggered by tuning the intensity of an irradiating light, the possibility of using an optical source to drive the PL of the SCO/MoS2-1 hybrid was evaluated. 58 As shown in Fig. 5c, we measured the A peak of the hybrid at two green laser excitation intensities, 0.08 mW (purple curve), and 0.8 mW (green curve). For the lower intensity, the SCO-NPs in the heterostructure are in the LS state while for the higher one they are expected to undergo a spin transition to the HS state (Fig. S14). Under these conditions, SCO/MoS2-1 spectra show that the A peak redshifts of about 60 meV and decreases in intensity upon the laser power increase, as previously observed for thermal spin transition in Fig. 5a. Hence, these measurements prove that we can also optically induce a spin transition at room temperature by increasing the power of the excitation laser. Comparing these results with the PL modulation thermally achieved, the shift of the A peak is clearly higher when a light source is used. This suggests a cooperative effect between the heating of the nanoparticles, due to laser irradiation, and the strain induced by the spin transition. Interestingly, when using CE-MoS2/2H samples for blank measurements, we observe that under these laser intensities (0.8 and 0.08 mW), the PL remains unaffected (Fig. 5d), suggesting that MoS2 is not directly heated by the laser in these experimental conditions, despite the fact that a thermal heating of the SCO-NPs is expected. 58 These studies thus demonstrate the possibility of inducing strain in the MoS2 by light irradiation, opening the door to the fast optical modulation of 2D material properties. Moreover, as far as the spin state is concerned, it is possible to sense the spin in these SCO-NPs by following the change in the MoS2 luminescence. This result is quite remarkable: on the one hand, because this kind of optical detection cannot be achieved in SCO/graphene heterostructures, 39 and, on the other, because we are providing a new tool for optical identification of spin states which is much more sensitive, simple and local than that obtained from transport measurements. ## Discussion We have reported here a two-step protocol in solution to chemically design smart molecular/2D heterostructures, formed by SCO-NPs covalently linked to semiconducting MoS2 flakes. In a properties. Notice that the present approach is radically different from those previously reported in this area, 25 in which the strain has been generated on pure MoS2 layers by different strategies like the direct application of pressure on a suspended layer, with an AFM tip; 24 by using substrates that can transfer strain to the MoS2 layer by a mechanical bending (in a flexible substrate) 23 or stretching (in an elastomeric substrate); 27 or by applying voltage (in a piezoelectric substrate), 59 or temperature (in a thermo-responsive substrate). 60 In contrast to all these cases, here the strain is generated in a single material, based on a chemically-designed hybrid heterostructure, either by a temperature modulation or by light irradiation. Hence, the intrinsic properties of the SCO component have opened the possibility for the first time, of using light as external stimulus to induce strain in TMDCs layers. In conclusion, these innovative results demonstrate the fabrication of a multifunctional material where properties of the two components have been reciprocally boosted. Compared to pure SCO-NPs, MoS2 confers to this hybrid luminescence and higher resilience and conductance. Concurrently, MoS2 gains new degrees of freedom thanks to SCO-NPs spin which can be addressed by temperature or light and read-out electrically or optically. All in all, a new smart heterostructure which mimics the bistability displayed by spin-crossover materials but ignores the typical drawbacks of these compounds have been synthesized. Interestingly, our approach could be easily expanded to other 2D materials offering a yet unexplored modulation of their properties and opening new frontiers for strain engineering, towards their application in multifunctional devices for beyond conventional electronics. For the synthesis of the analogue hybrid with smaller size nanoparticles, at different Fe:Mo ratios, an equivalent procedure was followed but adding 40 nm SCO-NPs suspension in 10, 1, or 20 mg•ml -1 for: SCO/MoS2-2a, SCO/MoS2-2b and SCO/MoS2-2c, respectively. ICP-OES Fe:Mo ratios of 2:1 for SCO/MoS2-2a, 0.

chemsum

{"title": "Smart molecular/MoS2 heterostructures featuring light and thermally-induced strain driven by spin switching", "journal": "ChemRxiv"}

transferable_neural_network_potential_energy_surfaces_for_closed-shell_organic_molecules:_extension_

## Abstract: Transferable high dimensional neural network potentials (HDNNP) have shown great promise as an avenue to increase the accuracy and domain of applicability of existing atomistic force fields for organic systems relevant to life science. We have previously reported such a potential (Schrödinger-ANI) that has broad coverage of druglike molecules. We extend that work here to cover ionic and zwitterionic druglike molecules expected to be relevant to drug discovery research activities. We report a novel HDNNP architecture, which we call QRNN, that predicts atomic charges and uses these charges as descriptors in an energy model which delivers conformational energies within chemical accuracy when measured against the reference theory it is trained to. Further, we find that delta learning based on a semi-empirical level of theory approximately halves the errors. We test the models on torsion energy profiles, relative conformational energies, geometric parameters and relative tautomer errors. ## Introduction Over the last decade, techniques borrowed from the field of machine learning (ML) have greatly impacted many fields within computational chemistry. No field seems to be safe from the intrusion, even the historically empiricism-averse field of ab initio quantum chemistry. Techniques have emerged at nearly every level, between solving the electronic Schrödinger equation using a Neural Network (NN) based ansatz of the many-electron wave function, 1,2 machine learned density functionals, 3,4 empirical corrections of semi-empirical, density functional, or Hartree-Fock energies to higher level theories, machine learned force fields, and property prediction as complex as chemical reactivity. 12 Of particular interest to this work is what we might refer to as a NN potential energy surface (NN-PES). The goal of a NN-PES is to compute the potential energy of a chemical system, given the atomic positions. These models are typically trained to reproduce a particular model chemistry and are expected to reproduce the total electronic energy to chemical accuracy. 5, A NN-PES can be considered a type of atomistic force field: it maps atomic coordinates to energies, it is comparatively efficient to compute, and its parameters are empirically determined. However, a NN-PES differs from traditional biomolecular force fields 18 in that the total electronic energy is reproduced, as opposed to the energy relative to an arbitrary reference conformation. Additionally, the energy is purely a function of the chemical elements, coordinates, and net charge (as in an ab initio method), not relying on additional discontinuous input data such as atom types or assigned bonds. As such, NN-PES models promise both to increase the accuracy of empirical force fields and to expand their application domain to include important processes such as chemical reactions. The most common approach to construct a NN-PES is to first transform the atomic coordinates into local atomic descriptors, or "features", which describe the local environment of each atom. Alternatively some models allow the features to be "learnable" parameters, typically referred to as an embedding. 9,10 These features are transformed and reduced in dimensionality by a machine learning method to produce an energy. The parameters of the model are determined by minimizing the error of predicted energies relative to a selected reference level of theory, typically density functional theory (DFT). In one of the simplest forms, and the form we focus on here, the feature vector describing each atom's environments is independently transformed by a neural network to output an atomic energy; these atomic energies are then summed to produce a molecular energy. This approach is often termed a high dimensional neural network potential (HDNNP). 19,22 Many of the works that have applied this algorithm have focused on applications in which the HDNNP is trained to the same chemical system to which it is applied. In this application, a model is trained very accurately within a small and well defined chemical space and then used to perform energy sampling within this chemical space, on larger simulation cells or time scales than would otherwise be possible. There is no expectation that the model could be applied to systems for which it is not trained. Conversely, it is also possible to develop a transferable HDNNP in which it is expected that the model will be applied to chemical systems for which it is not trained. Smith et al. demonstrated that an HDNNP, trained only to a representative, but still large, subset of the vast diversity of organic molecules, could produce accurate energies for molecules not in the training set. 8 A transferable HDNNP rests on two pillars: A model with the capacity to reproduce energies to within chemical accuracy, and a dataset which contains a representative sample of the space of relevant atomic environments. Previously, some of the current authors have extended the work of the Roitberg lab to increase element coverage 23 (which was initially only four elements but has recently increased to seven 24 ) and increased the precision when tested on rotamer scans of a diverse set of druglike organic molecules. The motivation for that work was to develop an HDNNP that could be confidently used to generate training data for parameterization of intramolecular terms in a biomolecular forcefield such as OPLS. 18 We termed this model Schrödinger-ANI (here abbreviated SANI) in homage to the ANI model from which it descends. There are several shortcomings of that model, the most significant for parameterization of force fields is that molecular ions were not in the domain of applicability. SANI and other models like it have no way to distinguish charged from uncharged species, nor did we have a dataset with sufficient coverage for such training. Here we report our solutions to both of these problems. In what follows, we will describe a model called QRNN (charge recursive NN), trained directly to DFT, and QRNN-TB, trained to the difference between the energies of DFT and a lower level, highly parameterized, quantum mechanical theory which is efficient to compute. In this work we will utilize the GFN2-xTB method 25 for this purpose -a tight binding DFT methodology which has been broadly parameterized and shown to have good performance when compared to other semi-empirical methodologies, such as PM7, 26 to which we will also compare. In addition, we will compare to a version (QeqNN) with a charge equilibration method like that of Ko et al, 15 which has similar accuracy but less favorable computational scaling. We will demonstrate the effectiveness of these models by testing conformational energies of a broad set of neutral and ionic systems, as well as the Hutchison conformer test set 27 and relative energies of tautomers in the Tautobase dataset. 28 ## Models The main issue to be resolved is that the electronic energy is not uniquely defined by the nuclear positions alone: one must also define the system's net charge and spin multiplicity. In a model such as SANI these two electronic inputs are implicitly defined to be zero and one (neutral closed-shell), simply by having only neutral closed-shell examples in the training set. It is tempting to add ionic examples without modifying the model and to hope that the model can interpret which of these systems are ionic and which not. Most chemists could guess on sight that a deprotonated carboxylic acid will be negatively charged, or a protonated amine will be positively charged. It is possible that one could proactively curate a dataset of ions such that the charge state can be inferred from coordinates alone, but we believe this approach is flawed and will eventually fail as the training set achieves broad coverage of chemical space. A simple example of the problem is any tertiary carbocation, which (for a given set of coordinates) would be indistinguishable from a tertiary carbanion. Any model that has features only depending on nuclear positions would unavoidably fail to distinguish these systems. Thus, one must specify the net charge to a model if the training set has broad coverage of geometries of closed shell ions. Here we do not consider systems that are not closed-shell, and thus the spin multiplicity continues to be defined as one for all inputs. The crucial idea of the models we study in this report is to use a simple physical model 29 to predict atomic charges, and then to use these charges as inputs to an energy model along with the usual geometric features. Essentially we transform the global net charge into local charges that can be used as part of description of the local atomic environment in a natural way. While there are numerous works which have used NN predicted atomic charges to compute long range interactions, 15, Ghasemi et al. 34 were the first to use a NN to predict parameters entering a set of equations which, in turn, determine a set of atomic charges obeying an overall charge constraint. Before describing the details of the models studied here we would like to highlight four recent highly relevant works. The first is the work of Zubatyuk et al 35 which extends AIMNet to charged systems. These authors have resolved the dilemma of adding a net charge in a novel and interesting way: a network which predicts energies of multiple charge states simultaneously. By definition, this scheme requires either the ionic state or the neutral state to be a radical. This is a disadvantage for our desired use case (closed-shell systems), but it could be useful in other contexts. Since this algorithm requires at least three energy labels for each training point (charge states +1, 0, -1), it seems that extending the scheme to any other charge states (+2, -2, etc) would require large increases in the size of the training set. In the same work Zubatyuk have also described a second method which simultaneously predicts atomic spin populations and energies and can handle arbitrary charge and spin states. These models were primarily trained on radical ions and used to predict ionization potentials and electron affinities whereas we focus here on conformational energies of closed shell ions. A second work from Ko et al 15,22 reports a "fourth generation" HDNNP (4G-HDNNP) which is quite similar to one of the two models we present below. While our model was independently developed, it was motivated by earlier work by some of the same authors, 34 and so it is not unexpected that both would develop in the same direction. One can therefore interpret our work as an extension of Ko et al in which we demonstrate the ability to construct a transferable charge-aware HDNNP with broad coverage of organic molecules, and with improved computational scaling. Third is the work of Xie et al. which reports BpopNN 36 in which charges (or populations) are varied in order to minimize the full energy expression, as opposed to a simple model expression as we do here. While this is very natural and gives a variational energy expression, it also significantly increases the expense and complexity of training and evaluating the model because the energy expression depends non-linearly on the atomic populations. As an alternative to minimizing this energy expression during training, Xie et al. utilize data generated for non equilibrium atomic populations from constrained density functional theory (CDFT). 37 Fourth is the work of Qiao et al. who reports OrbNet, 5,6 a method which makes extensive use of features from the tight binding quantum mechanics method GFN1-xTB. 25 While the authors of that work have not explicitly demonstrated that OrbNet accurately reproduces energies of ions, our work suggests that their model likely has the capacity to work for such systems. In fact, one could interpret the charge features we use as coarse-grained approximations to the quantum mechanical charge density which is used as a feature in OrbNet. Further, OrbNet relies on delta learning and is trained to the difference between GFN1-xTB and their reference level of theory. We demonstrate here that one can use this technique to boost the accuracy of our HDNNP models as well, at the cost of having to perform a tight-binding calculation. The geometric features for the neural networks are modified Behler-Parrinello symmetry functions, 19 as described by Smith et al. 8 The Cartesian coordinates are transformed into a set of element resolved radial, and angular symmetry functions, ( Here R ij is the distance between atoms i and j, f C (R) is a switching function 8 which decays to zero at a radial cutoff, Z is the atomic number defining an element type and θ ijk is the angle formed by atoms i, j and k, centered on atom i. The primed summation indicates that repeated indices are to be excluded. R s , θ s , η and ζ are hyper-parameters which direct these symmetry functions to probe different regions in distance and angle space. Together, the radial and angular symmetry functions for atom i form a fixed length vector (G AEV A commonly used algorithm to calculate atomic charges (q i ) is the QEq method. 29 This algorithm defines the charges as minimizing a simple energy expression, Here λ is a Lagrange multiplier, Q tot is the total charge of the system, χ is the electronegativ-ity of atom i and J ij is the Coulomb interaction matrix. Following others 34,38 we parameterize the Coulomb interaction by assuming atom centered, spherical Gaussian charge distributions with a standard deviation σ i , yielding To determine the atomic charges from Eq. 4 one can set the derivative with respect to q i and λ to zero and solve the resulting linear equations. 15,29,34,38 The atomic electronegativity, χ i , and width parameters, σ i , are both allowed to be environmentally dependent and are predicted from an ANN. More precisely, we compute and then calculate the Qeq parameters as χ i = X 2 i and σ i = σ 0 + S 2 i . This is done to ensure χ i > 0 (for physicality) and σ i > σ 0 (for numerical stability). We have found σ 0 as small as 0.05 Angstrom to be sufficient to avoid numerical issues. Once the charges have been determined they can be used as features to a second ANN which defines the energy. For atomic features, in addition to atomic charge and the standard AEV, we also use a charge-weighted radial AEV which describes the local charge environment: Here we have used the superscript qR to distinguish the charge-weighted radial AEV from the geometric radial and angular AEV described above. Finally, we compute the total energy as where E disp is the empirical dispersion correction of the ωB97X-D functional 39,40 and E coul is a truncated Coulomb energy which decays smoothly at short range. 41 Eq. 8 along with Eq. 4 can be seen to be a fairly straightforward extension of the recently reported 4G-HDNNP 15 with the main difference being the charge-weighted radial AEV shown in Eq. 7 which we have found increases the capacity of the model by providing information about the local distribution of charge around an atom. The truncated Coulomb expression we use is given by At large distances this expression becomes the standard coulomb interaction whereas at short distances the coulomb interaction is attenuated, in this work we use the parameters 2.2 and 8.5 −1 for a and b, respectively. In order to solve the Qeq equations and determine the charges one must first compute the Coulomb matrix in Eq. 5 and then solve a set of linear equations that has the dimension of the number of atoms. This yields a method with the same asymptotic scaling as the linear solver, approximately O(N 3 atoms ), a distinct disadvantage compared to standard force fields which can be computed in quasi-linear time O(N atoms log(N atoms )). The model we focus on in this report is an approximate form of the Qeq method with reduced computational scaling. This model removes the need to solve a system of linear equations, by shifting the burden onto the neural networks to predict a more difficult parameter χi . We begin by separating the diagonal and off diagonal contributions to the Coulomb sum in Eq. 4, thus defining an "effective" electronegativity: Without the explicit off-diagonal terms in J ij , this expression now has a simple analytic solution in terms of the effective electronegativities χi , We interpret the effective electronegativities, χi , as environmentally-dependent learnable atomic properties. From Eq. 10 it is seen that these parameters have an implicit dependence on all other atomic charges. To approximate this effect the charges are predicted iteratively, with each iteration using only local information for each atom. On iteration I we predict effective electronegativities which then enter Eq. 12 to predict the atomic charges. Note that it is not necessary to compute the full coulomb matrix for this method. Further, we fix the diagonal hardness parameters and use those taken from Caldeweyher et al. 38 The charge-weighted AEV, G qR i , of each atom is computed at each iteration, allowing charge information to propagate locally in a way reminiscent of message passing NNs, 9,10,35 though with a pre-defined type of message (atomic charges). We generally utilize only two cycles in the recursive procedure and have found that the statistical errors in predicted energies do not improve with additional recursive cycles. The charges are then used in an energy model identical in form to Eq. 8. The resulting model, which calculates accurate system properties but preserves the quasi-linear scaling of standard force fields, we call QRNN for charge recursive neural network. ## Dataset Construction We have constructed a large dataset of ions of druglike molecules and their tautomers using active learning which consists of approximately 18 million examples. We initialize the process with a dataset of neutral molecules which has been previously described. 23 A QeqNN model trained on the previous data at each active learning step is used to perform geometry optimization and normal model sampling (NMS). 8 As such, we must begin by appending to the neutral dataset an initial set of ionic data that roughly represents ionic species in general. Active learning cycles are initiated with a dataset of very small ionic fragments optimized by DFT. This is done by fragmenting molecules appearing in ChEMBL 46 and ZINC 47 molecular datasets with the BRICS 48 implementation in the RDKit. 49 The fragmentation points are methyl-capped, unless they are carbon atoms already in which case they are hydrogencapped. We retain all fragments with five heavy atoms or fewer, and then generate tautomers of charge states with a relative charge difference of -1, 0, and +1 electrons using Jaguar and EPIK tautomer enumerators. 50,51 All unique fragments (according to canonicalized SMILES strings) are retained and for each fragment we generate a single starting conformer with our in-house methodology, Fast3D. Each resulting conformer is then optimized with our reference level of DFT. For any optimization not resulting in a chemical reaction (see below), In the initial stages tautomers were not filtered and we retained all found tautomers. As the complexity of molecules grows the number of tautomers also grows combinatorially; to avoid an explosion of the number of high energy tautomers we filter them using an acceptance probability distribution given by where t is the tautomer energy relative to the lowest known tautomer and the mean and standard deviation are 10.0 and 20.0 kcal/mol respectively. In the final rounds of active learning we narrow our focus further by only using ionic examples directly from the molecular datasets, we expect that these examples are low lying tautomers as judged by some expert or software program that constructed the dataset or SMILES string. The rounds of active learning are summarized in Table 1. After each round DFT labels are generated and an ensemble of five members is trained. When generating tautomers by our enumeration protocol, potentially with very high energies, it is not uncommon that a geometry optimization will produce a chemical reaction. This reactivity may be enhanced because we do not use a solvent model, which could stabilize ## Details We train models directly to DFT energies (see below) as well as to the difference between GFN2-xTB and DFT energies. We refer to these as direct and delta learning. For direct learning we train models of the form of Eq. 8 to the atomization energy, that is, we subtract per-element atomic energy offsets from the DFT energies and train to the (much smaller and more tractable) residuals. For delta learning, we first generate delta energy labels then fit per-element atomic energy offsets to this difference (as for DFT energies) and again train to the residual of the labels and the per-element atomic energy offsets. When performing delta learning we omit the long-range terms E disp and E coul (q) from Eq. 8 since we expect GFN2-xTB to reproduce the reference DFT reasonably well at long range. All models are trained using a multitask loss function, 41,52 where the two tasks are charge prediction and energy prediction, L E and L q are the loss functions for the energy task and charge task, respectively. The inverse weights, σ E and σ q are trainable parameters. This approach obviates the need to hand tune the weights of the tasks in the overall loss function and we have found that models trained with this method outperform models trained sequentially. The energy loss function is taken to be the squared error between the predicted energies and energy labels. For delta learning we use the squared error between the predicted charges and GFN2-xTB charges. For direct learning we train to the squared error of predicted dipole moments to those predicted by DFT. The dipole moment is a physical observable which avoids the wellknown arbitrary nature of atomic partial charge schemes. More importantly, training to dipole moments ensures correct long range electrostatic interactions, which is not true for charge decomposition schemes in general. This is less important for the delta learned models where we rely on GFN2-xTB to provide a good description of the long range interactions. When applying a delta learning approach, one could use the atomic charges directly from the low level QM method (GFN2-xTB). However, this complicates the evaluation of energy gradients because one is required, by the chain rule, to compute the derivative of the quantum mechanically computed atomic charges with respect to Cartesian coordinates. While it is still possible to formulate an analytic gradient in such a case (see the energy gradient formulas reported for OrbNet as an example 53 ) it increases the complexity of the All DFT calculations were computed with the Jaguar molecular electronic structure package 56 and utilize the pseudo-spectral approximation to accelerate computation of J and K matrices. The reference functional is ωB97X-D 39 and we utilize the 6-31G* basis set. All calculations are run with default accuracy settings. All minima optimized with DFT are verified as such by checking that the number of imaginary vibrational frequencies is zero. Optimizations using the HDNNP models and GFN2-xTB were performed inside Jaguar with a local modification that allows specification of an external program which returns energy and gradient data. All PM7 calculations are performed with MOPAC2016 57 and all GFN2-xTB energy evaluations are computed with the xtb python API. 58 ## Results and Discussion We 31G* level which we will refer to as direct learning as well as the difference between DFT and GFN2-xTB, which we will refer to as delta learning and denote by appending -TB to the name of the model, as in QRNN-TB. We evaluate the performance of these four models on test datasets which probe the accuracy of predicted atomic charges, molecular dipole moments, torsional energy profiles, relative conformational energies, optimized geometries and relative tautomer energies. We focus on testing the models against our chosen reference level of theory, not against the highest possible level of theory. We expect that our chosen reference will have (potentially significant) basis set incompleteness errors for some systems, and this will be addressed in future work. Nonetheless we expect that if our model accurately reproduces the chosen reference it will still be generally useful for many applications and that we can have success in recapitulating higher levels of theory after some subset of the data is recomputed at that higher level. 59 All test sets that were generated in this work are reported in the supporting information with DFT, GFN2-xTB and model energies labeled to facilitate reproduction of this work and comparison to other works. We hope the addition of test sets of ionic molecules contributes to the growing body of test sets available for machine-learned force fields. We start by evaluating the effectiveness of delta learning on a standard dataset, QM9. 60 We were inspired by recently reported results of OrbNet, 5,6 a model which uses quantum mechanically derived features. It was reported that OrbNet can achieve higher accuracy results with fewer training data points as compared to other methods. Besides its unique featurization, OrbNet also differs from other methods in its use of delta learning between GFN1-xTB and DFT, rather than learning DFT directly. We were interested in how much of the reported impressive performance could be replicated in a much more simple HDNNP, simply by using delta learning. To this end we trained an HDNNP to the ground state energy labels of QM9 using train/test splits reported by Qiao et al. 5 Details of the training procedure for this section are available in the supporting information. QM9 is a popular dataset used to test the "capacity" of a model, or its ability to accurately train to complex data. In this section we also train a message passing model, SchNet, 61 as its design more closely resembles OrbNet and can be thought of as a more modern design. geometries, whereas our ionic training set and our test sets contain geometries away from potential energy surface minima. For this reason, as well as a potential lack of chemical diversity 62 in QM9 we expect our total energy errors to be higher on other datasets. ## Charge prediction accuracy Prior to evaluating the performance of our charge-aware energy models on various test sets it is interesting to inspect the ability of the models to reproduce reference dipole moments and atomic charges. For this test we use the geometries from the conformational energy benchmark of Folsmbee and Hutchison 27,63 which consists of both neutral and ionic druglike molecules containing up to 50 heavy atoms, significantly exceeding the size of the largest molecules in our training set (24 heavy atoms). This test set is further described in Sec. 5.4 where performance on conformational energies is analyzed. As described in Sec. 4 our direct learned models are trained to reproduce ωB97X-D/6-31G* dipole moments whereas our delta learned models are trained to reproduce GFN2-xTB atomic charges. As such, we report the errors in these dipoles and charges for the respective methods in Table 3. In both charge and dipole reproduction the QRNN method outperforms the Qeq method. The QRNN-TB method is able to reproduce atomic charges with an RMSE of only ∼0.005 electrons for neutral systems and less than ∼0.01 electrons in ionic systems. QRNN-TB is able to reproduce dipole moment components with an RMSE of ∼0.4 Debye for neutral systems and ∼ 0.8 Debye in ionic systems. It is hard to judge the accuracy of the dipole moments by inspecting the RMSE of the dipole components alone, since the dipole moment can grow with the size of the system. As such we also report mean absolute percent errors (MAPE) of dipole moment magnitudes in Table 3. This metric shows that the error in the magnitude of dipole moments are, on average, less than 10% for QRNN and around 15% for QeqNN. This point is also supported by Figure S1 of the supporting information, which shows that the correlation of the dipole moments and charges for Qeq type methods is excellent, indicating a high degree of similarity between the model and reference values. The correlation for QRNN type models is not shown but as indicated by the errors in Table 3 it is slightly better than the Qeq type models. We attribute the increased error for charged systems for both dipole moments and atomic charges to the increased magnitude of both partial atomic charges and dipole moments in such systems. QRNN was developed in Section 2 as an approximate form of the Qeq method, and thus it is interesting to ask how it can be more accurate than Qeq over these test sets. Similar to Qeq, QRNN utilizes a NN to define effective atomic electronegativities which are then used to determine atomic charges, in Eq. 12. While the effective electronegativities are functions of the local environment, the atomic charges are globally defined, as in the Qeq method, with every charge being a function of every electronegativity (this can be seen through the dependence of the Lagrange multiplier on the electronegativities of all atoms in the molecule). Thus, the atomic charge on an atom can be modified by the presence of a second atom that is far outside of the cutoff radius, this is demonstrated in Figure 1 where it is seen that the atomic charge on the carbonyl oxygen is modulated by the presence of a distant functional group. The examples in this figure (motivated by similar results reported by Ko et al 15,22 ) show that all the methods capture the correct trends in atomic charges and place equivalent charge on atoms with equivalent atomic environments, as seen in panel (c). While Figure 1 demonstrates charge transfer over a distance slightly over twice the radial cutoff, we have extended this analysis to much greater distances in Figure S2; it is shown that both QeqNN and QRNN are able to describe charge transfer at a range of 200 . The ability of a model to accommodate changes in atomic charge due to either modification at a distant site (changes in functional group or a chemical reaction) or a global charge state are the hallmarks of the fourth generation (4G) HDNNPs. 16 While QRNN neglects explicit long-range Coulombic interactions when determining atomic charges, the global dependence of these charges on the effective electronegativities and net charge state mean that the model can still describe long-range charge transfer. Therefore, QRNN is an example of a 4G HDNNP, as defined in Ref. 16. However, since QRNN explicitly neglects long-range Coulomb interactions during the determination of atomic charges, we expect that it cannot fully describe long-range polarization effects. This could result in a degradation of accuracy for highly extended systems. In this work, we focus on small, organic, druglike molecules and their ions, and leave a detailed analysis of the long-range polarization behavior of QRNN for future work. The ability of QRNN to transfer charge over large distances is both a blessing and a curse, and means that QRNN inherits well-documented problems associated with Qeq. Notably, unphysical long-range charge separation; the component parts of a dimer or other molecular system may be predicted to have non-integer charges when separated to some very large distance. 64,65 One interesting difference between QRNN and Qeq is that the QRNN method, due to its recursive nature, is itself charge-aware. The predicted electronegativity of each atom is dependent on its fractional charge, as determined in the previous cycle. We believe this allows for additional accuracy in the predicted electronegativities, particularly for elements with significant variation in atomic partial charge. Finally, we would like to point out that it is possible, though we leave it for future work, to use recursive determination of electronegativities in the Qeq method as well. This might allow for further accuracy enhancements in the resulting hybrid "QeqRNN" method, though at a higher computational cost than either Qeq or QRNN alone. ## Torsion energies We next turn to testing energy predictions of models trained on the full training set described in Sec. 3. Here our intent is to validate the models for describing geometries and energetics of conformers of organic molecules and their ions. We are primarily concerned with transferability and therefore test on molecules outside of the training set. One of our target applications is the use of HDNNP energies as reference data for fitting torsion parameters of traditional force fields. 18 As such, we test the accuracy of torsion scans of druglike molecules. We have generated relaxed torsion scans (optimized at the reference level of DFT) by fragmenting 1000 ionic or zwitterionic druglike molecules randomly selected from an internal dataset of such species (CACDB). This fragmentation results in 388 unique torsion scans of species containing at least one non-zero formal charge, which we will refer to as "charged" species, and 112 unique "neutral" molecules which contain no formal charges. Using each method, we compute single-point energies for the DFT-optimized geometries in this test set. The errors are listed in Table 4 and shown graphically in Fig. S2 which can be found in the supporting information. The errors reported here are computed as root mean squared error (RMSE) in relative energy (relative to the DFT minimum geometry) over each torsion scan, which consists of 12 equally spaced points. From these data it is clear that QRNN and QeqNN behave quite similarly and that delta learning improves the errors by nearly a factor of two. The direct learning models perform much worse on charged systems (∼1.0 kcal/mol) than does a model trained only to neutrals and evaluated on neutrals (0.56 kcal/mol). This situation recovers when applying delta learning and we are able to achieve a mean error of only ∼0.5 kcal/mol. All of the charge-aware models dramatically outperform the tested semi-empirical theories, which for charged systems perform similarly to our previous HDNNP trained solely to neutral systems. 23 These results are displayed as a box and whisker plot in the supporting information which shows that remarkably, the largest outliers with the delta learned models are only slightly higher than the upper quartile of errors for the semi-empirical theories. As a general rule we find that QRNN and QeqNN models perform very similarly, with the QRNN version slightly outperforming regardless of whether or not we use direct or delta learning. For this reason and the fact that QRNN has superior formal computational scaling we will focus on this model for the remainder of this article. Tables which include results for the QeqNN models can be found in the supporting information. ## Relative conformational energies We now turn our attention from torsional profiles to relative conformational energies of flexible molecules. Torsion scans probe energies which include torsion barrier heights, whereas relative conformational energies probe minima on the potential energy surface. To evaluate the conformational performance of our models we will use a test set reported by Folmsbee and Hutchison. 27 This test set is constructed from, as decomposed by Folmsbee and Hutchison, 63 622 neutral molecules and 86 charged systems, all of which are expected to show some degree of conformational freedom. Hutchison and co-workers performed a conformational search on each of these molecules and reported up to ten low-lying conformers of each species. 27 The test molecules contain up to 50 heavy atoms and 23 rotatable bonds, representing a strong test of transferability for our models, since our training molecules are substantially smaller. We have re-computed DFT energy labels at our reference level of theory and filtered any geometries that had unconverged self-consistent-field equations or contained chemical elements not supported by our model. This filtration left us with 576 neutral and 81 charged systems to study. We follow the analysis performed in the original work on this dataset and compute the Mean Absolute Relative Error (MARE) and the square of the Pearson correlation coefficient (R 2 ) for each set of conformers. The median of these two metrics over all conformer sets are used to assess the quality of reproducing the reference energies and rank ordering of conformations. Hutchison showed that when comparing against a DLPNO-CCSD(T)/cc-pVTZ reference, DFT methods typically have median R 2 values of greater than 0.8 and MARE of less than 0.3 kcal/mol whereas the best empirical methods (GFN2-xTB, ANI) have R 2 less than 0.65 and MARE greater than 0.4 kcal/mol. We would consider results for our models (versus our reference level of theory) that are similar to those reported for DFT versus DLPNO-CCSD(T) to be a very encouraging sign. Tables 5 and 6 show our results for the neutral and charged subsets, respectively. Indeed, comparing the delta learned models to our reference level of theory, we see that we can achieve MARE less than 0.2 kcal/mol and R 2 values greater than 0.9 for both subsets. Again, we see that delta learning provides an impressive gain in accuracy over direct learning and that both methods show improved error statistics relative to the semi-empirical QM methods PM7 and GFN2-xTB alone. When comparing to the DLPNO-CCSD(T) references the delta learned models perform only slightly worse than when comparing to our reference level of theory, ωB97X-D/6-31G*. We are also encouraged by the observation that the difference between ωB97X-D/6-31G* and DLPNO-CCSD(T)/cc-pVTZ is similar to that between our delta learned models and ωB97X-D/6-31G*. Further, when comparing directly to DLPNO-CCSD(T), the delta learned models perform nearly as well as ωB97X-D/6-31G*. ## Accuracy of optimized geometries In order to assess the accuracy of optimized geometries we have constructed a dataset of ionic conformers of druglike molecules. Two hundred molecules were drawn at random from the ZINC dataset. 47 The first hundred of these molecules were required to have sixteen heavy atoms or fewer, and the second hundred examples were required to have thirty-two heavy atoms or fewer. For each molecule we perform a mixed-mode conformational search with MacroModel utilizing the OPLS3e force field. 18 For each molecule, a maximum of 200 conformations is returned, with a maximum energy range of 12.0 kcal/mol. In an effort to increase the diversity of our test set, we choose not to select only lowenergy conformers for each molecule, since these are often similar to one another. Instead, we take the minimum-energy conformer and then up to nine other conformers drawn uniformly at random from the remainder. Each of the conformers is then geometry-optimized at the ωB97X-D/6-31G* level and then re-optimized, starting with the DFT geometry, with each model tested. After removing saddle point geometries we were left with 190 conformer sets that contained more than one conformer. We do not filter duplicate conformers after the DFT optimization and thus some similar minima will remain. However, conformational diversity of our conformer set is much greater than that of the Hutchison test set, with a mean relative energy range per molecule of 10.9 kcal/mol for our test versus only 2.9 kcal/mol for the Hutchison set. The results of the geometry optimizations are shown in Table 7. We see that errors in bond lengths, angles, torsion angles as well as overall Cartesian RMSDs improve from semi-empirical to QRNN and improve further for the delta learned model, QRNN-TB. The difference between direct and delta learned models for geometries is small. Our best models achieve errors of 0.004 , 0.52 • , 6.62 • and 0.24 for bond distances, angles, torsion angles and Cartesian RMSDs respectively. For comparison, the geometric errors for bonds, angles and torsion angles not involving atoms bearing a formal charges are given in Table S4 of the supporting information. While the neutral errors are uniformly lower, the errors follow the same trends as the ionic group focused errors with errors growing from semi-empirical theories to QRNN to delta learned QRNN-TB. The difference in performance on neutral and ionic functional groups also follows the same trend, with QRNN-TB performance being nearly indistinguishable and semi-empirical methods showing a larger difference. We also compute relative energy errors and R 2 values for this dataset and again find very good correlation with the reference. We believe the linear correlations and energy errors are higher than for the Hutchison dataset due to the larger range of energies in our test set, as discussed above. Overall the geometries produced by the delta learned models are in excellent agreement with DFT references and warrant future work related to energy ranking of conformers with these methods. Finally, we evaluate our models' ability to reproduce relative tautomer energies. Our primary interest in this task is in the impact it may have upon workflows which compute the pKa of organic molecules, as in drug design. In order to compute pKa it is often necessary to rank-order a large number of tautomers of a certain charge state. 51 It is very difficult to rank these tautomers using a purely rules-based scheme, and computationally expensive to do so with DFT or other ab initio methods. Generally, this type of bond-changing energy difference is outside of the range of applicability of classical force fields, so this is an example of an area where an HDNNP could have a large impact in computational life science. Currently in common workflows, tautomers are ranked with semi-empirical methods, but the low accuracy of these methods for this task (see below) means that a wide energy window must be used for selecting samples for re-ranking with DFT, increasing the DFT workload. ## Relative tautomer energies Tautomerization free energies have been recently studied with an ML/MM model based on ANI. 66 In addition Vazquez-Salazar et al. 67 have recently explored the impact of the diversity of training set on the ability to compute relative tautomer energies in a public dataset, Tautobase, 28 and we use the same dataset here. Tautobase consists of 1673 tautomer pairs stored as SMIRKS strings. Here we neglect solvent effects and focus on reproduction of relative tautomer energies in the gas phase. (Solvent effects would need to be accounted for in order to make contact with the experimentally observed populations, which are also available in Tautobase.) We convert each tautomer to a single (arbitrary) three-dimensional starting conformation and optimize with ωB97X-D/6-31G*. Relative energies are then computed using each of the tested models. After filtering unsupported elements, failed SCF or geometry optimization jobs, and optimizations that landed on saddle-points, we are left with 1552 The mean error over the tautomer pairs are listed in Table 8; again we see a dramatic improvement in going from GFN2-xTB to direct learned, charge-aware HDNNP and finally to a delta learned model. For the best model the overall mean error is only 0.57 kcal/mol and only a very few tautomers are mis-ranked as shown in Fig. 3, where qualitative misrankings appear as negative values on the vertical axis. These results suggest that it is possible to replace semi-empirical ranking of relative tautomer energies with a charge-aware HDNNP, and with this replacement utilize a significantly smaller energy window to re-rank tautomers with an ab initio method, greatly reducing costs. This motivates further work to incorporate solvation effects into such a model. ## Conclusions In this work we have reported on the construction of a transferable, charge-aware, HDNNP with broad applicability to organic molecules, including their conformational energies, ions, and tautomers. We have presented the results of two models, QRNN and QeqNN, when these models are trained directly to DFT based energy labels and to the difference between GFN2-xTB and DFT. While a model almost identical to QeqNN has recently been reported 15 this is, to our knowledge, the first report of a transferable HDNNP applied to a broad range of closed shell ionic systems and their tautomers. We also report a novel charge model, QRNN, which performs at least as well as Qeq methods like the previously reported 4th generation HDNNP 15 and has superior scaling properties. Despite the fact that the charge-aware models have higher computational complexity we find that the time to evaluate energy and forces on the size of molecules found in this work is not burdensome, as shown in Table 9. The more expensive, charge-aware methods are still at least 10 4 times faster than ωB97X-D/6-31G* for a 48 heavy atom system. Due to the improved scaling of QRNN we expect the performance of QRNN over QeqNN to improve with increasing system size for a performant implementation of the method. We leave further study of this to future work. We find that all of the ML based models studied here are able to achieve errors below 1 kcal/mol on relative energies of conformations on a broad range of ionic systems outside of the training set. Further, we find that delta-learning based on GFN2-xTB can nearly halve this error. We believe this technique could make a large impact in workflows involving conformational energy analysis and ranking, by providing geometries and energies that are highly accurate relative to the level of theory they are trained to. Finally, we show that our models are able to rank relative tautomers to less than 1 kcal/mol accuracy, a large advance for efficient energy ranking of tautomers. These results neglect solvent effects, which are important in both conformation and tautomer energy rankings, and this will be addressed in future work. In addition, we do not expect inter-molecular interactions to be well described by any model trained to the current data set due to the fact that the training examples are all single molecules. This will be extended in future work. A final limitation of this work is that our scheme cannot discriminate different spin states and we would expect this model to perform poorly on a training task involving multiple spin states, just as a model which is not charge-aware would perform poorly on a training set of ions. A potential solution would be to extend the type of atomic charge prediction used here to atomic spin populations, as was done with AIMNET 35 or, equivalently, to predict an atomic charge and an atomic spin polarization. This should allow the model to discriminate between systems of arbitrary spin.

chemsum

{"title": "Transferable Neural Network Potential Energy Surfaces for Closed-Shell Organic Molecules: Extension to Ions", "journal": "ChemRxiv"}

isonitrile-responsive_and_bioorthogonally_removable_tetrazine_protecting_groups

## Abstract: In vivo compatible reactions have a broad range of possible applications in chemical biology and the pharmaceutical sciences. Here we report tetrazines that can be removed by exposure to isonitriles under very mild conditions. Tetrazylmethyl derivatives are easily accessible protecting groups for amines and phenols. The isonitrile-induced removal is rapid and near-quantitative. Intriguingly, the deprotection is especially effective with (trimethylsilyl)methyl isocyanide, and serum albumin can catalyze the elimination under physiological conditions. NMR and computational studies revealed that an imine-tautomerization step is often rate limiting, and the unexpected cleavage of the Si-C bond accelerates this step in the case with (trimethylsilyl)methyl isocyanide. Tetrazylmethyl-removal is compatible with use on biomacromolecules, in cellular environments, and in living organisms as demonstrated by cytotoxicity experiments and fluorophore-release studies on proteins and in zebrafish embryos. By combining tetrazylmethyl derivatives with previously reported tetrazine-responsive 3-isocyanopropyl groups, it was possible to liberate two fluorophores in vertebrates from a single bioorthogonal reaction. This chemistry will open new opportunities towards applications involving multiplexed release schemes and is a valuable asset to the growing toolbox of bioorthogonal dissociative reactions. ## Introduction Performing chemistry in living organisms with bioorthogonal reactions makes it possible to study biological processes in their natural environments. 1,2 Recently, reactions have emerged that release diverse molecules under physiological conditions. 3,4 These reactions have opened unprecedented possibilities in chemical biology and drug delivery. 5,6 Dissociative bioorthogonal chemistry has been applied to the on-demand dissolution of polymers and micelles, site-specifc actuation of prodrugs, and control of enzyme activity in vivo. Although a growing number of "click-to-release" reactions has provided a solid foundation for applications in the life sciences, extending the reaction scope will be necessary to access the full range of capabilities. Moreover, there is a need for chemistry to allow for the controlled and simultaneous release of more than one molecule. 24 Dual-release reactions could be used for the concomitant delivery of synergistic drugs, in theranostic applications, and in multiplexed detection schemes. Bioorthogonal chemistry, both ligating and dissociative, mainly revolves around pericyclic reactions. 25,26 In particular, inverse-electron demand cycloadditions offer rapid reaction kinetics and high biocompatibility. 27, 28 1,2,4,5-Tetrazines are the most prevalent dienes in such reactions. 29,30 These heterocycles react with and subsequently trigger the release of payloads from allyl-modifed trans-cyclooctenes, 17, benzonorbornadiene derivatives, 21,34 and vinyl ethers. 7,19,20 Tetrazines also undergo bioorthogonal cycloaddition reactions with isonitriles, 35,36 and we have recently shown that they can induce the release of payloads from 3-isocyanopropyl (ICPr) groups (Fig. 1). 37,38 Given the prominent role and favorable properties of tetrazine-based cycloadditions in dissociative bioorthogonal chemistry, it would be valuable to have tetrazine-based protecting groups that release a payload upon reaction with some of these dienophiles. An example of such a molecule was disclosed by Wang et al. as demonstrated by a tetrazine-based prodrug that was activated through a reaction with a cyclooctyne modifed with a hydroxyl group at the propargylic position. 23 Such tetrazine derivatives, when combined with complementary release reagents, could be used for dual-release applications. Running two bioorthogonal release reactions in parallel is one possibility to achieve such dual-release as has been demonstrated by combining the reaction of tetrazines and benzonorbornadienes with that between sulfonyl sydnonimines and dibenzoazacyclooctyne. 39 A second example involved the reaction between vinyl ethers and tetrazines, which released alcohols but was limited to generating pyridazine and had slow reaction kinetics. 24 A single pair of reactants that releases two molecules in a single fast reaction would bring such approaches to the next level. Here we describe tetrazylmethyl (TzMe) protecting groups that can be rapidly removed by a reaction with isonitriles. The rationale behind our design is based on the precedent that isonitriles convert tetrazines into 4-aminopyrazoles 35,36 and that 5membered heterocycles with amine substituents spontaneously eliminate diverse functional groups. 40,41 In a series of experiments, we demonstrated that TzMe-modifed molecules reacted readily with isonitriles to release amines (from tetrazylmethyloxycarbonyl (Tzmoc) derivatives) and phenols (Fig. 2a). We analyzed the reaction mechanism, and in the case of (trimethylsilyl)methyl isocyanide (TMS-MeNC), observed an intriguing C-Si bond cleavage that accelerated release. The reaction was compatible with living systems, and we demonstrated that when TzMe-derivatives were combined with ICPr-derivatives (Fig. 1), 37 two fluorophores could be simultaneously released in zebrafsh embryos. This innovative chemistry will open new possibilities for biomedical research and drug delivery. ## Isonitrile-induced removal of Tzmoc-groups from amines To prove the concept of isonitrile-induced deprotection of TzMe groups, we synthesized a colorimetric reporter probe (4a, Fig. 2b and c) consisting of p-nitroaniline (pNA) caged by a Tzmoc group (Fig. 2a). pNA has a characteristic maximum absorbance signal at l abs ¼ 385 nm, which is hypsochromically shifted in derivatives with acyl-modifed amine. 4a was accessed by a dibutyltin dilaureate-catalyzed reaction of (6-(tert-butyl)-1,2,4,5-tetrazin-3-yl)methanol (2) with 4-nitrophenyl isocyanate (Fig. 2b). 2 was prepared in three steps from the nitrile precursors to obtain 1, followed by deprotection of the methoxy group by BBr 3 (Fig. 2b). We evaluated the liberation of pNA from 4a upon reaction with several isonitriles (Fig. 2d). As designed, a primary isonitrile, n-butyl isocyanide (n-BuNC, Fig. 2c), reacted with the tetrazine and elicited the release of pNA as monitored by the emergence of the pNA absorbance signal (Fig. 2d). As a control, we performed the experiment with tert-octyl isocyanide (t-OcNC, Fig. 2c), which we expected not to release pNA because tertiary isonitriles form stable 4H-pyrazol-4-imine conjugates. 36,38,42 Indeed, pNA-release was undetectable in experiments with t-OcNC confrming that TzMe-removal follows the designed release principles. We were interested whether electrondonating groups adjacent to the isocyano functionality would accelerate the inverse-electron demand cycloaddition step. We therefore tested the reaction of 4a with (trimethylsilyl)methyl isocyanide (TMS-MeNC, Fig. 2c). As predicted, TMS-MeNC reacted $3-fold faster with 4a (k 2 ¼ 0.344 AE 0.013 M 1 s 1 ) than did n-BuNC (k 2 ¼ 0.117 AE 0.001 M 1 s 1 ), and an isonitrile with an electron-withdrawing substituent (methyl isocyanoacetate) lead to a 2-fold decrease (k 2 ¼ 0.05 AE 0.01 M 1 s 1 ) in the cycloaddition rate (Table S1 †) but still released pNA (data not shown). Unexpectedly however, the TMS-substituent also greatly accelerated the release step (Fig. 2d). The rate of pNA release was $30-fold faster for TMS-MeNC (k 1 ¼ 3.4 10 4 AE 1.1 10 6 s 1 ) than for n-BuNC (k 1 ¼ 1.1 10 5 AE 1.2 10 7 s 1 ). Reactions with n-BuNC led to gradual, continuous, elimination of pNA with a release yield of 35.4 AE 1.0% measured at the 8 hour time-point in contrast to reactions with TMS-MeNC leading to near-quantitative release yields in this period as quantifed by the absorbance signal (Fig. 3c). The bimolecular reaction rates of tetrazines and isonitriles were in the range of those observed in previous studies (Table S1 †). 37,38 Under these conditions, (DMSO : PBS pH 7.4, 4 : 1, v/v at T ¼ 37 C) the rate constants of the reactions with 4a ranged from k 2 ¼ 0.05-0.38 M 1 s 1 . The water content strongly influences the kinetics of the cycloaddition step, and based on previous studies, 34,38 we extrapolate the reaction to be about 10-fold faster in purely aqueous solutions. These initial results indicate that the TMSgroup promotes pNA release as the faster bimolecular rate of TMS-MeNC compared to n-BuNC is insufficient to explain the rapid elimination of pNA for the former. Although TMS-MeNC effectively elicits the release of amines from Tzmoc groups, there are applications for which the rapid release by simple alkyl isocyanides will be preferred. For example, TzMe-molecules could be combined with ICPr derivatives 37 in dual-release strategies (Fig. 1). Serum albumins catalyze diverse chemical transformations, and we hypothesized that albumin might also accelerate the release step. Indeed, both human serum albumin (2 mg mL 1 HSA in 4 : 1, PBS pH 7.4 : DMSO at T ¼ 37 C) and bovine serum albumin at 2 mg mL 1 (data not shown) greatly accelerated the liberation of pNA in reactions with n-BuNC (Fig. 2e). n-BuNC was able to effectively elicit the near-quantitative release of pNA in a little over an hour while the same reaction without HSA led to less than 20% release during the same timeframe. In contrast, 4a incubated alone in a solution of HSA (2 mg mL 1 in PBS, T ¼ 37 C) did not result in a detectable pNA release signal (data not shown), indicating that HSA catalyzes the elimination step whereas the Tzmoc-probe is stable in the absence of isonitrile. To differentiate between a catalytic activity of the protein and simple base-catalysis by its surface amines, we tested the effect of tris-base (concentration equal to that of surface amines in HSA experiments; 2 mM) on the release rate of pNA. The base had no detectable effect on the isonitrile-induced release of pNA from 4a (Fig. S1 †). It is therefore possible to achieve rapid and high-yielding uncaging of amines from stable Tzmoc precursors with simple alkyl isocyanides in serum. ## Isonitrile-induced removal of TzMe-groups from phenols Having demonstrated the release of carbamates from Tzmocgroups, we aimed to determine whether the chemistry would be applicable to other functional groups. We were especially interested in phenols because aromatic hydroxy groups are present in tyrosine, diverse drugs, and fluorophores. For these experiments, we synthesized a TzMe-caged O-carboxymethyl fluorescein (4b, Fig. 2c) and 7-hydroxycoumarin (4b 0 , Fig. S2 †) that report on TzMe-removal by a fluorescence turn-on signal. 46,47 The fluorogenic probes were synthesized by ether-ifcation of the phenolic dyes with 3-(bromomethyl)-6-(tertbutyl)-1,2,4,5-tetrazine (3), which can be accessed by bromination (PBr 3 ; 88%) of 2 (Fig. 2b). The reaction of TMS-MeNC with 4b (Fig. 2f) and the 7-hydroxycoumarin derivative 4b 0 (Fig. S3 †) led to near-quantitative release yields by 2 h as quantifed by fluorescence emission and HPLC analysis (Fig. S4 †). TzMeremoval was associated with a characteristic fluorescence increase (152-fold for 4b 0 and 30-fold for 4b; Fig. S5 †). The kinetics of the release reaction of 4b with TMS-MeNC was determined by measuring the fluorescence turn-on signal (k 1 ¼ 3.5 10 3 AE 1.7 10 4 s 1 ). The faster release rate compared to the release of 4a may in part be explained by the higher water content. Surprisingly, the rate of 7-hydroxycoumarin (Fig. S3 †) and O-carboxymethyl fluorescein (Fig. 2f) elimination by n-BuNC was signifcantly faster than for 4a without the need for the addition of albumin, an effect that was assessed by NMR studies. We further evaluated the stability of TzMe-caged probes. First, we assessed a TzMe-caged 7-hydroxycoumarin dye (4b 0 , Fig. S2 †) in a human liver microsome stability assay (Creative Bioarray, USA) and the probe exhibited good stability (t 1/2 ¼ 52.11 min; Cl int ¼ 33.36 mL min 1 kg 1 ; Fig. S6 †) even under these harsh conditions. Second, the TzMe-derivative of fluorescein (4b) used for zebrafsh studies was stable in serum for hours (t 1/2 ¼ 19 AE 4 h). The decomposition product was not the released fluorophore and therefore the contribution to fluorescence background is low (Fig. S7 †). These experiments establish that TzMe-groups are removed rapidly and in high yields from key functional groups. ## Effect of structural modications on isonitrile-induced TzMe deprotection We were interested to determine if the reaction kinetics and release yields from isonitrile-induced release from TzMe-groups could be enhanced by modifying the structure of the tetrazine. We designed a series of tetrazyl-derivatives of pNA (Fig. 3a) and analyzed such parameters upon reaction with isonitriles n-BuNC and TMS-MeNC. Methyl and phenyl groups at the methylene position (R 0 in Fig. 3a) modestly accelerated the release of pNA upon reaction with TMS-MeNC (Fig. 3b) with minor impact on the bimolecular kinetics (Table S1 †). A methyl substituted tetrazine (4c, Fig. 3a) released pNA with a half-life of 24 min and a phenyl-substituted derivative (4d, Fig. 3a) with a half-life of 19 min, both near-quantitatively (Fig. 3b). Intriguingly, modifcations drastically decreased the ability of n-BuNC to trigger the release of pNA; after 8 h only 20.2 AE 0.8% and 6.9 AE 0.4% of the pNA was deprotected from 4c and 4d, respectively (Fig. S8 †). The effect on the bimolecular reaction rates did not cause the modest release yields of pNA (Table S1 †). Next, replacing the C-6 tert-butyl group of 4a by a phenyl substituent (4e, Fig. 3a) led to a marginally faster release of pNA triggered by n-BuNC (Fig. S9 †). However, the phenyl group decreased the rate of TMS-MeNC triggered release (Fig. 3b). This effect may in part be because of a slightly slowed bimolecular reaction rate (Table S1 †), which agrees with the lack of dispersion forces between the tert-butyl group and the incoming isocyano group. 38 These results show that the substituents on both the tetrazyl ring and the methylene position are important to achieve prompt and high-yielding release. These insights provide guidance for further improvement of probe performance in future studies. ## TzMe removal by alternative dienophiles Inspired by the effective release of payloads from TzMederivatives by isonitriles, it was of interest to fnd out whether other dienophiles provide a similar outcome. Tetrazines react with diverse strained alkenes, 29,30 and we tested whether such dienophiles (methylcyclopropene (Cp), norbornene (Nb), and trans-cyclooctene (TCO); see Fig. S10 † for structures) induce the release of pNA from 4a. Elimination of pNA occurred; however, the yields of amine release were modest (t ¼ 8 h, 37 C; Cp ¼ 32.4 AE 0.2%; Nb ¼ 31.0 AE 0.3%; TCO ¼ 20.2 AE 4.0%; Fig. 3c). Addition of HSA, which catalyzed pNA release from 4a (Fig. 2e) in the reaction with n-BuNC, had an insignifcant effect on the TCO-mediated reaction (data not shown). We further tested the ability of Cp, Nb, and TCO to elicit the release of O-carboxymethyl fluorescein from 4b. Analogous to the results for pNA release, only a fraction of the product was eliminated (t ¼ 8 h, 37 C; Cp ¼ 22.1 AE 0.3%; Nb ¼ 12.4 AE 0.5%; TCO ¼ 8.3 AE 1.0%; Fig. 3c). These experiments demonstrate that isonitriles have a unique ability to remove TzMe-based protecting groups. In the case of TCO, the rapid bimolecular reaction with tetrazines may open opportunities for interesting applications in drug delivery where the rate of the release step may not be limiting. ## Studies on the mechanism of TzMe removal Having established that isonitriles remove TzMe-moieties from phenols and amines (Fig. 2), we aimed to gain a mechanistic understanding of the reaction. Several reaction pathways are conceivable. Carbamate release could in principle occur by heterolytic cleavage of the benzylic C-O bond or by a cyclization step involving the attack of a nucleophilic intermediate on the carbonyl (Fig. S11 †). To elucidate the reaction pathway, we performed a time-dependent NMR experiment between 4a and n-BuNC (DMSO-d 6 : D 2 O (9 : 1, v/v) at T ¼ 25 C; Fig. 4a-c and S12, S13 †). At a lower aqueous content and T ¼ 25 C, as opposed to the higher water content and T ¼ 37 C we preformed kinetics studies with previously (Fig. 2d), we expected slower reaction kinetics to allow for rigorous examination of the intermediates formed along the reaction pathway. As determined by 1 H NMR, the formation of one equivalent of the 4H-pyrazole intermediate (I1) paralleled the disappearance of 4a in the reaction with n-BuNC (Fig. 4b and c). I1 subsequently tautomerized to the 1H-pyrazole intermediate (I2) that gradually released pNA (Fig. 4b and c). The triplet peak of I2 centered at 7.78 ppm with a normalized integration value corresponding to one proton is characteristic for the N]CH-CH 2 proton present in the postulated structure of I2. In reactions between n-BuNC and di-methyl-tetrazine (Fig. S14 †) or di-tert-butyl-tetrazine (Fig. S16 †), the same characteristic triplet peak at $7.8 ppm was present (Fig. S15 and S17 †), which indicated that the signal originated from the n-BuNC portion providing additional support for the structural assignment of I2. The observed reaction cascade mirrored the predicted mechanism (Fig. 1). 35,36 Interestingly, the 1 H NMR signals of pNA (d, 2H, 6.60 ppm; d, 2H, 7.94 ppm) emerged before those of the aldehyde (s, 1H, 9.64 ppm). It therefore appears that the elimination step can occur from the imine intermediate I2. We proceeded to study the reaction between TMS-MeNC and 4a by NMR (DMSO-d 6 : D 2 O (9 : 1, v/v), T ¼ 25 C; Fig. 4d-f and S18, S19 †). TMS-MeNC was completely stable for >7 days under the experimental conditions (data not shown) ruling out the possibility that a decomposition product caused the fast release. Time-dependent 1 H NMR spectra revealed a single intermediate (I1 0 ; Fig. 4e). I1 0 exhibited a strong coupling peak pattern centered at 7.45 ppm with a normalized integration value corresponding to two protons that was absent from I1 and I2 (Fig. 4b). Repeating the experiment in DMSO-d 6 , without the addition of the 10% D 2 O, led to a peak at 5.28 ppm corresponding to one proton, which could not be assigned to the pyrazole species (Fig. S20 and S21 †). To further examine the transformation, we performed the reaction between TMS-MeNC and di-tert-butyl-tetrazine in DMSO-d 6 at T ¼ 25 C (Fig. S22-S27 †). This reaction provided an adduct with the same strong coupling pattern with a normalized integration value corresponding to two protons and this species persisted for days in DMSO-d 6 , making it possible to thoroughly analyze its structure by various NMR experiments. The strongly coupled protons that centered at 7.60 ppm in the 1 H spectrum (Fig. S23 †) correlated in the gCOSY spectrum (Fig. S24 †) and according to gHSQC analysis, were bonded to the same carbon with a chemical shift of 159.7 ppm (Fig. S25 and S26 †). Furthermore, these protons showed a multi-bond correlation with one of the aromatic ring carbons in gHMBC (130.9 ppm; Fig. S27 †). The spectroscopic data is consistent with the formation of a methanimine intermediate, which would indicate cleavage of the C-Si bond (Fig. 4d). In agreement, trimethylsilanol was detected in the 1 H NMR spectrum (s, 1H, 5.28 ppm; s, 9H, 0.01 ppm; Fig. S23 †). The peak corresponding to the trimethylsilyl protons in the reaction of 4a with TMS-MeNC remained unaffected as the reaction proceeded to generate several unidentifed side products, further corroborating the formation of trimethylsilanol (Fig. S19 †). Cleavage of the C-Si bond under these conditions is surprising as documented cases required harsher conditions. 48 We analyzed this reaction step using density functional theory (DFT) calculations. The analysis was conducted in Gaussian 09 using M06-2X-D3/def2TZV in water (SMD). 52 3,6-Di-methyl-1,2,4,5-tetrazine derived intermediates A1 and B1 were used as model substances (Fig. 5) with water as the initial nucleophile or proton source in all pathways to reflect the neutral experimental conditions. The S N 2 reaction between A1 and water was identifed as the minimum energy pathway for the formation of imine A3 going through a highly stabilized anion A2 making this structure an excellent leaving group, allowing for low barriers even with weak nucleophiles such as water (Fig. 5a and e). The barrier was calculated to be 16.4 kcal mol 1 and this pathway is therefore in accordance with the fast reaction observed experimentally (Fig. 2d). In contrast, the deprotonation of B1 by water to initiate the tautomerization had a calculated barrier of 29.8 kcal mol 1 with the resulting intermediate B2 being 15 kcal mol 1 higher in energy than the reactants (Fig. 5b). While the omission of tunneling effects may overestimate barriers for proton transfers calculated with a classical treatment of the nucleus, it is plausible to assume that the barrier is above the 16.4 kcal mol 1 calculated for the A1 > A2 transformation, given that the intermediate B2 is already at +15.2 kcal mol 1 . This computational prediction agrees with the experimental observation that the tautomerization in case of reactions with n-BuNC proceed signifcantly slower than the cleavage of TMS (Fig. 4). Analogous pathways involving OH instead of water showed the same trend with overall lower barriers (Fig. S30 †). Alternative pathways that involve protonation of A1 and B1 followed by transfer of TMS + to water, or deprotonation, respectively, were also explored (Fig. 5c and d). Protonation of A1 or B1 was disfavored by 23.9 and 27.3 kcal mol 1 , respectively. The barriers for the following abstraction of TMS + or H + are lowered considerably compared to the pathways described above. However, this pathway also favors removal of TMS + from protonated A1 over proton abstraction from protonated B1 in accordance to experimental results. In addition, stability of TMS-MeNC against nucleophilic attack of water was investigated computationally. While the transition state structure could not be located, the transformation is disfavored by over 51.0 kcal mol 1 because of the inability of the adjacent isonitrile group to stabilize a carbanion at the a carbon, leading to poor leaving group qualities. The geometry of the anion shows a tetrahedral center at the a methyl group, consistent with isolation of the negative charge on this center without any stabilization by the adjacent p-systems (Fig. 5f). The high barrier and thermodynamically disfavored nature of this transition corroborate the observed high stability of TMS-MeNC in aqueous solution. We further examined the mechanistic steps of phenol release triggered by n-BuNC. The photospectrometric studies had revealed a puzzling discrepancy in the rate of carbamate versus phenol elimination induced by n-BuNC (Fig. 2). To obtain mechanistic insight into this discrepancy, we analyzed the reaction of 4b 0 and n-BuNC (DMSO-d 6 : D 2 O (9 : 1), T ¼ 25 C) by 1 H NMR (Fig. S28 and S29 †). In this experiment, the formation of the corresponding 4H-pyrazole species, which was noticeable in the reaction between n-BuNC and 4a (I1; Fig. 4a), was unobservable. The tautomerization step to the aromatic 1Hpyrazole following the bimolecular cycloaddition step therefore seems to proceed substantially more rapidly for the phenol than for the carbamate. The remaining gap in the mechanism is the actual elimination step. We observed a striking dependence of pNA release on the presence of water (Fig. S31 †). In anhydrous DMSO, pNA release was quasi-absent; however, traces of water induced the rapid release of pNA. Water therefore participates in the release step. Several possible release pathways are conceivable. One possible mechanism could be elimination of the benzylic leaving group induced by deprotonation of the pyrazole. Alternatively, water could attack the imine with concerted electron migrations and elimination of the leaving group (Fig. S32 †). In summary, through a combination of DFT analysis and empirical studies, it was possible to establish and validate a likely reaction mechanism. The reaction cascade largely followed the predicted steps of cycloaddition, N 2 expulsion, tautomerization, and elimination, with the unexpected cleavage of the C-Si bond in case of TMS-MeNC. ## Demonstration of TzMe-deprotection on biomacromolecules, in cells, and in living vertebrates Many potential applications of the presented chemistry would require it to be compatible with living systems. We aimed to show that the developed chemistry can be performed under physiologically relevant conditions. We frst tested whether it is possible to conjugate a TzMe-modifed probe to a protein and unmask it with TMS-MeNC. For proof of principle, we used the SNAP-tag system 53 for protein labeling. We synthesized an O 6 -benzylguanine derivative of 4b (4b-BG, Fig. S33 †) and labelled a purifed SNAP protein (New England Biolabs) with it. The SNAP protein-4b-BG conjugate was exposed to TMS-MeNC (100 mM) and after 2 h, the uncaging of the fluorophore was examined by protein gel analysis (Fig. 6a and S34 †). The protein incubated with TMS-MeNC was visible by a strong in-gel fluorescence signal, whereas the fluorescence signal for controls was low. A 11-fold increase in the fluorescence signal was measured upon treatment with TMS-MeNC relative to untreated controls (Fig. S34 †). Mass-spectrometry experiments confrmed the labeling of the SNAP tag to afford the SNAP protein-4b-BG conjugate and the efficient removal of the TzMe group by TMS-MeNC (>80%) in the given time window (Fig. S35-S37 †). These results demonstrated that it is possible to conjugate TzMe-modifed groups to biomacromolecules and to actuate them by treatment with isonitriles thereafter. Next, we tested our chemistry with cultured cells. Restoration of the cytotoxicity of a Tzmoc-caged doxorubicin prodrug (Tzmoc-Dox (5), Fig. 6b; synthesis described in the ESI †) was tested with cultured A549 lung adenocarcinoma cells. In the presence of TMS-MeNC (100 mM) the prodrug was as toxic (EC 50 ¼ 0.239 AE 0.014 mM; Fig. 6b) as genuine doxorubicin (EC 50 ¼ 0.202 AE 0.025 mM; Fig. 6b). Tzmoc-Dox alone showed almost no toxicity below 10 mM, confrming the traceless activation of the doxorubicin prodrug. Exposure to 100 mM TMS-MeNC for 72 h caused no cell toxicity (Table S3 †). To demonstrate that TMS-MeNC can activate TzMe-modifed molecules in vivo, we performed experiments in zebrafsh embryos (Fig. 6c). The non-fluorescent TzMe-modifed fluorescein derivative 4b was injected into the yolk sac of zebrafsh embryos. The fsh were then incubated in either medium containing 20 mM TMS-MeNC or only its vehicle (DMSO) for 2 hours. Subsequently, the fsh were washed, and fluorescence turn-on signal analyzed by fluorescence microscopy (Fig. 6d). Strong green fluorescence staining localized to the yolk sac was observed for 4b-injected fsh incubated with TMS-MeNC, whereas 4b-injected control fsh treated with vehicle (DMSO) exhibited low fluorescence (Fig. 6d). A 3.8-fold higher fluorescence signal was measured in TMS-MeNC treated fsh relative to untreated controls (Fig. S38 †; p-value # 0.001). Exposure to 20 mM TMS-MeNC for the duration of the study caused no developmental issues to the zebrafsh embryos. These experiments establish that the reaction of TMS-MeNC and TzMe-groups is suitable for experiments with biomolecules and living organisms. ## Dual-release by combining TzMe-with ICPr-modied molecules There is considerable interest in developing reaction schemes that allow for the release of two molecules simultaneously in vivo. 24,39 We rationalized that combining the TzMe-release chemistry with our previously disclosed 3-isocyanopropyl (ICPr) chemistry, 37 would liberate two independent payloads. Dual release was frst tested in vitro. The TzMe-caged fluorescein dye 4b was incubated with ICPr-rsf, an ICPr-caged resorufn probe 37 (c(4b) ¼ 0.5 mM, c(ICPr-rsf) ¼ 1 mM, DMSO : PBS pH 7.4 (4 : 1), T ¼ 37 C, l ¼ 480 nm) and concurrent fluorophore release was analyzed by HPLC. The traceless release of both O-carboxymethyl fluorescein and resorufn was observed (Fig. S39 †). Dual release from combinations of TzMe/ICPr-reagents was then tested in vertebrates. Zebrafsh embryos were either injected with 4b or left untreated (Fig. 7a). The fsh were then incubated in media containing 10 mM ICPr-rsf for 2 hours, washed, and fluorescence turn-on signals analyzed by fluorescence microscopy (Fig. 7b and c). Strong emission signals were detected in the yolk sac in both green and red fluorescence channels for fsh injected with 4b. (Resorufn: pvalue # 0.0001; fluorescein: p-value # 0.0001; Fig. S38 †). Neither 4b (Fig. 6d) nor ICPr-rsf (Fig. 7c) alone produced obvious fluorescence signal confrming that it was the reaction between the isonitrile and the tetrazine that led to the concurrent release of the two fluorophores. While it is acknowledged that precisely controlling the injected probe volume into the yolk sac is challenging, the 60-fold higher resorufn (p-value # 0.0001) and 3.6fold higher fluorescein signal (p-value # 0.0001) in zebrafsh treated with both reactive species relative to controls indicate unmasking of a considerable fraction of the fluorophores (Fig. S38 †). Conclusively, combining TzMe-and ICPr-reactants can simultaneously liberate pairs of molecules of interest. ## Conclusions In summary, the study introduces TzMe-substituents as protecting groups that are removed under physiological conditions by isonitriles. In a series of steps, we demonstrated that TzMegroups could reversibly cage amines and phenols with nearquantitative release yields. Fast elimination occurred for phenols and for amines could be achieved by the addition of HSA or the use of TMS-MeNC. NMR and DFT studies revealed that the reaction followed the expected mechanism of cycloaddition and tautomerization to the imine. Unexpectedly, the elimination step could occur from the imine intermediate and this step involved the need for water. Furthermore, it was observed that in reactions with TMS-MeNC, the C-Si bond dissociated to generate a methanimine intermediate. TzMecaged fluorophore release on a protein, cytotoxicity experiments with a doxorubicin-prodrug and cultured cells and fluorophore release in zebrafsh embryos demonstrated the potential utility of the reaction in chemical biology and in the context of living systems. It is worth noting that the reaction with TMS-MeNC generates formaldehyde as side product. However, endogenous levels of formaldehyde (50-100 mM in serum and 200-500 mM in cells) 54 exceed the levels that would be released in most foreseeable applications. Furthermore, metabolic pathways counteract formaldehyde toxicity primarily mediated by glutathione 55 and formaldehyde is diverted to the one-carbon metabolism. 56 The toxicity of aldehyde side products of other isonitriles (e.g. butanal for nBuNC) is typically even lower. Combining TzMe-with ICPr-molecules allowed for the frst time the unmasking of two pro-fluorophores by a single bioorthogonal reaction. Multiple synergistic drug combinations would beneft from simultaneous and controlled delivery. Achieving release in a single reaction is important because controlling the delivery and stability of four individual reactants required for two reactions occurring in parallel would be challenging. This versatile protecting group chemistry constitutes a valuable addition to the dissociative bioorthogonal chemistry and synthetic methodology toolbox with potential utility for a broad range of applications. In addition to uses in drug delivery and controlling biomolecules, it may also be valuable as a protecting group for the synthesis of sensitive molecules allowing for late-stage deprotection under extremely mild conditions.

chemsum

{"title": "Isonitrile-responsive and bioorthogonally removable tetrazine protecting groups", "journal": "Royal Society of Chemistry (RSC)"}

high_yielding_acid‐catalysed_hydrolysis_of_cellulosic_polysaccharides_and_native_biomass_into_low_mo

## Abstract: Ionic media comprising 1-butyl-3-methylimidazolium chloride and the acidic deep eutectic solvent choline chloride/oxalic acid as co-solvent-catalyst, very efficiently convert various cellulosic substrates, including native cellulosic biomass, into watersoluble carbohydrates. The optimum reaction systems yield a narrow range of low molecular weight carbohydrates directly from cellulose, lignocellulose, or algal saccharides, in high yields and selectivities up to 98 %. Cellulose possesses significant potential as a renewable platform from which to generate large volumes of green replacements to many petrochemical products. Within this goal, the production of low molecular weight saccharides from cellulosic substances is the key to success. Native cellulose and lignocellulosic feedstocks are less accessible for such transformations and depolymerisation of polysaccharides remains a primary challenge to be overcome. In this study, we identify the catalytic activity associated with selected deep eutectic solvents that favours the hydrolysis of polysaccharides and develop reaction conditions to improve the outcomes of desirable low molecular weight sugars. We successfully apply the chemistry to raw bulk, non-pretreated cellulosic substances. ## Introduction Biomass is the principal renewable resource for sustainable industrial production of high-volume and high-value chemicals. Within natural sources, cellulose is the most abundant substrate with the scale to reduce reliance on fossil fuel-derived bulk chemicals. There is an intense effort to efficiently transform polysaccharides into small organic building block molecules (platform chemicals) generating a source of renewable replacements to crude oil-based products. In the presence of an acid catalyst, cellulose hydrolyses into low molecular weight glucans and monomer glucose, which are convertible into a range of value added molecules with high potential for manufacturing applications (Scheme 1). Platform chemicals are readily accessible on industrial scale but the present production is mostly based on refined edible sugars (such as glucose, fructose, sucrose, or starch). This reliance undermines the sustainability of the biorefinery and becomes a source of controversy and public concern. Cellulose-derived low molecular weight sugars can potentially provide an inexhaustible source of substrates for useful chemicals (Scheme 1). Cellulose is composed of monomer units suitable for conversion into platform chemicals but cellulose has an intractable structure which renders challenging the direct transformation thereof into platform chemicals. The synthesis of platform chemicals directly from cellulose thus remains mostly industrially unviable. The ongoing major challenge relates to the efficient depolymerisation of cellulose and its concomitant hydrolysis into low molecular weight water-soluble saccharides under green processing conditions. An excellent example of the direct transformation of biomass into platform chemicals is the Biofine Process. High temperature reaction of cellulosic biomass with dilute mineral acids yields levulinic acid, formic acid, furfural and biochar. Stage 1 processing of a high cellulose-content substrate at 210 °C affords low molecular weight saccharides and 5-(hydroxymethyl)furfural (HMF), while Stage 2 reaction at below 200 °C converts the sugars into levulinic and formic acids as major products. Importantly, the procedure requires the preprocessing of the lignocellulosic raw material to remove hemicellulose and generate the high cellulose-content feed for the reactors. Equally importantly, the two-stage process requires the hydrolytic formation of low molecular weight sugars from cellulose, which are converted into the desired platform products. The hydrolysis of cellulose is therefore a critical step in the chemical transformation of this substance. Ionic liquids (ILs) are efficient solvents for reactions of cellulosic materials. [9a] These ionic media can fully dissolve polysaccharides, and thereby can promote chemical transformations into highly desirable products under mild reaction conditions; in contrast, common aqueous media cannot dissolve cellulose and require forcing processing conditions to promote catalytic reactions. [3,9b,c] In particular, mineral acids (e. g., hydrochloric, sulfuric or phosphoric acids) or solid acid catalysts (e. g., acidic resins or carbonaceous acids) in imidazolium-based ILs have been heavily explored in the hydrolysis of microcrystalline cellulose (MCC) to low molecular weight reducing sugars and sometimes for its conversion directly into platform chemicals such as HMF. In some instances the imidazolium quaternary salts possess acidic functional groups, which avoids the need to add catalysts. Importantly, ionic solvents and saccharides can be fully recovered after catalytic processing by potentially scalable methods. [14a] The low molecular weight sugars so produced are amenable to fermentation into alcohols. Much of the work presented in the current literature relating to the hydrolysis of cellulose employs pretreated cellulose such as MCC, obtained by the treatment of cellulose with mineral acids, or ball-milled cellulose. [9b,c,10,11,13,15,16] Some works present the use of unfractionated biomass. However, many chemical treatments detailed in the literature have not been demonstrated on native-or raw biomass, or even the type of cellulose that is available via the large volume cellulose refining technologies applied in the paper and pulp industry. While common imdazolium-based systems present some advantages in the processing of cellulose, deep eutectic solvents (DESs) are somewhat overlooked in this arena. DESs are alternative media (to common ILs) usually formed from eutectic mixtures of Lewis or Brønsted acids and bases under solvent-free conditions. These solvents can be produced from inexpensive plant-based substances and many DESs are considered to be environmentally benign reaction systems for the processing of carbohydrates. Especially, the combinations of choline chloride (ChCl) with organic acids prove to be useful green media for chemical conversions of some poly-and oligosaccharides. The intrinsic acidity of such DESs facilitates the transformation of inulin and hemicellulose into monomer sugars and ultimately into furan-type molecules in a single solvent system. Despite the promising characteristics of DES systems, cellulose largely resists dissolution in DES. The optimum conditions deliver solutions containing a maximum of 6.5 wt% cellulose. In this instance, a specific cotton linter pulp with a low degree of polymerisation is used and the observations are not generalised (this material is known as a dissolving pulp, which is soluble in aqueous solutions of mineral bases). If native cellulose can be dissolved and selectively chemically transformed in the presence of DESs then substantial progress would have been made in the processing of biomass into useful products, offering an area with significant scope for exploration. The present work uncovers and demonstrates the functionality of DESs as reactive (catalyst) co-solvents in imidazoliumbased solvents for the conversion of cellulose and native biomass into low molecular weight carbohydrates under green processing conditions. It researches the chemistry associated with catalytic reactions of polysaccharides, and develops novel acidic systems for their effective employment in biorefinery settings. Most importantly, we successfully apply the optimised conditions to the hydrolysis of raw bulk cellulosic substrates of terrestrial and marine origin, including an example of an agricultural waste product. Many current state-of-the-art systems employ vigorous pretreatment methods such as extensive ball milling. [9b,c] Such energy intensive techniques are more suited to scientific studies than to large scale industrial manufacture. In contrast to most related studies which typically report for glucose, we track the formation of (glucose) n Scheme 1. Catalytic conversion of cellulose into value added molecules. n = integer, R = H, or alkyl. saccharides for n = 1-4 by LC-MS, thereby improving our understanding of the course of the hydrolysis chemistry. ## Results and Discussion To explore the course of the hydrolytic reaction of cellulose in various ILs, we investigated the conversion of MCC and of typical low-molecular-weight sugars that commonly appear during the hydrolysis of cellulose (i. e., cellobiose, glucose, fructose; Scheme 1). Reactions were conducted in 1-butyl-3methylimidazolium chloride ([C 4 mim]Cl) or in DESs based on ChCl and oxalic or citric acids, or on ChCl and TsOH, at 120 °C for 2 h (Table 1 and in Table S1). Neither the imidazolium IL nor the DES systems were suited to the task, for different reasons. While MCC is freely soluble in [C 4 mim]Cl, it does not hydrolyse notably in this solvent under our conditions, and the masses of the input and recovered cellulose were practically identical (Table 1, entry 1). Within the low molecular weight saccharides, cellobiose was hydrolysed into glucose in [C 4 mim]Cl in only low yield (8 wt%), while glucose or fructose provided a little HMF (up to 6 mol%), but were otherwise unchanged (Table S1). The imidazolium-based solvent clearly possesses some Brønsted acidity which catalyses certain reactions to a limited extent (e. g., hydrolysis of glycosidic bonds, dehydration into HMF), but this solvent-derived acidity is insufficient to catalyse the transformation of cellulose into products. On the other hand, DES systems based on ChCl/oxalic acid or ChCl/TsOH promoted more comprehensive transformation of MCC under the selected reaction conditions (conversion up to 46 %, Table 1, entries 2 and 4) but mostly into undesirable high molecular weight humins (even though MCC remained as a suspension in the DES). Humins are thought to form by the condensation of intermediate sugars and aldehydes. Similarly, low molecular weight carbohydrates in ChCl/oxalic acid DES were almost completely transformed (total conversion to product of 96, 97, and 99 % for cellobiose, glucose, and fructose, respectively, Table S1) into a bulk of undesirable humins and small amounts of HMF (yields up to 11 mol%). The results clearly suggest that neither imidazolium-based solvent nor the acidic DESs are suitable reaction media under our conditions, due to a) the low acidity of [C 4 mim]Cl which fails to promote the catalytic conversion of cellulose and b) the high acidity of DESs inducing the formation of by-products. These drawbacks are likely to be mutually exclusive and a combined solvent system based on [C 4 mim]Cl and acidic DESs should favour the more selective transformation of cellulose into desirable low molecular weight sugars. To test this hypothesis, the transformation of MCC was conducted in mixed [C 4 mim]Cl/DES systems, in which the intention was to employ the acidic DES to catalyse the hydrolysis (Table 1, entries 5-7, 9-14). A combination of [C 4 mim]Cl and ChCl/oxalic acid 10 : 1 w/w showed high selectivity towards low molecular weight carbohydrates (total selectivity is 81 % for the products glucose, cellobiose, cellotriose, and cellotetraose) after reaction of MCC at 120 °C for 2 h (Table 1, entry 5). This mixed solvent readily dissolves cellulose and, based on observation, has lower viscosity than neat [a] Yields are specified in wt % based on input of cellulose for carbohydrates and in mol % based on anhydroglucose units present for HMF; '0' means that product was identified in trace amounts based on HPLC analysis; X = conversion; S = total selectivity of carbohydrates (glucose, cellobiose, cellotriose, and cellotetraose). Reaction conditions: MCC (50 mg), IL or DES (1.000 g), 120 °C. [b] Reaction conditions: MCC (50 mg), [C 4 mim]Cl (1.000 g), DES (0.100 g), 120 °C. [c] Reaction conditions: MCC (50 mg), [C 4 mim]Cl (1.000 g), oxalic acid dihydrate (48 mg), 120 °C. [C 4 mim]Cl, which alleviates one of the downsides attributed to the use of imidazolium-based ILs. Other DESs based on citric acid or TsOH, combined with [C 4 mim]Cl, are less effective for the hydrolysis of cellulose under similar conditions (Table 1, entries 6 and 7). Possibly, the solvent system with ChCl/citric acid possesses insufficient acidity to catalyse the hydrolysis under the prevailing conditions (see 1, entries 13 and 14) led to the formation of HMF and by-product humins, which reduced the overall selectivity. HMF likely derives from fructose, the isomerisation product of glucose, while humins form by condensation of sugars and HMF. Scheme 2 summarises the findings. We sought to better understand the origins of the catalyst activity in the ChCl systems. pH readings of dilute aqueous solutions of the DES systems and of their parent acids revealed that the ionic solvents possess enhanced Brønsted acidity compared to the parent acids, in the cases of the organic acids (Table 2). All readings were performed at the same concentration of the components being measured and are therefore a measure of acid strength. ChCl, which is considered to be a Lewis acid, apparently forms a Lewis acid-assisted Brønsted acid complex with oxalic acid; this complexation assists to deliver the higher acidity of the DES (Scheme S1). Most likely, the complex so formed consists of choline cation and oxalate anion, as proposed in Scheme S1. Interestingly, the induced Brønsted acidity catalyses the esterification of oxalic acid with ChCl (Scheme S1). The esterification of carboxylic acids with ChCl in molten media, especially with added hydrochloric acid, has been previously noticed by Florindo et al. In our hands, 13 C NMR analysis of ChCl/oxalic acid (1 : 1 mixture, dissolved in deutero acetonitrile) demonstrates three chemical shifts associated with the carboxylic groups of oxalic acid (one peak associated with free oxalic acid at 160.2 ppm, and two new peaks at 158.9 and 158.7 ppm that are consistent with the formation of the ester, Figure S1), along with six chemical shifts assigned to the carbon atoms of choline cation (three peaks associated with free ChCl at 68.5, 56.5 and 54.7 ppm, and three new peaks assigned to the ester at 64.8, 60.5, and 54.6 ppm, Figure S1). Integration of the peaks in the 1 H NMR spectrum of ChCl/oxalic acid diluted in deuterated solvents (CD 3 CN or D 2 O, Figure S2) or using neat DES (with DMSO-d 6 for external lock employing a coaxial insert tube, Figure S2), showed the ester and ChCl to be present in a ratio 1 : 10. The FTIR spectrum of the DES indicates a band at 1723 cm 1 characteristic of the asymmetrical C=O stretching mode and a band at 1185 cm 1 corresponding to asymmetrical vibration of C C(=O) O (Figure S3). Free oxalic acid or ChCl do not provide these vibrations, which we propose to be associated with the ChCl/ oxalic acid ester. Now with reference to the experimental results (Table 1) and the pH measurements (Table 2), while Table 2 shows only incremental differences between the various acidic systems, there are distinct experimental outcomes associated with the different reaction media (Table 1). Of the media probed, the acidity secured through ChCl/oxalic acid is optimal for the conversion of cellulose into water-soluble low molecular weight carbohydrates. Other DESs showed either high Brønsted acid activity leading to the rapid conversion of cellulose into byproducts, or too low acidity for the effective hydrolysis of glycosidic bonds. The experimental data point to a need for sufficient acidity to cause the hydrolysis but for a tight balancing act such that the acidity is not so high as to cause secondary reactions under the conditions. Although cellulose-derived oligoglucans (such as those produced as described above: cellobiose, cellotriose, cellotetraose) are valuable products, glucose is usually considered to be the desired product of the hydrolysis of polysaccharides. Metal triflates are efficient green Lewis acidic catalysts for a range of chemical transformations, including the processing of cellulose, especially when mixed with Brønsted acids to form Lewis acid-assisted Brønsted acid complexes. Because DESs possess intrinsic Brønsted acidity as shown above (Table 2, Scheme S1), in theory, it is possible to advantage the conversion of cellulose in favour of glucose by modifying solvent/catalyst system with Lewis acid. Accordingly, the activity of metal triflates (Al(OTf) 3 , Y(OTf) 3 , AgOTf, In(OTf) 3 , Sn(OTf) 2 , La (OTf) 3 , Yb(OTf) 3 and Hf(OTf) 4 ) in mixed ionic solvent [C 4 mim]Cl/ ChCl/oxalic acid was investigated for the conversion of MCC. Table 3 presents that metal triflates, specifically AgOTf, In(OTf) 3 , Sn(OTf) 2 , La(OTf) 3 , and Yb(OTf) 3 , promote the rapid hydrolysis of cellulose into low molecular weight sugars. Those that afford higher acidity (Al(OTf) 3 , Y(OTf) 3 and Hf(OTf) 4 ) favour the formation of HMF and the unwanted by-product humins. La (OTf) 3 possessed the highest activity to catalyse the transformation of MCC directly into desired monomer glucose in high yield (35 wt% based on cellulose), and in shorter reaction time compared to the process without metal triflate catalysts. La(OTf) 3 and oxalic acid formed a Lewis acid-assisted Brønsted acid in the DES, improving the overall activity of the catalytic media for hydrolysis of glycosidic bonds. The addition of La (OTf) 3 leads to slightly higher acidity (Table 2). This is caused by complexation of the La to the oxalic acid, thereby providing a complementary form of Lewis acid-assisted Brønsted acidity. This complexation was unambiguously demonstrated by FTIR analyses of [C 4 mim]Cl ionic liquid solutions of La(OTf) 3 mixed with oxalic acid and separately La(OTf) 3 mixed with ChCl/oxalic acid (Figure S4). Both mixtures showed the appearance of an asymmetrical stretching band at 1620-1600 cm 1 of the carboxylate anion confirming the complexation. As we have already discussed for the La(OTf) 3 /H 3 PO 4 system, La holds a special place relating to Lewis acid-assisted Brønsted acidity, providing a complex acid system with exceptional catalyst activity in several different chemical transformations. pH readings of the mixed acid systems (La(OTf) 3 + oxalic acid) show the increased acidity, albeit that it too appears incremental in dilute solution. Notwithstanding this incremental change to the pH, the experimental results (Table 3) provide strong evidence for the benefits brought by the addition of the Lewis acid, leading to 73 % conversion of cellulose within two hours into 69 % low molecular weight saccharides (w/w based on the input cellulose) and 95 % selectivity for such saccharides. Care needs to be taken with reaction times, though. The heightened activity of the reaction media with La(OTf) 3 is detrimental after extended periods due to the transformation of glucose into HMF and ultimately into humins (Table 3, Scheme 2). The hydrolytic conversion of cellulose in ILs is more efficient in the presence of water, favouring the formation of glucose. We therefore conducted the transformation of MCC in the mixed ionic solvent [C 4 mim]Cl / ChCl/oxalic acid at 120 °C with the addition of water (Figure 1). MCC was first dissolved in the ionic solvent at 100 °C for 2 h, after which the temperature was raised to 120 °C to cause the reaction, and water was added to the reaction media after 0.5 h (water content 20 wt% based on IL) and 1 h (water content 30 wt% based on IL) of reaction time. The glucose yield is dramatically enhanced in the presence of water. Recall that in the absence of added water, cellulose yielded (with a maximum at 12 h) 28 wt% glucose, along with 43 wt% oligosaccharides consisting of 11, 16 and 16 wt% of cellobiose, cellotriose and cellotetraose, respectively (Table 1, entry 12). In contrast, in the presence of water, glucose was the major product (maximum at 4 h, 49 wt %, Figure 1), with only little accumulation of oligosaccharides (maximum at 2 h, 8, 9, 7 wt% of cellobiose, cellotriose and cellotetraose, respectively). These results suggest that added water improved the rates of hydrolysis of glucans. Other studies also note that water suppresses the conversion of glucose into HMF in ILs, most likely related to the reduced acidity of the diluted media. Longer reaction times nonetheless led to diminished yields of glucose by the formation of HMF and humins (Figure 1). All of the reactions performed to this stage had been conducted using MCC as a substrate. MCC is a polysaccharide obtained after the acid-catalysed depolymerisation of native cellulose and is a commodity product for many industries. However, the conversion of non-pretreated substrates is desirable. A subsequent set of reactions was performed employing non-pretreated cellulose of various origins (cotton linter, cellulose extracted from eucalyptus and Pinus, microalgal biomass, macroalgal biomass). Processing of non-pretreated bulk cellulose in the co-solvent system [C 4 mim]Cl / ChCl/oxalic acid at 120 °C for 6 h afforded lower yields of low molecular weight reducing sugars compared to MCC (Table 4, entries 1, 3, 5, 8, 10), showing the difficulties experienced when working with native biomass, and the need to improve and modify reaction conditions. Depolymerisation of bulk cellulose requires more forcing conditions compared with MCC, and therefore the overall rate of hydrolysis is lower. The addition of La(OTf) 3 slightly improved the transformation of cellulose extracted from eucalyptus to glucose (yield 20 wt%, Table 4, entry 6), but the effect of the Lewis acid is less prominent, when compared to the conversion of MCC under identical reaction conditions (glucose yield 35 wt%, Table 3). Similar to the processing of MCC, longer reactions were accompanied by the formation of humins. Pleasingly, the addition of water in two steps as before, improved the hydrolysis of polysaccharides, affording excellent yields of glucose (45-50 wt%, based on substrate, Table 4, entries 4, 7, 9, 11). This is an efficient process for the conversion of a range of non-pretreated cellulosic substrates into glucose, involving environmentally benign solvent-catalyst media. The results compare favourably with outcomes in the literature where mineral acids or zeolites are employed as catalysts for cellulose hydrolysis in [C 4 mim]Cl (Table 4, entries 21-23). The conversion of native cellulosic biomass is a significant challenge, and we applied our hydrolysis protocols to biomass derived from terrestrial (chips obtained from softwood or corncob) and marine sources (macroalgae Ulva lactuca and microalgae Porphyridium cruentum), respectively. Optimal conditions for each source of cellulose or biomass and the reaction outcomes are given in Table 4, entries 12-20. The direct processing of biomass is inherently difficult because native cellulose is usually entangled into plant cell walls with other polysaccharides (e. g., hemicellulose, polymannosides, glycoproteins, etc.) and aromatic polymers (e. g., lignin) forming a rigid polymer system. In our hands, the transformation of softwood chips (Table 4, entry 13) in the mixed solvent [C 4 mim]Cl/ ChCl/oxalic acid yielded low molecular weight carbohydrates (glucose + glucosyl oligosaccharides) in a respectable 38 wt% yield based on the glucan content in the biomass. This required processing of wood chip biomass at 120 °C for 6 h, followed by further processing for 4 h after the addition of water (30 wt% of water, based on solvent, added as detailed in Table 4, entry 13). Under these conditions, the xylans (part of the hemicellulose) were hydrolysed into monomer xylose in 25 wt% yield (based on xylan content in the biomass), demonstrating the hydrolysis of both linear and branched polysaccharides (Table 4, entry 13). Most likely, the lower yields of low molecular weight carbohydrates are caused by complicated depolymerisation of wood biomass in ILs. Nevertheless, higher yields of glucose (25 wt%) and xylose (30 wt%, Table 4, entry 14) were attainable after somewhat extended processing of the wood chips (120 °C, 12 h) before addition of water and further heating. In distinct contrast to softwood, the conversion of corncob provided excellent yields of glucose (54 wt%) and xylose (35 wt%) under milder processing conditions before the dilution (100 °C, 2 h, Table 4, entry 16), likely due to the less rigid molecular structure and larger amount of structurally branched polysaccharides present (e. g., hemicellulose and starch). For the same reason, marine cell walls were more amenable to hydrolysis. For example, the conversion of the seaweed Ulva lactuca in [C 4 mim]Cl / ChCl/oxalic acid at 120 °C for 6 h, followed by addition of water (30 wt% of water) and further heating at 120 °C for 4 h, provided glucose in 40 wt% yield (based on the glucan content, Table 4, entry 17). This yield can reach 43 wt% when adding water at the 4 h mark (instead of at 6 h, Table 4, entry 18). The experimental data confirm that the branched saccharides which are predominant in marine plant cell walls require less forcing reaction conditions for selective conversion into desirable low molecular weight saccharides. In a remarkable example, the direct processing of microalgae P. cruentum, as raw biomass, in [C 4 mim]Cl / ChCl/oxalic acid, yielded 55 wt% of glucose and 40 wt% of xylose (based on the content of glucans and xylans in biomass, respectively, Table 4, entry 20), giving 98 % conversion of the biomass. These experimental outcomes shine light on the exquisite promise held by the direct conversion of biomass into significantly higher value and useful monosaccharides. One persistent drawback is the need to employ large volumes of the IL solvent, and this challenge remains to be solved. Nevertheless, the results serve as a springboard towards scalable processes to manufacture sustainable and renewable chemicals (Scheme 1). ## Conclusions The combined ionic liquid mixture of [C 4 mim]Cl/ChCl/oxalic acid is an excellent solvent-catalyst system for the high yielding and selective conversion of cellulose and native biomass, of terrestrial and marine origin, into the low molecular weight saccharides glucose, cellobiose, cellotriose, cellotetraose and xylose. We demonstrate that the acid-catalysed transformation of cellulose in mixed solvents occurs predominantly into glucan oligomer 'chunks' (cellotetraose, cellotriose and cellobiose) from which glucose emerges. The conversion into glucose can be improved by modifying the natural acidity of the DES with added Lewis acid, producing a Lewis acid-assisted Brønsted acid complex, or by the addition of water during the course of the reaction. Importantly, the mixed system avoids the need for pretreatment of the native cellulosic materials. While efficient in this process, the need for large volumes of the ionic system remains to be solved. ## Experimental Section Materials Reagents and metal trifluoromethanesulfonate (metal triflate) catalysts (Al(OTf) 3 , Y(OTf) 3 , AgOTf, In(OTf) 3 , Sn(OTf) 2 , La(OTf) 3 , Yb (OTf) 3 , or Hf(OTf) 4 ) were used as supplied from commercial sources. Cotton linter, and cellulose extracted from eucalyptus and Pinus (unbleached and bleached, BKT, Kinleith, New Zealand) were a 13 [b] 14 [d] Wood chips (softwood) [g,11] Sigmacell ---38 ----22 [h,11] ---21 ----23 [i,15] MCC ---50 (mol %) 5 (mol %) --24 [a] Yields are specified in wt % based on input of cellulose for carbohydrates and in mol % based on anhydroglucose units present for HMF; yields of glucose, cellobiose, cellotriose and cellotetraose obtained from lignocellulose, or algal biomass, are specified in wt % based on the glucans content in substrate; yields of xylose are specified based on the xylans content in biomass; '0' means that product was identified in trace amounts based on LC analysis; X = conversion; S = total selectivity of carbohydrates (glucose, cellobiose, cellotriose, cellotetraose °C, 45 min. [h] Reaction conditions: substrate (0.32 g), [C 4 mim]Cl (4.0 g), HCl (36 wt%, 0.285 g), water (0.063 g), 100 °C, 11 min. [i] Reaction conditions: substrate (0.1 g), [C 4 mim]Cl (2.0 g), 130 °C, to complete dissolution, then addition of HY-zeolite (11 mol%) and water in three steps (step 1: water content 5 wt %, based on IL, t = 0; step 2: water content 20 wt%, based on IL, t = 0.5 h; step 3: water content 33 wt%, t = 60 min ), 130 °C, 2 h. generous gift from Dr Simon Hinkley, The Ferrier Research Institute, Victoria University of Wellington (New Zealand). Lignocellulose (softwood chips and corncob) was sourced from local growers (Australia). Macroalgae Ulva lactuca was provided as a generous gift by Dr Wayne O'Connor, Department of Primary Industries Fisheries, Port Stephens Fisheries Institute (Australia). Microalgae Porphyridium cruentum was grown and supplied by Climate Change Cluster (C3), University of Technology Sydney (Australia). Biomass for acidcatalysed reactions was vacuum oven-dried (60 °C, 12 h). Compositional analysis of biomass was performed using standard analytical procedures: NREL/TP-510-42618 for lignocellulose, NREL/TP-5100-60957 for algal biomass. The total amount of carbohydrates in the given biomass is specified in the Supporting information Table S2. [C 4 mim]Cl was prepared according to reference, while ChCl/acid (1 : 1 molar ratio for oxalic acid dihydrate or ptoluenesulfonic acid monohydrate, respectively; 1 : 0.5 for citric acid monohydrate, respectively) solvents were prepared according to reference. High Performance Liquid Chromatography ## IR Spectrometry IR spectra were collected using a thin film on a Thermo Scientific Nicolet 6700 spectrometer in a range 4000-450 cm 1 . pH Readings pH Readings were recorded at room temperature (22-23 °C) using a Mettler Toledo pH meter adapted with a standard glass electrode with prior calibration by two buffer solutions (pH = 4.00, pH = 7.00). Measurements were performed in triplicate and the average values are presented. ## Acid-Catalysed Conversion of Cellulosic Substrates MCC, cellobiose, glucose, or fructose (50 mg) and solvent ([C 4 mim] Cl, or DES, 1.000 g) were loaded to a glass pressure tube equipped with a magnetic follower and the reactor was sealed. The mixture was heated and stirred at 120 °C for 2 h. After completion of the process, the reaction mixture was cooled and diluted with deionised water (9.00 mL) to precipitate any unreacted cellulose. The mixture was centrifuged (10,000 × g for 15 min) and decanted. The recovered solids were washed with deionised water (3 × 10 mL), vacuum oven-dried (60 °C, 1 mbar, 12 h) and weighed to calculate the conversion of cellulose. The decanted liquid phase was quenched by the addition of an aqueous solution of sodium hydrogen carbonate (1.00 mL, 0.05 M) and centrifuged (10,000 × g for 15 min). The recovered solutions were diluted with a known volume of deionised water, if required, and analysed by HPLC. To recover HMF from the diluted and neutralised ILs (50.0 mL, combined aqueous phases after conversions of carbohydrates), an extraction with ethyl acetate (3 × 100 mL) was performed. The ethyl acetate layers were combined and dried over MgSO 4 and the solvent was removed under reduced pressure. The crude product was subjected to flash column chromatography (hexane/ethyl acetate, 60 : 40 v/v) to isolate HMF, which gave satisfactory analytical data. For the conversions in mixed ionic solvent, cellulose (50 mg), [C 4 mim]Cl (1.000 g) and DES (0.100 g), and in some instances metal triflate catalyst (Al(OTf) 3 , Y(OTf) 3 , AgOTf, In(OTf) 3 , Sn(OTf) 2 , La(OTf) 3 , Yb(OTf) 3 , or Hf(OTf) 4 , 10 mol% based on the number of anhydroglucose units present in cellulose), were introduced to a glass pressure tube equipped with a magnetic follower and the reactor was sealed. The mixture was heated and stirred at the predetermined temperature for a fixed period of time. The products were recovered and analysed as detailed above. For the conversions with gradually added water, cellulosic substrate (cellulose, lignocellulose, or algal biomass, 50 mg), [C 4 mim]Cl (1.000 g) and DES (0.100 g) were charged to a glass pressure tube equipped with a magnetic follower and the reaction mixture was heated and agitated for a fixed amount of time. After this fixed period of time (which varied from case to case), deionised water was added (0.220 mL, water content 20 wt%, based on IL) followed by heating, and 30 minutes later a second portion was added (0.110 mL, total water content 30 wt%, based on IL) and the reaction system was additionally heated and stirred for a fixed period of time. The products were recovered and analysed as detailed above.

chemsum

{"title": "High Yielding Acid\u2010Catalysed Hydrolysis of Cellulosic Polysaccharides and Native Biomass into Low Molecular Weight Sugars in Mixed Ionic Liquid Systems", "journal": "Chemistry Open"}

enhanced_aging_properties_of_hkust-1_in_hydrophobic_mixed-matrix_membranes_for_ammonia_adsorption

## Abstract: Metal-organic frameworks (MOFs) in their free powder form have exhibited superior capacities for many gases when compared to other materials, due to their tailorable functionality and high surface areas.Specifically, the MOF HKUST-1 binds small Lewis bases, such as ammonia, with its coordinatively unsaturated copper sites. We describe here the use of HKUST-1 in mixed-matrix membranes (MMMs) prepared from polyvinylidene difluoride (PVDF) for the removal of ammonia gas. These MMMs exhibit ammonia capacities similar to their hypothetical capacities based on the weight percent of HKUST-1 in each MMM. HKUST-1 in its powder form is unstable toward humid conditions; however, upon exposure to humid environments for prolonged periods of time, the HKUST-1 MMMs exhibit outstanding structural stability, and maintain their ammonia capacity. Overall, this study has achieved all of the critical and combined elements for real-world applications of MOFs: high MOF loadings, fully accessible MOF surfaces, enhanced MOF stabilization, recyclability, mechanical stability, and processability. This study is a critical step in advancing MOFs to a stable, usable, and enabling technology. ## Introduction With an annual production over 200 million tons, ammonia is one of the most widely manufactured chemicals in the world. 1 Ammonia has been identifed as a chemical that frequently creates a high risk for accidents such as spills at manufacturing facilities or explosions at fertilizer plants. 2 Furthermore, the availability and toxicity of ammonia make it a potential chemical for insurgents to utilize in asymmetric warfare. In the 1990s, Serbian forces targeted chemical plants during the war in Croatia as a method of attack toward civilians by causing ammonia release into the environment. 3 For these reasons, the development of engineered materials that can remove large amounts of ammonia for air purifcation applications is paramount. Metal-organic frameworks (MOFs) are porous materials comprised of inorganic metal nodes, known as secondary building units (SBUs), linked together by polydentate organic ligands. 4,5 The various combinations of SBUs and organic linkers allow for tuning of the physical and chemical properties. Many MOFs are microporous, making them ideal for gas storage, gas separations, 10,11 molecular sensing, 12,13 toxic chemical adsorption, 14,15 and catalysis. 16,17 While MOFs have been examined extensively in their primitive powder form, far fewer studies have been conducted on their properties in engineered forms or as part of a matrix. To incorporate MOFs into materials for applications in textiles, flters, or sensors, engineered forms, such as pressed pellets, particles with binders, or flms, must be fabricated. Studies of pure-MOF membranes have been reported, but typically only small area samples can be achieved, and delamination typically proves difficult. 23 Mixed-matrix membranes (MMMs) of MOFs have been reported primarily for the study of gas separations. MMMs have the potential to enhance the utility of MOFs by allowing for the facile fabrication of supported or freestanding flms with variable material composition that exhibit mechanical and material properties beyond that of single crystals or free flowing powders. Recently, the use of polyvinylidene difluoride (PVDF) to prepare MMMs for a wide variety of MOFs was described. 24 PVDF-based MMMs of MOFs exhibited good mechanical properties, while allowing for high weight percent loading of MOFs. Furthermore, the crystallinity, surface area, and chemical reactivity of the MOFs was largely unperturbed in these MMMs. HKUST-1 (aka. Cu-BTC, Cu 3 (BTC) 2 , MOF-199, HKUST-1 ¼ Hong Kong University of Science and Technology) is a copper based MOF, in which paddlewheel Cu-dimers are linked together by benzene 1,3,5-tricarboxylate to form a 3-dimensional pore structure, as seen in Fig. 1a. 25 HKUST-1 has been shown to be superior to other MOFs for the adsorption of basic gases such as ammonia, due to coordinatively unsaturated Cusites. 26,27 Recently, HKUST-1 was incorporated into biological chitin fbers at loadings up to 55% (w/w), while maintaining approximately 75% of its ammonia capacity (based on HKUST-1 content). 28 One of the major shortcomings of HKUST-1 is its instability toward liquid water and high humidity conditions. Upon exposure to 90% RH at 25 C, HKUST-1 exhibits a substantial loss of its porosity and ammonia uptake capacity, with a transformation of the crystal structure. 31 It has been shown that the chemical stability of HKUST-1 can be enhanced via plasma enhanced chemical vapour deposition of perfluoroalkanes into the MOF structure. 32,33 The hydrophobic nature of fluoroalkanes prevents water molecules from clustering in the pores and subsequently breaking the Cu-carboxylate linkages. Despite the success of this approach, it does not address the myriad of other issues required to make a suitable gas capture device from MOFs, including retaining fully accessible MOF surfaces, facile recyclability, mechanical stability, and processability. Here we address all of these issues with a single solution. We explore the use of a hydrophobic polymer (PVDF) to prepare HKUST-1 MMMs of various composition for ammonia removal, as well as the effect of humidity on the performance of the MOF in the MMM. We fnd that the ammonia removal performance of HKUST-1 in these MMMs is unprecedented, with greatly improved stabilization of the MOF toward both humidity and ammonia, when compared to HKUST-1 powder. 31 ## Mixed-matrix membrane preparation The preparation and initial characterization of PVDF MMMs with a variety of MOFs was recently reported. 24 Fabrication of these MMMs, described in depth in the ESI, † involves an ink comprised of a solution of PVDF polymer and HKUST-1 that is applied to a substrate followed by solvent evaporation to form a flm that can be delaminated to give a freestanding MMM (Fig. 1). The samples throughout this manuscript are named [wt%]-HKUST-1 MMM, where [wt%] ¼ 30, 50, or 67. For the MMMs, as the HKUST-1 content increases, so does the intensity of the HKUST-1 peaks in the PXRD spectra (Fig. S1 †). Likewise, the FTIR spectra (Fig. S2 †) show that as the PVDF/HKUST-1 content varies for each sample, the corresponding infrared bands for PVDF and HKUST-1 vary in the same manner. ## Ammonia adsorption Ammonia uptake capacities were measured for each material using dynamic microbreakthrough tests at a concentration of 2000 mg m 3 (see ESI †). The ammonia capacities, as determined from the breakthrough curves (Fig. S3 †), for the HKUST-1 powder and HKUST-1 MMMs are shown in Table 1. HKUST-1 exhibited an ammonia capacity of 7.4 mol kg 1 , which is commensurate with a loading of approximately 1.5 NH 3 molecules per Cu atom. It has been shown that even at low pressures each Cu atom of HKUST-1 ligates one NH 3 molecule. 27 The additional ammonia sorption is likely due to hydrogen bonding with the chemisorbed NH 3 molecules in the MOF pore. The ammonia capacity of each MMM varies proportionally with HKUST-1 content. Interestingly, as the HKUST-1 content increases in the MMM, the experimental ammonia capacity agrees better with the hypothetical capacity, as determined from the weight percent of HKUST-1 in each MMM. The good agreement between the experimental and hypothetical capacities, especially for 50-HKUST-1 and 67-HKUST-1 MMM, is strong evidence that the HKUST-1 crystals even within the interior of the MMM are largely accessible to the contaminated airstream. The pre-and post-ammonia exposed HKUST-1 MMMs, HKUST-1 powder, and PVDF were analyzed using PXRD (Fig. S1 and S4 †) and FTIR (Fig. S2 and S5 †). The HKUST-1 powder exhibited a change in the PXRD pattern upon exposure to ammonia, with new reflections at 2q ¼ 18.1, 25.3, and 27.0 , indicative of a substantial phase change and the loss of the HKUST-1 structure. In contrast, the PXRD patterns of each HKUST-1 MMM showed minimal change, indicating that HKUST-1 maintains its crystallinity better in the MMMs upon exposure to ammonia, when compared to the HKUST-1 powder. HKUST-1 powder exposed to ammonia also showed a loss of the FTIR band at 1646 cm 1 , which is indicative of Cu-carboxylate bonding. 31 These characteristics are consistent with ammonia degrading the MOF structure, as seen in earlier reports. 26 The appearance of the bands at 1610 cm 1 in the FTIR spectrum upon the exposure of HKUST-1 to ammonia is characteristic N-H bending mode, indicative of the presence of ammonia. 34 The FTIR spectra of the HKUST-1 MMMs upon exposure to ammonia show that the Cu-carboxylate band is retained, which is consistent with the PXRD data confrming the retention of the HKUST-1 crystal structure. Furthermore, the HKUST-1 MMMs displayed an N-H bend at 1610 cm 1 indicative of ammonia binding. The presence of Cu-carboxylate and N-H modes shows that even upon ammonia adsorption HKUST-1 can be stable when confned in a PVDF MMM. Taken together, the microbreakthrough, PXRD, and FTIR data show that ammonia sorption in HKUST-1 powder induces rapid degradation, but that incorporation of HKUST-1 into a MMM stabilizes the MOF, without loss of ammonia sorption capacity. Typically, engineered forms of sorbent materials have decreased activity toward an analyte of interest due to the blocking of pores and/or active sites. However, that is not observed here, likely due to the vast majority of active sites being located within the micropores of the MOF, instead of on the outer surface as is typically the case with metal/metal oxide nanoparticles. Even though many of the outer MOF pores may be in contact with the polymer binder, many of these pores must still be accessible such that ammonia can diffuse into the MOF crystallites, as evidenced by the high capacities observed for HKUST-1 MMMs. Furthermore, based on the sorption capacity observed here, it is improbable that the PVDF polymer penetrates into the inner pores of the MOF, impeding the diffusion of adsorbates within the MOF. ## Effect of water on HKUST-1 MMMs It was previously found that HKUST-1 in its powder form degrades upon exposure to humid conditions over the course of weeks. 31 The most aggressive aging condition was found to be 90% RH at 25 C, corresponding to an absolute humidity of 20.5 g m 3 , where HKUST-1 has a water uptake of 32 mol kg 1 or 38 wt%. We used these same conditions in this study to examine the moisture stability of the HKUST-1 MMMs. The ammonia loading for each HKUST-1 MMM, compared to the HKUST-1 powder, 31 after aging for various times is shown in Fig. 2 (as determined from the microbreakthrough experiments in Fig. S6-S8 †). For the HKUST-1 powder, $90% of the ammonia capacity is lost in the frst 7 days, without much further change over the full 28 days of the experiment. In contrast, for the 50-HKUST-1 and 67-HKUST-1 MMMs the ammonia capacity over the full 28 days of aging varies less than 20%. 30-HKUST-1 MMM shows $20% loss in ammonia capacity after aging for 14 days, and $50% after aging for 28 days. Although 30-HKUST-1 MMM loses more ammonia capacity than higher loading MMMs, the relative loss in capacity of 30-HKUST-1 is still substantially less than the HKUST-1 powder. Upon examination of the PXRD patterns (Fig. 4) of the HKUST-1 MMMs, it becomes more evident that the structures of 50-HKUST-1 and 67-HKUST-1 MMMs withstand the harsh aging conditions that HKUST-1 powder cannot. In general, the PXRD pattern remains unchanged over the 28 days studied for these materials. Furthermore, as expected from the loss in ammonia capacity of the 30-HKUST-1 MMM as it is aged, a degradation in the HKUST-1 crystal structure is observed. Interestingly, there is no formation of the secondary crystal structure in 30-HKUST-1 MMM that is seen upon degradation of the HKUST-1 powder, evidenced by reflections at 2q z 7.9, 9.2, 12.1, and 14.3 . It should also be noted that the HKUST-1 PXRD reflections are observed in 30-HKUST-1 MMM after 28 days of aging, but they are decreased in intensity. In previous work, we identifed the breakdown of the HKUST-1 structure by water to proceed mechanistically through the breaking of the Cu-carboxylate bond to form a carboxylic acid, seen through FTIR bands at 1708 and 1243 cm 1 . 31 Over the course of 28 days we did not observe the appearance of these FTIR bands for any of the MMMs (Fig. S9-S12 †). Interestingly, FTIR bands at 1620 and 1540 cm 1 , which represent physisorbed water, were observed for 30-HKUST-1 MMM, which was also the only MMM that showed any signs of degradation. These same bands were observed in the powder form of HKUST-1 at the early stages of MOF degradation (days 1-3), with a modest decrease in the ammonia uptake, but no noticeable change in the crystal structure was observed. Conversely, these FTIR bands are not observed in 30-HKUST-1 MMM until day 14, nor are they observed in the 50-HKUST-1 and 67-HKUST-1 MMMs over the 28 day study. More importantly none of the HKUST-1 MMMs show the appearance of the carboxylic acid modes typically observed upon breakdown of HKUST-1. Physically no change was observed in the color or appearance of the 50-HKUST-1 and 67-HKUST-1 MMMs during the aging process; however, 30-HKUST-1 MMM showed a fading of the characteristic blue color from what appeared to be visible degradation of the material (Fig. S13 †). The preparation of the MMMs results in the side against the substrate having greater polymer content (referred to as substrate-facing side), while the other side is richer in MOF crystals dispersed throughout the PVDF (referred to as outward-facing side), which is more representative of the bulk MMM. SEM images (Fig. 3 and S14 †) further show the physical state of the 50-HKUST-1 and 67-HKUST-1 MMMs remain essentially unchanged over the 28 days of aging. The different sides of the MMM have varying degrees of hydrophobicity, as do the MMMs with various amounts of HKUST-1, as determined by contact angle measurements (Table 2). Two distinctive trends were observed: (1) the contact angle increases as the HKUST-1 content of the MMM increases, and (2) the outward-facing side of the MMM has a higher contact angle than the substrate-facing side of the MMM. The water contact angle of a material is driven by not only the chemical make-up of the material, but also by the surface roughness. 35,36 The air trapped in the space between the MOF particles is an important contributor to the increased hydrophobicity as the water contact angle of air is considered to be 180 . 36 We propose that the increased surface roughness from the individual MOF crystals is the primary contributor to the increased contact angle of the outward facing-side (more MOF-like) of the MMMs verse the substrate-facing side (more PVDF-like). As expected, due to increased surface roughness, the HKUST-1 MMMs with a higher percentage of HKUST-1 exhibit an increased contact angle. Of particular signifcance, the water contact angle of an HKUST-1 pellet has been reported to be 59 , 33 much lower than that observed for the any of the MMMs. The increased hydrophobicity of the MMMs compared to HKUST-1 powder causes a signifcant decrease in the total water uptake as observed in the water isotherms performed at 25 C (Fig. S15 and S16 †). When corrected for MOF content, the water loading is similar for each of the MMMs. It has been shown elsewhere that hydrolysis of HKUST-1 requires the clustering of water molecules near the Cu-carboxylate bonds, in order for hydrolysis to occur. 30 The decreased water uptake of the MMMs causes there to be less water per SBU, signifcantly decreasing the potential for degradation of the MOF via hydrolysis. The PXRD and FTIR data, along with the ammonia capacities of the HKUST-1 MMMs, clearly show that the 50-HKUST-1 and 67-HKUST-1 are quite stable to humid environments over the period studied here. As seen previously, the presence of hydrophobic fluoroalkanes enhance the water stability of HKUST-1. 32,33 However, in humid environments 30-HKUST-1 begins to lose its crystallinity and consequently its ammonia adsorption capacity over time, due to the early stages of MOF degradation, which can be observed through the sorption of water in the FTIR spectra, and loss of the HKUST-1 crystal structure in the PXRD spectra. 30-HKUST-1 MMM contains the highest ratio of the PVDF polymer, which in the case of the MMM acts as a binder and supplies hydrophobic CF 2 groups. Intuitively, one might hypothesize that the 30-HKUST-1 MMM would withstand the humid environment that the materials were subject to better than the MMMs with a higher HKUST-1 content; however, this was not the case. It was observed that the 30-HKUST-1 MMM sample had much lower contact angles (similar to that of PVDF) than the other MMMs, likely due to the increased surface roughness of the 50-HKUST-1 and 67-HKUST-1 MMMs. In the SEM images of the higher HKUST-1 content MMMs, it can be observed that the MOF crystals dominate the surface of the MMM, which is not the case for 30-HKUST-1 MMM. In turn, the flattened surface of 30-HKUST-1 MMM does not repel water droplets as well as the other MMMs, and over the course of 28 days may attract more water into the MOF structure, promoting eventual degradation of the MOF, even though this was not observed on the timescale of the water isotherm. Nevertheless, the degree of degradation for the MMM is much less than is observed for the pure HKUST-1 powder. In fact, after 7, 14, and 28 days of aging, the 30-HKUST-1 MMM even has a higher ammonia capacity than the HKUST-1 powder, despite 30-HKUST-1 MMM only having 30% of the material being the active HKUST-1 MOF. ## Conclusions HKUST-1 MMMs are strong candidates for use in gas fltration applications, due not only to the high ammonia capacities of the engineered form of the MOF, but due to the increased water and ammonia stability of the MOF over the primitive powder form. Furthermore, the MMMs can easily be shaped or molded into useful forms for a given application. PVDF acts as an effective binder for HKUST-1, while still allowing the pores to be permeable and accessible to adsorbates such as ammonia, giving ammonia capacities that are scalable to the HKUST-1 content of the MMM. Remarkably, no degradation was observed over the course of 28 days for the 50-HKUST-1 and 67-HKUST-1 MMMs upon exposure to 90% RH at 25 C. Furthermore, the ammonia capacity for these samples was relatively constant over the period studied. The increased water stability of HKUST-1 in the MMM is due to the increased chemical hydrophobicity stemming from the fluorocarbon groups of the PVDF polymer and the surface roughness created by the MOF crystals bound in the matrix, making these materials resistant towards both water vapor and liquid water. Overall the fndings here strongly support the use of MOF MMMs as a means to enhance the utility, processability, stability, and performance of MOFs in gas sorption applications in a comprehensive manner that has not been demonstrated before.

chemsum

{"title": "Enhanced aging properties of HKUST-1 in hydrophobic mixed-matrix membranes for ammonia adsorption", "journal": "Royal Society of Chemistry (RSC)"}

selective_high-resolution_dnp-enhanced_nmr_of_biomolecular_binding_sites

## Abstract: Locating binding sites in biomolecular assemblies and solving their structures are of the utmost importance to unravel functional aspects of the system and provide experimental data that can be used for structurebased drug design. This often still remains a challenge, both in terms of selectivity and sensitivity for X-ray crystallography, cryo-electron microscopy and NMR. In this work, we introduce a novel method called Selective Dynamic Nuclear Polarization (Sel-DNP) that allows selective highlighting and identification of residues present in the binding site. This powerful site-directed approach relies on the use of localized paramagnetic relaxation enhancement induced by a ligand-functionalized paramagnetic construct combined with difference spectroscopy to recover high-resolution and high-sensitivity information from binding sites. The identification of residues involved in the binding is performed using spectral fingerprints obtained from a set of high-resolution multidimensional spectra with varying selectivities.The methodology is demonstrated on the galactophilic lectin LecA, for which we report well-resolved DNP-enhanced spectra with linewidths between 0.5 and 1 ppm, which enable the de novo assignment of the binding interface residues, without using previous knowledge of the binding site location. Since this approach produces clean and resolved difference spectra containing a limited number of residues, resonance assignment can be performed without any limitation with respect to the size of the biomolecular system and only requires the production of one protein sample (e.g. 13 C, 15 N-labeled protein).D-galactose according to the crystal structure. Table of assigned residues of LecA based on Sel-DNP spectra. Chemical analysis of compounds 1, 2, 4, and 5. See ## Introduction Addressing site-specifc atomic scale information of biomolecular binding sites is of primary importance. 1 The main experimental methods have relied so far on the use of X-ray crystallography, cryo-electron microscopy (cryo-EM) and nuclear magnetic resonance (NMR). Beyond solving structures, NMR is also uniquely suited to perform dynamics studies at the atomic scale, does not require crystals of the bound-ligand complex and is compatible with measurements performed under physiological conditions, including membrane or cellular milieu. In addition, solid-state NMR, possibly combined with a hyperpolarization technique called Dynamic Nuclear Polarization (DNP), 2 has recently emerged as a key technique to extract precise structural details of the ligand (or drug) in functional biomolecular complexes (including membrane and fbrillar proteins). 3 This approach relies on the use of isotopic labeling and allows the ligand structure to be solved without requiring prior knowledge of the host biomolecular architecture. Even if such information does not give a direct insight into binding sites, it can be efficiently combined with computational docking studies. 10 Nevertheless there is still a lack of NMR methodologies capable of locating binding sites in biomolecular systems and solving their structures at the atomic scale. Indeed, access to this type of information requires the use of advanced nonuniform isotopic spin labeling strategies in order to label residues involved in the binding site while reducing spectral crowding. 11,12 Although possible, this approach suffers from strong limitations, since it is restricted to systems that can be over-expressed accordingly, often relies on previous knowledge of the binding site location and is limited to biomolecular assemblies composed of monomeric units of 50 kDa or less. 13 The present contribution provides a solution to this problem. We report a new approach that allows identifcation of hyperpolarized binding sites in biomolecular complexes using DNPenhanced NMR. This method, abbreviated as Sel-DNP (Selective Dynamic Nuclear Polarization), relies on the use of a ligand functionalized polarizing agent (PA), and allows extraction of site-specifc information from the interaction site without prior a Univ. Grenoble Alpes, CEA, CNRS, INAC-MEM, Grenoble, France. E-mail: sabine. hediger@cea.fr; gael.depaepe@cea.fr b Univ. Grenoble Alpes, CNRS, DCM, Grenoble, France c Univ. Grenoble Alpes, CNRS, CERMAV, Grenoble, France † Electronic supplementary information (ESI) available: Aromatic region of Sel-DNP spectra of LecA for k ¼ 1, 0.8, and 0.7. Sel-DNP spectra of LecA for k ¼ 0.6 and 0.4. Conformations of the PA-ligand in the LecA binding site. DQ/SQ dipolar recoupling sequence. Table of residues involved in the binding of knowledge of its location. The application of Sel-DNP is demonstrated on the galactose-specifc lectin LecA 14 for which highly resolved multi-dimensional experiments are obtained that allow the determination of residues involved in the binding site. The approach does not have any limitations with respect to the size of the biomolecular system, and only requires the preparation of one protein sample. ## Results DNP-enhanced NMR under Magic-Angle-Spinning (MAS-DNP) is a recently developed hyperpolarization technique which has been demonstrated on a large range of systems with new applications in surface chemistry, materials science and structural biology. The increase in NMR sensitivity is achieved by transferring the polarization from unpaired electrons, contained in dopant molecules called Polarizing Agents (PAs), to the surrounding nuclear spins of the system of interest. PAs are typically composed of stable organic mono-or bi-radicals. The transfer of polarization is driven by high-power high-frequency microwave irradiation of the sample, typically generated by a gyrotron. 18 PAs are usually dissolved in a glass-forming matrix that is used to suspend or impregnate the sample of interest. 15,16,19 Where appropriate, further removal of the solvent can also be performed such as in the Matrix Free or Film Casting sample preparation approaches, as long as radical aggregation is avoided. Recently several groups have discussed the use of a localized PA, either covalently linked to the system of interest or to a bound-ligand, or just using a PA with high-affinity for the system of interest. 21, The motivations behind these contributions were different from those for the work presented here, ranging from optimizing sample sensitivity through PA co-localization, 21, along with a reduction of the cryo-protectant content, 23,24 to the development of targeted DNP experiments in complex environments 27 or the experimental investigation of the bleaching effect using specifc isotopic labelled samples. 28,29 Fig. 1 (a) Schematic of the functionalized ligand used in Sel-DNP and composed of a ligand tethered to a paramagnetic tag (PT) via a short linker. In this work, the ligand corresponds to D-galactose, the linker is made of a phenylglycine unit, and the paramagnetic tag chosen is the bisnitroxide TOTAPOL. 30 (b) Structure of the binding site of LecA highlighting the residues known (from the crystal structure) to interact with the galactose ligand (in cyan). (c) and (d) DNP-enhanced 13 C- 13 C DQ/SQ one-bond correlation spectra of LecA, using (c) AMUPol (reference spectrum S 0 ) and (d) the functionalized ligand of (a) (spectrum S). Positive contours are in black and negative ones are in red. Contours are set to the same level throughout the entire manuscript. The Sel-DNP approach: introducing selectivity in DNP NMR The approach reported here and dubbed Selective DNP (Sel-DNP) combines for the frst time the use of localized Paramagnetic Relaxation Enhancement (PRE) and difference spectroscopy to recover high-resolution and high-sensitivity information from hyperpolarized binding sites. It is well known that the presence of paramagnetic centers in MAS-DNP samples induces signal broadening for nearby nuclear spins (so-called bleaching/quenching effect). Consequently, the introduction of a functionalized ligand, composed of the ligand moiety, a linker, and a paramagnetic species (see Fig. 1a), is expected to induce a localized PRE effect. This is not the case when paramagnets (used either to increase relaxation or as a PA) are uniformly distributed in the sample of interest. Sel-DNP requires the acquisition of two datasets of a given ssNMR experiment recorded for two different sample preparations of the protein. The frst spectrum, called reference (S 0 ), is obtained for the biomolecular bound-ligand complex, for which the PA is homogeneously distributed in the sample. The corresponding NMR signals are uniformly enhanced by DNP, including the binding region. The second spectrum (S) is obtained with the same pulse sequence for the same protein, but using a specifc ligand tethered to a paramagnetic tag (PT). The proximity of the paramagnetic moiety to the binding region in the receptor-ligand-PT complex induces a broadening and attenuation of the closest resonances. Upon microwave irradiation, 1 H spins from the protein are polarized by DNP, thanks to spin-diffusion of the hyperpolarized magnetization from the bleached zone (resonances broadened and attenuated beyond detection) to the distant protons. 31,32 As demonstrated in this work, residues involved in the binding region can be revealed by the subtraction of the two spectra. ## Application to carbohydrate-binding proteins The approach was tested on lectins, which are carbohydratebinding proteins that act as molecular readers to decipher information coded in glycoconjugates on the cell surface. They therefore play important biological roles in recognition processes relevant for human health. Lectins from pathogens are involved in host recognition of human tissues 33 and are targets for the development of anti-infectious compounds that could provide alternatives to antibiotics. The lectin LecA from the opportunistic pathogen Pseudomonas aeruginosa consists of 121 amino acids (MW 12.75 kDa), is specifc for D-galactose-containing molecules, and forms a tetramer in solution. 14 The X-ray structure of the complex between LecA and galactose is presented in Fig. 1b, and unveils details of hydrogen bonds and Ca 2+ bridge interactions. 14 Note that no NMR study has been reported so far, and thus chemical shift assignment is unknown for this protein. Two DNP samples of LecA were prepared as described in detail in the Experimental section. The frst sample, used for the reference spectrum S 0 , was prepared with one equivalent of Dgalactose and impregnated with an AMUPol-biradical 34 containing solution. For the second sample (used for spectrum S), a specifc ligand-PT was synthesized (see Experimental section), composed of D-galactose tethered to the well-known polarizing agent TOTAPOL 30 via a short phenylglycine linker (Fig. 1a). One equivalent of this ligand-PT was added as the PA to the protein solution prior to evaporation of the excess solvent. For each sample, a 13 C-13 C Double-Quantum (DQ) dipolar correlation spectrum was acquired with the mixing time of the recoupling sequence set to produce only one-bond correlations (see Experimental section for more details). 35 Both spectra, recorded at 9.4 T and 100 K sample temperature, look very similar and illustrate the typical limited resolution obtained with MAS-DNP NMR experiments performed on biomolecular systems (Fig. 1c and d). Overall, linewidths in our experiment S appear broader compared to the ones in S 0 . This results in part from a higher overall concentration in ligand-PT (compared to the AMUPol concentration used in the reference sample), and from the fact that the paramagnetic species are not distributed the same way in both samples. ## Tuning the selectivity focus using Sel-DNP The Sel-DNP spectra can be obtained by combining the two datasets S 0 and S, using the following expression: ## S k Sel-DNP ¼ S 0 kS By tuning the relative weight of both spectra through the selectivity factor k, we can progressively adjust the focus on the binding region, selecting resonances as a function of the distance between the corresponding site and the ligand-PT. As S and S 0 are obtained from different DNP samples, their respective spectral intensity is arbitrary. Prior to Sel-DNP subtraction, the spectra are relatively adjusted such that the difference using k ¼ 1 gives rise to the least number of positive Sel-DNP peaks (highest focus). This arbitrary pre-scaling is then kept constant for all other Sel-DNP spectra produced with lower k. After subtraction, only positive peaks are kept in the Sel-DNP spectra. This Sel-DNP procedure is illustrated in Fig. 2. It is important to note that the residual positive NMR signal itself originates from the protein with its native ligand (S 0 ). Locating the binding site using de novo resonance assignment Following the subtraction procedure described in the previous section, several binding-site selective spectra are produced with decreasing selectivity factor, k ¼ 1.0, 0.8, 0.7, 0.6 and 0.4 (Fig. 3a-c, e-g, S1a-c, and S2 †). They all show excellent resolution arising from the limited number of residues selected in each Sel-DNP spectrum, with linewidths at half height of 0.5 to 1 ppm. This contrasts severely with the very limited resolution obtained in the S 0 spectrum (using uniformly distributed AMUPol), which contains all the 121 residues of the protein but cannot be used to perform resonance assignment. In contrast, each Sel-DNP spectrum can be used to identify and connect carbon resonances (from carbonyl to sidechain carbons) within a given residue. These 13 C NMR fngerprints can be compared to tabulated values to identify and assign the residues involved in the binding site. The Sel-DNP spectrum obtained with k ¼ 1 contains crosspeaks corresponding to $6 residues labeled in black in Fig. 3a and e, and S1a. † For two of them, an almost complete carbon fngerprint is observed, which easily allows these residues to be assigned to a proline and a histidine. This is a very positive result since it is known from the crystal structure that 51 Pro and 50 His are involved in the binding of galactose in LecA (Fig. 1b). 14 A second set of residues with only a partial carbon fngerprint in the k ¼ 1 Sel-DNP spectrum is observed. With the help of Sel-DNP spectra of lower k, in which the missing resonances appear, they could be assigned to a tyrosine, a threonine, a glutamic acid, and an aspartic acid, respectively. Based on the amino acid sequence of the protein, and assuming that these residues belong to 2 or 3 groups of spatially close residues involved in the binding site, it is straightforward to identify 36 Tyr (involved as well in the binding of galactose according to the crystal structure), 39 Thr, 49 Glu, and 52 Asp, in addition to 51 Pro and 50 His. With k ¼ 0.8, further cross-peaks of the 6 residues appear clearly, as well as new resonances corresponding to residues color-coded in blue in Fig. 3b and f, and S1b. † Following the protocol described above, these carbon fngerprints can be identifed and assigned to residues that are either located at neighboring positions to the black residues, or belong to a third binding region, as highlighted in Fig. 4c. The same analysis can be conducted for k ¼ 0.7, revealing new residues that are color-coded in purple in Fig. 3c and g, and S1c, † and for k ¼ 0.6 and 0.4, whose new residues are colorcoded in burgundy and orange in Fig. S2a-c and S2d-f, † respectively. Gradually, by decreasing the selectivity factor k, the full binding site (see Fig. 1b and Table S1 †) is highlighted with the appearance of 53 Gln, 100 Asp, 101 Val, 104 Thr, 105 Tyr, 106 Gly, 107 Asn, and 108 Asn, as well as other residues located outside the binding pocket (e.g. 34 Ala, 60 Ala, and 55 Leu). The list of assigned resonances is given in ESI Table S2. † Note that we were not able to use Sel-DNP spectra with k < 0.4 because of spectral crowding which prevents us from readily extracting site-specifc information. ## Discussion Relating selectivity factor and distance to the paramagnetic tag The X-ray structure of the LecA/Gal complex (PDB ID 1OKO) 14 was then used to check whether residues assigned using the Sel-DNP methodology are indeed located in or close to the binding region. Fig. 4a and b show the crystal structure of the proteinligand complex with the assigned residues color-coded in black, blue, purple, burgundy and orange, depending on in which Sel-DNP spectrum (with k ¼ 1, 0.8, 0.7, 0.6 or 0.4, respectively) their NMR resonances appear. This confrms that all residues assigned with Sel-DNP are in the proximity of the ligandbinding site. This is also clearly seen in Fig. 4c which uses the same color code in combination with the sequence of the protein, in which the residues identifed in the crystal structure to be in direct interaction with the ligand or the calcium ion are highlighted in bold (see Table S1 †). Note that the residues from the binding site that are located behind the calcium atom (residue numbers 100 to 108) appear in Sel-DNP spectra with k ¼ 0.7-0.4, and not with k ¼ 1-0.8, since they are further away from the ligand (and thus from the PT) compared to residues (and their direct neighbors) that have either a direct interaction with the ligand, or interactions with both the calcium and the ligand. We then tried to relate the selectivity factor k to a distance cutoff from the center of the paramagnetic tag. The position of the latter was estimated by averaging three possible conformations of the functionalized ligand inside the binding pocket (as represented in Fig. S3 †). Using the X-ray structure of the binding site, it is possible to then relate the changes in the selectivity factor k to an approximate distance cut-off from the PT. For instance, k ¼ 1, 0.8 and 0.6 correspond to distance cutoffs of $10, $17 and $25 from the PT, respectively. This translates into a gradual increase in the number of residues appearing in the Sel-DNP spectra, from 6 residues for k ¼ 1 to about 33 residues for k ¼ 0.4 (Fig. 4d). The Sel-DNP spectra obtained with k < 0.4 and corresponding to a distance cut-off larger than 30 do contain a number of residues, which precludes further assignment under our experimental conditions due to spectral crowding. ## Paramagnetic relaxation enhancement of NMR resonances at low temperature Although PRE has been well described for solution and solidstate NMR experiments at ambient temperature, 36,37 it has not yet been thoroughly investigated under the specifc experimental conditions of DNP (e.g. 100 K, MAS, nitroxide-based PA). As demonstrated here, the presence of paramagnetic centers affects the signal of nuclear spins in two ways: frst a broadening of the nearby resonances (i.e. increased signal dephasing during detection, T * 2 ), and second an attenuation of the detected peak intensities, resulting from the paramagnetic relaxation occurring during each block of pulses and delays present before the detection period (i.e. CP step, dipolar recoupling period, etc.). This can be seen from the 1D traces through the doublequantum frequency of spectra S and S 0 shown in Fig. 5. In depth quantitative study of the PRE effect at low temperature is beyond the scope of this article but will be investigated in the near future. Nevertheless, we have demonstrated with this work that the presence of localized paramagnetic centers as found for ligand-PT induces an intensity gradient of the detected signal around the position of the electron spin, as shown in Fig. 2. According to our DQ-SQ 2D data, this effect is noticeable up to at least 30 from the paramagnetic center, in agreement with DNP bleaching radii reported by other groups. 28,29 When the PA is uniformly distributed in the sample, partial broadening and signal attenuation statistically affect all resonances. ## Experimental section Expression and purifcation of the LecA protein E. coli BL21 STAR (DE3) carrying the LecA plasmid was progressively adapted in four stages over 48 h to a minimal M9/ D 2 O medium (5.3 g L 1 Na 2 HPO 4 , 3 g L 1 KH 2 PO 4 , 0.5 g L 1 NaCl, 1 g L 1 15 NH 4 Cl, 5 mg thiamine, 1 mM MgSO 4 , 0.1 mM CaCl 2 , 50 mM ZnSO 4 , 100 mM FeCl 3 and a vitamin cocktail) with 100 mg L 1 ampicillin and 2 g L 1 U-[ 13 C, 2 H]-D-glucose. Eight 500 mL cultures in minimal M9/D 2 O medium were inoculated from the previously described preculture and grown at 37 C. When the OD 600 reached 0.75, protein expression was induced by the addition of 0.25 mM IPTG and the culture was incubated at 20 C overnight. Cells were harvested by centrifugation (6000g, 15 min) and the pellet was re-suspended in 80 mL of buffer A (20 mM Tris/HCl, 100 mM NaCl, pH 7.5, 100 mM CaCl 2 ). Purifcation proceeded through affinity to D-galactose resin (Sigma) with elution with buffer B (20 mM Tris/HCl, 100 mM NaCl, pH 7.5, 100 mM EDTA). After 5 days' dialysis in MilliQ H 2 O supplemented with 100 mM CaCl 2 , and 3 days' dialysis in H 2 O alone, the protein was lyophilized. A fnal yield of 238 mg for 4 L of culture was obtained. The triple-labelled protein samples were characterized by 15% SDS gel electrophoresis. Activity was checked by titration microcalorimetry and the triple-labelled protein displayed the same affinity for galactose as the non-labelled one. ## Chemical synthesis of the ligand-PT 5 (Scheme 1) The glyco-amino acid 1 was obtained through solid-phase peptide synthesis, following the Fmoc/ t Bu strategy and using a 2-chlorotrityl resin (m ¼ 1.000 g, loading ¼ 1.04 mmol g 1 ). Coupling reactions were performed using, relative to the resin loading, 2 eq. of ethylenediamine, N-Fmoc-protected L-glycine and 1-(4-carboxyphenyl)-2,3,4,6-tetra-O-acetyl-b-D-galactopyranoside activated in situ with PyBOP (2 eq.) and DIPEA (4 eq.) in Fig. 5 1D-traces through the double-quantum frequency at 85 ppm corresponding to the 50 His Ca-Cb cross peaks from the original data S 0 and S (see Fig. 1c and d), as well as from the Sel-DNP spectra with k ¼ 1, 0.8, 0.7, and 0.6 (see Fig. 3a-c Acetic acid was added to a solution of 4-oxo-TEMPO (50 mg, 0.29 mmol, 1 eq.) and glyco-amino acid 1 (1.1 eq.) in dry THF (10 mL) in order to reach pH ¼ 4. The mixture was stirred at room temperature for 4 h. Then sodium cyanoborohydride (93 mg, 0.44 mmol, 1.5 eq.) was added and the mixture was stirred overnight. After concentration under reduced pressure, the crude mixture was purifed by semi-preparative RP-HPLC (C18, l ¼ 214 nm, 5-40% B in 15 min) yielding the glyco-amino acid 2. A solution of 2 (0.1 mmol, 1 eq.), epoxide 3 (25 mg, 0.11 mmol, 1.1 eq.), and lithium perchlorate (106 mg, 1 mmol, 10 eq.) in dry acetonitrile (10 mL) was stirred at reflux for 72 h. Additional lithium perchlorate (10 eq.) was added after 24 h and 48 h. After concentration under reduced pressure, the crude mixture was purifed by semi-preparative RP-HPLC (C18, l ¼ 214 nm, 5-40% B in 15 min) yielding compound 4 (1 : 1 diastereomeric mixture). A solution of tetra-O-acetylated compound 4 (0.020 mmol, 1 eq.) in a mixture of methanol, triethylamine and water (5 : 1 : 2) (8 mL) was stirred at room temperature overnight. After concentration under reduced pressure, the crude mixture was purifed by semi-preparative RP-HPLC (C18, l ¼ 214 nm, 5-40% B in 15 min), obtaining the ligand-PT 5. The chemical analysis of 1, 2, 4 and 5 is given in the ESI (Fig. S5 to S18 †). ## Sample preparation for Sel-DNP experiments Two samples (of the same protein) were prepared in order to record the Sel-DNP spectra. For the frst sample, a solution of 1.5 mM 2 H, 13 C, 15 N-uniformly labeled LecA was prepared in a protonated buffer containing 50 mM ammonium acetate (pH ¼ 4.7), 1 mM calcium chloride, 1 mM magnesium chloride and 1.6 mM D-galactose. This solution was incubated for 24 h at 5 C, followed by evaporation of the excess solvent using a speedvac until a slightly wet solid was obtained. The resulting sample was impregnated with a solution of 60% 2 H-glycerol, 30% D 2 O and 10% H 2 O (v/v) containing 10 mM AMUPol. 34 Approximately 5 mg was packed into a 3.2 mm sapphire NMR rotor using a silicone plug and it was closed with a zirconia cap. The fnal concentration of AMUPol in the sample is estimated to be about one molecule for fve LecA units. This sample was used to record the reference spectrum S 0 (Fig. 1c). For the second sample, a solution of 1.5 mM 2 H, 13 C, 15 Nuniformly labeled LecA was prepared in a protonated buffer containing 50 mM ammonium acetate (pH ¼ 4.7), 1 mM calcium chloride, 1 mM magnesium chloride and 1.6 mM of the ligand-PT compound 5 (Fig. 1a). This solution was incubated for 24 h at 5 C, followed by evaporation of the excess solvent with a speed-vac. Approximately 14 mg of the resulting sample was packed into a 3.2 mm sapphire NMR rotor using a silicone plug and it was closed with a zirconia cap. This sample was used to record the spectrum S (Fig. 1d). ## DNP experiments All the experiments reported here were carried out on a Bruker 263 GHz DNP-NMR Avance III spectrometer (400 MHz 1 H frequency) equipped with a low temperature 3.2 mm wide-bore MAS probe. Experiments were conducted at $100 K sample temperature and 9.5 kHz MAS frequency, to prevent both rotational resonance effects 38 and spinning sidebands from overlapping with the NMR peaks. The ratio of the 1 H- 13 C CP MAS signals with and without microwave irradiation, 3 on/off , was evaluated for the two DNP samples, returning 70 and 43 for the AMUPol and the ligand-PT samples, respectively. The DNP buildup times (T B ), measured using a saturation-recovery experiment, were 5.3 s and 0.7 s for the AMUPol and ligand-PT samples, respectively. 13 C-13 C DQ-SQ dipolar correlation experiments were performed for each sample using the pulse sequence given in ESI Fig. S4. † The hyperpolarized 1 H magnetization is frst transferred to carbons using a 1 H- 13 C cross polarization (CP) step. The dipolar recoupling symmetry-based sequence R16 7 2 is then applied to generate double-quantum (DQ) coherences. 35 The excitation period was set to s DQ ¼ 0.4 ms to maximize one-bond polarization transfer. After free evolution during t 1 (indirect dimension) of the generated 13 C- 13 C DQ coherences (at frequencies corresponding to the sum of the chemical shifts of the coupled 13 C spins), a second block of R16 7 2 recoupling of the same duration s DQ is applied to reconvert DQ coherences into detectable single-quantum (SQ) coherences. No 1 H decoupling is required during the R16 7 2 recoupling periods. 39 SPINAL64 1 H decoupling at a feld strength of 100 kHz was applied during both indirect (t 1 ) and direct (t 2 ) detection times. Each two-dimensional spectrum was acquired with a maximum t 1 evolution time of 5.2 ms using states-TPPI for quadrature detection (256 complex points). For each increment, 16 transients were added with interscan delays of 4 s and 0.9 s for the AMUPol and the ligand-PT sample, respectively. Acquisition time in the direct t 2 dimension was 10 ms. Spectral widths were 492 ppm in the DQ dimension and 405 ppm in the SQ dimension. Cosine apodization was applied in both dimensions prior to Fourier transformation, followed by baseline correction with a polynomial function of grade 5. ## Sel-DNP spectra Sel-DNP spectra with k ¼ 1, 0.8, 0.7, and 0.6 have been used to address the NMR fngerprint of the LecA binding site. To avoid baseline artefacts, both experiments S and S 0 need to be acquired with similarly good signal-to-noise ratios. The assignment of the Sel-DNP spectra for k ¼ 1, 0.8, and 0.7 was possible without pre-knowledge of the crystal structure: the type of amino acid was identifed using chemical shift statistics for amino acid residues published by the Biological Magnetic Resonance Data Bank (BMRB) 40 and their position in the sequence was obtained considering that all residues have to be located close to each other in only a few regions of the sequence (see Fig. 4c). For k ¼ 0.6 and 0.4, calculations based on the crystal structure using the software SHIFTX2 (ref. 41) were used as well, in particular for ambiguous residues. ## Molecular modeling A three-dimensional model of glycoconjugate 5 was built with Sybyl (Certara, Princeton, NJ) graphical software using the atomic coordinates of TOTAPOL proposed from the PubMed site (https:// pubchem.ncbi.nlm.nih.gov). The molecule was graphically extended by adding a glycine, a benzene and a galactose residue from the Sybyl molecular libraries. The galactose was docked on the calcium ion in two adjacent binding sites of the LecA structure by superimposition on an equivalent sugar in the crystal structure with PDB ID 1OKO. 14 In order to illustrate the extension that the TOTAPOL group can attain on the surface of LecA, three conformations were generated by varying the torsion angles in the phenylglycine linker (Fig. S3 †). ## Conclusions The Sel-DNP approach enables the extraction of site-specifc information from native biomolecular binding sites. Since only a limited number of resonances are sequentially highlighted, this approach yields highly resolved multi-dimensional spectra that allow resonance assignment and identifcation of residues in close spatial proximity to the paramagnetic tag, and therefore location of the binding site. Sel-DNP does not have any intrinsic limitation in the size of the biomolecular system that can be studied, since only residues close to the ligand-PT are detected. The approach only requires one protein sample, either uniformly isotopically 13 C, 15 N labeled or not, but can of course be combined with advanced site-specifc labeling strategies. In addition, the protein-ligand complex does not need to be crystallized. It is important to stress that this approach can be implemented without any previous knowledge of the binding site location. Nevertheless, resonance assignment is greatly simplifed if a global fold or a structure of the protein alone is available. Extension to 15 N- 13 C correlation experiments using Sel-DNP is currently being implemented in our laboratory, as well as correlation experiments that can provide long-range distance contacts. This should help in performing sequential resonance assignment and extracting specifc structural constraints from the binding site. We expect Sel-DNP to become an important method for the study of protein binding sites, e.g. membrane proteins (possibly in their cellular environment), fbrillar assemblies, etc. Multidimensional difference spectroscopy and DNP, which was recently used to improve the spectral resolution of the various cell-wall components of bacteria, 42 are likely to become important tools for the study of complex systems. Finally, it is worth noting that on-going instrument development has recently demonstrated the feasibility of improving the sensitivity by several orders of magnitude compared to the current state of the art. This approach which relies on the use of sustainable cryogenic helium spinning at much lower temperatures, 43 should allow Sel-DNP to be used for the investigation of interaction interfaces in very large biologically relevant complexes. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Selective high-resolution DNP-enhanced NMR of biomolecular binding sites", "journal": "Royal Society of Chemistry (RSC)"}

solution_structure_of_a_europium-nicotianamine_complex_supports_that_phytosiderophores_bind_lanthani

## Abstract: We report the solution structure of a europium-nicotianamine complex predicted from ab initio molecular dynamics simulations with density functional theory. Emission and excitation spectroscopy measurements show that the Eu 3+ coordination environment changes in the presence of nicotianamine, suggesting complex formation, and strongly supporting the predicted Eu 3+ -nicotianamine complex structure from computation. We used our recently optimized pseudopotentials and basis sets for lanthanides to model Eu 3+ -ligand complexes with explicit water molecules in periodic boxes, effectively simulating the solution phase. Our simulations consider possible chemical events (e.g. coordination bond formation, protonation state changes, charge transfers), as well as ligand flexibility and solvent rearrangements. Our computational approach correctly predicts the solution structure of a Eu 3+ -ethylenediaminetetraacetic acid complex within 0.05 Å of experimentally measured values, backing the fidelity of the predicted solution structure of the Eu 3+ -nicotianamine complex. Emission and excitation spectroscopy measurements were also performed on the well-known Eu 3+ -ethylenediaminetetraacetic acid complex to validate our experimental methods. The electronic structure of the Eu 3+nicotianamine complex is analyzed to describe electron densities and coordination bonds in greater detail. Nicotianamine is a metabolic precursor of, and structurally very similar to, phytosiderophores, which are responsible for the uptake of metals in plants. Although knowledge that nicotianamine binds europium does not determine how plants uptake rare earths from the environment, it strongly supports that phytosiderophores bind lanthanides. ## I. Introduction The remarkable and unique characteristics of electronic states of lanthanide (Ln) complexes originating from partially filled 4f-electron shells, and their extremely localized nature, make studies of their compounds a very active area of research. Ln complexes are used in a multitude of high-tech applications. 11 At the same time, the role of lanthanides in naturally occurring biological systems was somewhat overlooked. Until recently, lanthanides were not considered as essential elements of biological systems and Ln-incorporated enzymes were viewed as useful, yet mainly artificial, systems. This perspective changed with the discovery of lanthanides in bacterial methanol dehydrogenases. Moreover, a number of studies show that the Ln elements, especially Ce, affect the growth and development of agriculturally important crops. 17,18 However, currently there is no certainty regarding Ln intake pathways and accumulation in plants. It is not clear whether accumulation happens through Ca or Fe uptake pathways involving the broad-spectrum metallophore nicotianamine (NA), or if Ln intake and accumulation are the result of production of lanthanophores by bacteria residing in the phyllosphere. 12,19 Low-molecular weight chelators with functional carboxy-, amino-and hydroxy-groups facilitating metal coordination in bacteria and plants are classified as siderophores and phytosiderophores, respectively. When it comes to understanding the role of these compounds in the metabolism of lanthanides in plants, NA, a metallophore naturally occurring in higher plants, 20,24 is of particular interest. NA is structurally very similar to phytosiderophores, and it is a precursor in phytosiderophores biosynthetic pathways, 20 which makes it an ideal model system. Apart from its importance for understanding the role of lanthanides in plant of metabolic transformations, elucidating the solution structure of Ln 3+ complexes with NA could potentially contribute to the development of novel ligands for Ln extraction. Phytosiderophores are polydentate and polyacidic, with a wide range of pKa sites, therefore solution pH changes the protonation state of each acidic site, impacting the Ln coordination structure. It is well known that changes in solution acidity facilitates Ln extraction, additionally, recent studies with europium, 25 gadolinium, 26 and terbium 27 demonstrate how solution pH affects Ln complex coordination structures and their surrounding environment. Moreover, previous studies show that tuning the flexibility of polydentate and polyacidic ligands is well suited for Ln separations, 28,29 highlighting the importance of ligand flexibility. Siderophores have been shown to bind Ln 3+ ions with a pH-sensitive binding behavior. 31,33 Like siderophores, the molecular structures of phytosiderophores are highly susceptible to changes in protonation states, 20, but, unlike siderophores, their ability to bind Ln 3+ ions has not been elucidated. Although lanthanides have been detected in plants and their roots, 37 and uptake mechanisms of other trivalent elements were previously reported, 38 apart from few hypotheses, the plant uptake mechanisms of Ln 3+ remains largely unknown. 19,39 . Interestingly, Liu et al. 40 demonstrated that Rubisco, 41 an enzyme crucial for photosynthetic CO2 fixation in higher plants, 40,42 binds cerium, which demonstrates Ln 3+ -binding capability of compounds native to plants. To date, there are no reported in vitro or in silico studies of Ln ions and phytosiderophores. Lanthanide coordination is affected by a number of factors in the coordination environment, such as flexibility and protonation state of the bound ligand. Due to the high coordination number of Ln 3+ ions, in addition to the ligand, water (or solvent) molecules will coordinate as well. Thus, the solution structure of Ln 3+ -ligand complexes and their stabilities strongly depend on the molecular structure of their coordination spheres, which include ligand and solvent molecules. 47 Resolving the solution structure of Ln 3+ -ligand complexes is a nontrivial task due to their highly dynamical nature in solution. To do so computationally in atomic resolution requires the use of density functional theory (DFT) and ab initio molecular dynamics (AIMD) simulations to simulate the breakage and formation of chemical bonds (e.g. coordination bonds, protonation state changes) simultaneously considering ligand flexibility and solvent rearrangements. Additionally, simulating solution structures (i.e. condensed phase) requires including explicit solvent molecules, and treating the system periodically in repeating unit cells. An accurate periodic treatment of solute-solvent systems requires periodic boxes of sufficient size, which makes such all-electron AIMD simulations computationally out of reach, thus requiring DFT. The employment of relativistic, norm-conserving, separable, dual-space Gaussian-type pseudopotential protocol of Goedecker, Teter, and Hutter (GTH) in a mixed Gaussian−plane wave scheme 51 has proved to be an effective and efficient way to perform AIMD simulations of larger systems (> 500 atoms) 52,53 and with longer trajectories (> 10 ps), 52,54 significantly reducing the computational cost of such simulations. Until recently, accurate GTH pseudopotentials and basis sets for the lanthanides, besides cerium, 55,56 were lacking. This prevented performing larger-scale DFT and AIMD simulations of lanthanide-containing systems in the condensed phase or solid state. Our previous work bridged this critical gap by producing LnPP1: a full set of well-benchmarked pseudopotentials along with the corresponding basis sets within GTH protocol specifically optimized for GGA PBE 57 calculations of lanthanide-containing systems. 58 Despite the fact that other types of lanthanide pseudopotentials (i.e. effective core potentials) and basis sets were previously reported, these were employed in electronic structure calculations with lanthanide systems containing less than ∼100 atoms. 66,67 Thus, simulating the solution structure of lanthanide-ligand complexes, with molecular dynamics sampling to explicitly include solvent rearrangement and ligand flexibility, was not possible until very recently. Our LnPP1 complete set of pseudopotentials and accompanying basis sets enabled us to perform DFT calculations and AIMD simulations of lanthanide-ligand complexes in explicit water boxes in periodic conditions (system > 500 atoms) in the present paper. This work pursued two goals: i) demonstrate that our pseudopotentials and basis sets (LnPP1) with AIMD simulations can predict the structures of lanthanide complexes in solution and, ii) elucidate the molecular and electronic structure of Eu 3+ -nicotianamine complexes in aqueous solution. Our previous work 58 demonstrated the accuracy of our LnPP1 pseudopotentials and basis sets to replicate Ln reactivity (e.g. oxidation reactions, heats of formation, ionization potentials). This work shows that our pseudopotentials and basis sets with AIMD simulations replicate the solution structure of lanthanide-ligand complexes, by correctly (within 0.05 ) predicting the structure of a complex whose structure is known (Eu 3+ -ethylenediaminetetraacetic acid [EDTA]). 68,69 Upon validation of the computational approach, we predict the structure of Eu 3+ -nicotianamine complexes in solution: this work describes their molecular structures (e.g. ligand conformation, water molecule inclusion, coordination bond geometry) and electronic structure (e.g. molecular orbitals, electron densities). Further, with excitation and emission spectroscopy, we measured an in vitro change in Eu 3+ coordination upon coming in contact with nicotianamine, supporting the formation of a Eu 3+ -nicotianamine complex. ## II.A.1. Ab initio molecular dynamics simulations All geometry optimizations and AIMD simulations were performed in 17.5 cubic periodic boxes within the PBE/LnPP1 GTH level of theory as implemented in the CP2K computational chemistry software package. 70 Core electrons were modeled with norm-conserving GTH pseudopotentials, while valence electrons including f electrons, were treated with polarizable double-zeta quality basis sets. 71 We used our recently developed LnPP1 pseudopotentials and basis sets for europium. 58 The long-range electrostatics terms were calculated with an additional plane wave basis set with a 500 Ry cutoff. Grimme's corrections 72 were used to account for van der Waals interactions with a 6.0 radius. All AIMD simulations were done in the NVT ensemble, with a 1.0 fs or 0.5 fs time step. A 1.0 fs time step is sufficient, although initially we used a 0.5 fs time step. Initial complex structures were placed in the center of the periodic box and solvated with water molecules. We used the following protocol to obtain optimized aqueous structures of the Eu 3+ -ligand complexes: we initially performed 5 ps of NVT AIMD simulations at 363 K, which were followed by a slow annealing to 0 K with rescaling factor for annealing velocities equal to 0.997, and final geometries optimizations. Starting from the optimized geometries, production AIMD simulations of Eu 3+ -ligand complexes were performed at 300 K for >10 ps trajectories. The analysis of radial distribution functions (RDFs) and coordination numbers (CNs) of the studied systems was done for equilibrated parts of the trajectories that corresponded to at least ~10 ps. Potential energy plots along the production AIMD trajectories appear in the Supporting Information (SI), Figure S1. All of the models containing Eu 3+ had a septet spin multiplicity. We modelled the Eu 3+ -EDTA complex with the EDTA protonation state corresponding to pH 11 (four carboxylate sites and two amine sites deprotonated) resulting in EDTA 4-. The EDTA protonation states were chosen to correspond to pH 11 so to be able to directly compare with the published Eu 3+ -EDTA complex structure. 68,69 Our model periodic unit cell contained 574 atoms, which included 33 atoms representing the [Eu 3+ -EDTA 4-]complex, 180 explicit water molecules, and 1 Na + counter ion, allowing for the overall neutral charge of the periodic unit cell (Figure 1a). Since the structure of the [Eu 3+ -EDTA 4-]complex in solution is known, we initially constructed the complex with the known structure, and subjected it to the AIMD protocol described. We modelled the Eu 3+ -NA complex with NA as a zwitterion (three carboxylate sites deprotonated and three amine sites protonated) resulting in a net uncharged ligand. NA has a zwitterionic protonation state from pH ~3.2 to pH ~7.7. 20,73 We modelled NA as a zwitterion to directly compare with our experiments performed at pH ~5. Our model periodic unit cell contained 586 atoms, which included 43 atoms corresponding to the [Eu 3+ -NA] 3+ complex, 180 explicit water molecules, and three Clcounter ions allowing for the overall neutral charge of the unit cell (Figure 1b). Because we did not know a priori the [Eu 3+ -NA] 3+ complex structure, and in order to account for different possible conformations, we independently constructed this system with three different initial structures: i) structure with all three NA-COOgroups having bidentate binding to Eu 3+ , ii) structure with two bidentate NA-COOgroups binding to Eu 3+ and one monodentate and iii) structure with one bidentate NA-COOgroup and two monodentate. All three initial structures were independently subjected to the AIMD protocol described. Due to the large number of degrees of freedom in explicitly solvated [Eu 3+ -NA] 3+ complexes, we performed additional ~4ps AIMD simulations at 500 K, to verify their stabilities. ## II.A.2. Electronic structure analysis We studied the electronic structure of the optimized complex structures in further detail by: i) examining how the electron density changes between the bound Eu 3+ -ligand complex and unbound Eu 3+ ion and ligand, and ii) analyzing the molecular orbitals (MOs) of the Eu 3+ -ligand complex coordination bonds. Starting from the optimized (annealing plus geometry optimization) solution structures of our Eu 3+ -ligand complexes with explicit water molecules, we extracted the atomic coordinates of the Eu 3+ -ligand complexes, including water molecules that are directly coordinated with the Eu 3+ ion, and performed single point energy calculations in the Gaussian software package 74 with PBE functional, 57 ECP28MWBSEG basis set with corresponding effective core potential for Eu, 64,75 and cc-pVTZ basis set for other chemical elements within the model systems. 76 All single point energy calculations were done with polarizable continuum model implicit water solvent. All geometry optimizations were done in the solution phase (i.e. with explicit water molecules, periodic conditions) with CP2K, and we extracted the coordinates of only the molecules (ligand and water) coordinated with Eu 3+ ions to analyze the electronic structure of the complexes. We did not further optimize the structures with implicit solvent to directly probe the electronic properties of the structures resolved with DFT and AIMD in aqueous solution. For each complex, we examined how the electron density changes between the formed Eu 3+ -ligand complexes (including coordinated water molecules) and their unbound ligand, coordinated water molecules, and Eu 3+ ion. This was done through subtraction of electronic densities 81 of the bound and unbound states, as represented by equation 1. Where ∆𝜌 is the change in electronic density upon binding, 𝜌 %&'()*+ is the electron density of ## II.B.2. Photophysical studies The photoluminescence data were obtained on a Fluorolog-3 spectrofluorimeter (Horiba FL3-22-iHR550), with a 1200 grooves/mm excitation monochromator with gratings blazed at 330 nm and a 1200 grooves/mm emission monochromator with gratings blazed at 500 nm. An ozone-free 450 W xenon lamp (Ushio) was used as a radiation source. The emission spectra were measured in the range 550-725 nm using a Hamamatsu 928P detector. All emission spectra were corrected for instrumental function. The emission decay curves were obtained using a Horiba TCSPC system and a Xe pulsed lamp as the excitation source. The number of coordinated water molecules (q) was obtained by measuring the emission decay lifetimes in water and D2O and equation 2. 82 All photophysical measurements are the average of at least three independent measurements, and, unless otherwise indicated, performed at 25.0 ± 0.1 o C. ## III. Results and Discussion III.A. Our computational approach replicates the known Eu 3+ -EDTA complex structure Our optimized [Eu 3+ -EDTA 4-×(H2O)3]complex structure matches the coordination structure previously resolved with X-ray crystallography 68 , where the Eu 3+ ion is 9-coordinate with 4 oxygen atoms from the coordinated monodentate carboxylates, two nitrogen atoms from the coordinated amines, and 3 oxygen atoms from the coordinated water molecules. This confirms that our Eu pseudopotential and basis set, as well as the AIMD protocol described, accurately replicate the structure of the [Eu 3+ -EDTA 4-×(H2O)3]complex in solution, in which the EDTA protonation states correspond to pH 11. The RDFs calculated for the [Eu 3+ -EDTA 4-×(H2O)3]complex, shown in Figure 3 ## III.B. The solution structure of Eu 3+ -nicotianamine complexes resolved with ab initio molecular dynamics In order to account for different possible conformations of [Eu 3+ -NA] 3+ complexes, we performed AIMD simulations of this system with three different initial conformations, as described in the computational section. We found two stable [Eu 3+ -NA] 3+ complex structures. One structure includes three water molecules directly coordinated to the Eu 3+ ion (Figure 4a), in which the Eu 3+ ion is coordinated to two carboxylates in a bidentate conformation and one in a monodentate conformation. In the [Eu 3+ -NA×(H2O)3] 3+ complex structure (Figure 4a), coordination sites 1, 2, and 3 correspond to water molecules, coordination sites 4, 5, 7 and 8 to bidentate -COOgroups, and coordination site 6 to a monodentate -COOgroup. The other structure has four water molecules directly coordinated to the Eu 3+ ion (Figure 4b), in which the Eu 3+ ion is coordinated to one carboxylate in a bidentate and two in a monodentate fashion. In the [Eu 3+ -NA×(H2O)4] 3+ complex structure (Figure 4b), coordination sites 1, 2, 3 and 4 correspond to water molecules, coordination sites 5 and 7 to the bidentate -COOgroup, and coordination sites 6 and 8 two each monodentate -COOgroup. The atomic cartesian coordinates of both complexes appear in the SI. Unlike the 9-coordinate [Eu 3+ -EDTA 4-×(H2O)3]complex, both Eu-NA complexes are 8-coordinate. While the [Eu 3+ -EDTA 4-×(H2O)3]complex includes nitrogen atoms (amine sites) in its first coordination sphere, both [Eu 3+ -NA×(H2O)n] 3+ complexes have only oxygen atoms in their first solvation shell. This is evidenced in the Eu-N RDF plots, which show a single peak for [Eu 3+ -EDTA 4-×(H2O)3] -~2.8 (Figure 3b), while more disordered peaks are seen for the [Eu 3+ -NA×(H2O)n] 3+ complexes at distances >4 (Figures 5b, 6b). This means that the amine NA sites are on the "outside" of the first coordination sphere of both Eu-NA complexes and in contact with solvent molecules, unlike the [Eu 3+ -EDTA 4-×(H2O)3]complex structure where most heteroatoms are coordinated to the Eu 3+ ion. Similarly, the Eu-C RDF plots of the [Eu 3+ -NA×(H2O)n] 3+ complexes are more disordered than the single Eu-C peak for [Eu 3+ -EDTA 4-×(H2O)3] -, which implies that the [Eu 3+ -EDTA 4-×(H2O)3]complex is more rigid than both [Eu 3+ -NA×(H2O)n] 3+ complexes. This has implications toward lanthanide-ligand complex stability and solubility. ## III.C. Eu 3+ coordination changes in the presence of nicotianamine Ln 3+ ions emit light, 4 a property we will use to determine Ln 3+ -ligand complex formation, and which arises from transitions within the 4f orbitals. These ions form mostly ionic bonds that, along with their core nature, leave the 4f orbitals substantially unaffected by the coordination environment. While f-f transitions are forbidden by the parity rule, the symmetry of the coordination environment partially lifts the restrictions, which enables the use of emission spectroscopy to characterize the symmetry of the coordination sphere of Eu 3+ . 83 Excitation and emission spectra are easily interpreted for the Eu 3+ ion, because it has a simple electronic structure in which the excited-and ground-state energy level multiplets do not overlap, 84 allowing for simple spectral interpretation. The fine structure in the emission spectra of Eu 3+ complexes is directly related to the coordination sphere around Eu 3+ and its symmetry. 85 Additionally, since the gap between the emissive and the ground electronic states of the ion is relatively narrow, the emission can be partially quenched by the presence of oscillators, such as OH functional groups. Excited state lifetimes in water and D2O can be used to evaluate the number of oscillators in the coordination sphere of Eu. 82,86 When Eu 3+ complexes are excited at 395 nm into the 7 F0 ® 5 L6 transition of Eu 3+ , the characteristic line-like emission spectrum of the metal ion is seen (Figure 7). Different coordination environments lead to slight changes in symmetry and thus fine structure, as can be seen from the spectra in Figure 7, which indicates that different species are obtained when EDTA coordinates to Eu 3+ in different protonation states (i.e. different pH values). Figure 7 shows that the coordination environment of Eu 3+ changed when in contact with EDTA, which agrees with the known fact that EDTA bind Eu 3+ , and validates our experimental approach. Figure 8 shows that the coordination environment of Eu 3+ changed when in contact with nicotianamine. The differences in emission and excitation curves between [Eu 3+ ×(H2O)n] 3+ and [Eu 3+ -NA×(H2O)n] 3+ strongly suggest that Eu and NA are forming a complex because similar differences are observed when Eu 3+ comes in contact with EDTA, which are known to form complexes. This supports, yet doesn't confirm, that nicotianamine binds Eu 3+ , as predicted by computation. (Section III.B). ## III.D. Electronic structure of Eu 3+ -ligand complexes To spatially elucidate differences in the electronic properties of the Eu 3+ -ligand complexes, we analyzed their electron density differences between bound and unbound states. Figure 9a shows that, although the Eu 3+ ion is coordinated to both nitrogen and oxygen atoms in the [Eu 3+ -EDTA 4-×(H2O)3]complex, the main change in ligand electronic density is observed at -COOcoordination sites. Three out of four monodentate -COOgroups of the [Eu 3+ -EDTA 4-×(H2O)3]complex display a substantial change in electronic density upon binding. It is unlikely that the particular -COOgroup in Figure 9a without a net electron density gain always remains so. Most likely, the time-averaged behavior is that three of four EDTA carboxylate groups are more tightly bound to the Eu 3+ ion than the fourth one, although the specific carboxylate groups displaying this behavior changes over time due to the dynamic nature of solution structures. This is shown from analysis of Eu 3+ -Ocarboxylate bond distances along the AIMD trajectory of the [Eu 3+ -EDTA 4-×(H2O)3]complex (Figure S3a), which reveals the distance's dynamics. Bond lengths vary between ~2.30 and ~2.70 . In most frames during the trajectory the distance of one carboxylate group is further from Eu 3+ that the remaining three. Therefore, the observation that a -COOgroup has no significant electron density loss (Figure 9a), is partly due to the fact that it is further from the remaining three in that particular optimized frame. Due to steric clashes and the coordination geometry around Eu 3+ ions, not all Eu-Ocarboxylate bond distances will be equal, and they will vary in the solution phase. Both [Eu 3+ -NA×(H2O)n] 3+ complexes in Figures 9b and 9c All complexes have water molecules coordinated to the Eu 3+ ion, however the [Eu 3+ -EDTA 4-×(H2O)3]complex has two water molecules with a significant net loss of electron density (Figure 9a), the [Eu 3+ -NA×(H2O)3] 3+ complex has one water molecule a significant net loss of electron density (Figure 9b), and the Eu 3+ -NA×(H2O)4] 3+ complex no bound water molecule with a significant net loss of electron density (Figure 9a). This indicates a higher contribution of water molecules to binding in the [Eu 3+ -EDTA 4-×(H2O)3]than the [Eu 3+ -NA×(H2O)n] 3+ complexes. Eu-Owater bond lengths along the AIMD trajectories for the [Eu 3+ -EDTA 4-×(H2O)3]and [Eu 3+ -NA×(H2O)n] 3+ complexes (Figure S5) show longer coordination bond lengths in the EDTA complex. This suggests that net electron density loss in a coordinating functional group has multiple causes, and bond lengths alone are not predictive of which coordinating functional group will contribute electron density to bound ion. To further investigate coordination bonds in [Eu 3+ -NA×(H2O)n] 3+ complexes, we analyzed their valence MOs, which were calculated with PBE/ECP28MWBSEG//cc-pVTZ level of theory. The MO diagrams for valence orbitals of [Eu 3+ -NA×(H2O)3] 3+ and [Eu 3+ -NA×(H2O)4] 3+ appear in Figure 10a, and a depiction of selected valence MOs are shown in Figure 10b. We focused on analysis of the valence orbitals rather than the electronic excitations, because, as shown in previous studies, 87,88 such analysis is sufficient to qualitatively elucidate coordination properties. TDDFT or multi-reference calculations are necessary for more robust quantitative validation of Ln 3+ complexes stabilities; 32, however, in this work we focus on describing the structure of the complexes, and we will quantify stabilities in the near future. Thus, in the present paper, we provide a qualitative description of the electronic structure of the previously unresolved Eu 3+nicotianamine complexes to complement our molecular structure findings (Section III.B). Figure 10a shows that the energy levels of MOs in the [Eu 3+ -NA×(H2O) 3+ -NA×(H2O)3] 3+ complex (Figure 9b) but not in the [Eu 3+ -NA×(H2O)4] 3+ (Figure 9c). Based on these observations, we hypothesize that the [Eu 3+ -NA×(H2O)3] 3+ complex is more stable than [Eu 3+ -NA×(H2O)4] 3+ , and nicotianamine will preferably bind Eu 3+ with two bidentate and one monodentate carboxylate groups. ## IV. Conclusions This work shows that our pseudopotentials and basis sets with AIMD simulations can accurately predict the structure of lanthanide-ligand complexes in solution (i.e. in the condensed phase), by replicating the structure of the [Eu 3+ -EDTA 4-×(H2O)3]complex in basic conditions within 0.05 , compared to X-ray fine structure absorption spectroscopy measurements from the literature. 68,69 Paired with the fact that our pseudopotentials and basis sets can accurately predict lanthanide reactivity, 58 a very powerful computational approach allows us to properly predict lanthanide structures and reactions in the condensed phase. This will be highly useful to efforts in rare earth separation and purification, lanthanide-ligand based medical contrast agents, single molecule magnets, luminescent molecules, and any field in which resolving the structure and reactivity of lanthanides in the atomic scale is relevant. This work also reports two very similar Eu 3+ -nicotianamine complex structures that were predicted using our computational approach, which is supported by the fact that the same computational methods and approach were used to replicate the known Eu 3+ -EDTA structure. While our excitation and emission spectroscopy measurements by themselves do not resolve the structure, they confirm that the coordination sphere of the Eu 3+ ions changes when in contact with nicotianamine, strongly supporting that the computationally predicted Eu 3+ -nicotianamine complex is forming. The spatial distribution of electron density gain or loss shows that not all coordinating groups contribute to the electron density gain in the bound ion. Different binding modes (mono-or bidentate) of a single carboxylate changes the electronic properties of all complex coordination bonds. We are currently quantifying lanthanide-ligand complex stabilities and thermodynamics, and will report them in the near future. Although knowledge that nicotianamine binds europium does not determine how plants uptake rare earths from the environment, it strongly supports that phytosiderophores bind lanthanides.

chemsum

{"title": "Solution structure of a europium-nicotianamine complex supports that phytosiderophores bind lanthanides", "journal": "ChemRxiv"}

molecular_signature_of_pseudomonas_aeruginosa_with_simultaneous_nanomolar_detection_of_quorum_sensin

## Abstract: Electroanalysis was performed using a boron-doped diamond (BDD) electrode for the simultaneous detection of 2-heptyl-3-hydroxy-4-quinolone (PQS), 2-heptyl-4-hydroxyquinoline (HHQ) and pyocyanin (PYO). PQS and its precursor HHQ are two important signal molecules produced by Pseudomonas aeruginosa, while PYO is a redox active toxin involved in virulence and pathogenesis. This Gramnegative and opportunistic human pathogen is associated with a hospital-acquired infection particularly in patients with compromised immunity and is the primary cause of morbidity and mortality in cystic fibrosis (CF) patients. Early detection is crucial in the clinical management of this pathogen, with established infections entering a biofilm lifestyle that is refractory to conventional antibiotic therapies. Herein, a detection procedure was optimized and proven for the simultaneous detection of PYO, HHQ and PQS in standard mixtures, biological samples, and P. aeruginosa spiked CF sputum samples with remarkable sensitivity, down to nanomolar levels. Differential pulse voltammetry (DPV) scans were also applicable for monitoring the production of PYO, HHQ and PQS in P. aeruginosa PA14 over 8 h of cultivation. The simultaneous detection of these three compounds represents a molecular signature specific to this pathogen.Bacterial species rely on a wide array of signaling molecules, signal detection systems, and signal-transduction mechanisms to coordinate gene regulation 1 . Quorum sensing (QS) is a cell-to-cell communication system, which involves the production and detection of diffusible signaling molecules, known as 'autoinducers' or bacterial pheromones 1,2 . Chemically diverse QS signaling molecules ranging from N-acyl homoserine lactones (AHLs) and 2-alkyl-4-quinolones (AHQs) to peptides and furanones, coordinate activities including bacterial secondary metabolite production, biofilm development, bioluminescence, competence, plasmid transfer and pathogenicity 2,3 . Pseudomonas aeruginosa, a Gram-negative opportunistic pathogen, is capable of surviving in a broad range of natural environments. It is an antibiotic-resistant human pathogen associated with hospital-acquired infections 4 and causes acute pneumonia and chronic lung infections in cystic fibrosis (CF) patients 5 . Two AHL-based systems (the las and rhl) together with the AHQ-based system form a complex QS hierarchy in P. aeruginosa which controls global gene expression 6 . The primary P. aeruginosa AHQ signaling molecules are 2-heptyl-3-hydroxy-4-quinolone 1 ("Pseudomonas Quinolone Signal", PQS) and its immediate precursor, 2-heptyl-4-hydroxyquinoline 2 (HHQ) 7 . Pyocyanin 3 (1-hydroxy-N-methylphenazine, PYO), one of the several phenazine-based secretory products of P. aeruginosa (Fig. 1), is also involved in QS 8 . an important virulence and pro-inflammatory factor. As a redox-active molecule, PYO is a potential source of damaging reactive species, which plays a significant role in oxidative stress and inducing lung injury infected by P. aeruginosa. In general, P. aeruginosa is the only species known to produce PQS and PYO while HHQ can be generated by Burkholderia strains 9 . Also, the lack of these molecules does not mean that P. aeruginosa is absent, since certain isolates lack a functioning AHQ system (e.g. P. aeruginosa PA7) 10 . Among different analytical methods for AHQs and PYO detection, an LC-MS/MS method was developed to profile a broad range of QS signaling molecules including PQS and HHQ 3 . The PYO levels of patient sputa ranging from 0 to 27.3 μ g mL −1 were detected by HPLC-UV detection 11 . Chromatography can be paired with mass spectrometry to enhance detection sensitivity, however, this approach requires sample pretreatment, lengthy analysis times and high costs, whereas fluorometry is sample-consuming and lacks selectivity 12 . Accurate and rapid point-of-care diagnosis has stimulated efforts to develop simple and sensitive electrochemical strategies. Biosensor-based assays for PQS and HHQ have been reported 2,13,14 in addition to simple cyclic voltammetry (CV) and amperometry using a boron-doped diamond (BDD) thin-film electrode as an excellent method for the detection of HHQ, PQS 15 and 2-(2-hydroxyphenyl)-thiazole-4-carbaldehyde (IQS) 16 . Adsorptive stripping voltammetry (AdSV) using a hanging mercury drop electrode (HMDE) 17 and differential pulse voltammetry (DPV) using graphite rods 18 were employed for PYO detection in biological samples. Disposable screen-printed electrodes were also used to probe the presence of PYO in the human biofluids by square wave voltammetry (SWV) . Sismaet et al. measured the production of PYO in the presence of various amino acids. The presence of the amino acids resulted in a faster and stronger electrochemical response 22 . Kim et al. 23 developed a bio-based redox capacitor for in situ monitoring of the production of PYO during P. aeruginosa cultivation. The catechol-grafted chitosan film amplified the electrochemical signals of PYO and lowered the detection limit. Miniaturized PYO sensors in the form of a 'smart-bandage' 24,25 or 'nanofluidic' 8,26 were also reported. There are, however, only two reports on the simultaneous electrochemical determination of PQS and PYO 27,28 . In the former, a conductive polymer film modified glassy carbon electrode has been applied to increase electroactive electrode surface. In the latter approach, a P. aeruginosa strain was grown on the electrode surface to concentrate electrochemical signals of PYO and PQS. The BDD thin-film electrode 29,30 featuring a wide potential range, high current density, extreme electrochemical stability, low background current and high resistance to fouling has proven promising for PQS detection 15 . This report unravels the use of a BDD electrode, without modification, for simultaneous determination of PYO, PQS and HHQ in a mixed solution. The influence of electrolyte pH on the peak potential separation and peak currents was also investigated. The optimized procedures were then applied to analyze supernatant extracts from P. aeruginosa wild-type strains, monitor the production of signaling molecules in the bacterial strain PA14 and detect the signals in CF patient sputa cultured with P. aeruginosa using a bare BDD electrode, taking advantages of the hydrogen surface termination and sp 3 carbon bonding without the extended π -electron system for sensitive detection over other solid electrodes. ## Results and Discussion Electrochemical behavior of PQS, HHQ, and PYO. The redox reactions of PYO and the quinolones on the BDD electrode were studied by CV. The CV obtained for PQS at pH 5.0 on the pristine BDD electrode exhibited three oxidation peaks at + 0.759 V, + 1.103 V and + 1.590 V (Fig. 2a). In contrast, only one well-defined peak at + 1.351 V was observed on the CV of HHQ (Fig. 2b), indicating the simultaneous voltammetric measurements of the quinolones is feasible due to the peak potential differentiation. The CV of PYO (Fig. 2c) presented a pair of redox peaks (oxidation at − 0.1 V and reduction at − 0.153 V) at the negative potential range, making these potentials unique for the identification/determination of PYO. PYO also provoked an oxidation peak at + 0.871 V, resulting in a peak separation (∆ Ep) of 112 mV to the first oxidation peak of PQS (+ 0.759 V). DPV was chosen for further experiments as a sensitive technique with good distinction against the background current. Effect of electrolyte pH on the detection of PYO, HHQ, and PQS. The influence of electrolyte pH on the peak potentials (Ep) and peak currents (Ip) of the electrooxidation of PYO, HHQ and PQS was investigated by DPV in the pH range of 4.0 to 7.0. As illustrated in Fig. 3a, the anodic peak potential for the oxidation of PYO shifted to negative values when increasing the buffer pH with linear regression of Ep (V) = 0.137-0.0551 (pH), indicating the role of protons in the oxidation process 31 . The determined slope was 55.1 mV/pH, close to the reported value of 61 mV/pH 27 and the expected value of 59 mV/pH. Such behavior confirmed the number of hydrogen ions was equal to electrons taking part in the electrode reaction. The peak current (Ip) was also dependent on pH, which increased when the electrolyte pH decreased from 7.0 to 4.0. As protonation was involved in the catalytic reactions, the electrocatalytic oxidation of PYO became more favorable at lower pH 32 thus resulting in higher peak current. However, the oxidation peak of PYO at pH 4.0 was 3.16% broader compared to the sharp peaks obtained at pH 5.0. The peak current intensity was decreased by increasing the pH to 6.0 and 7.0. The effect of buffer pH on the response of HHQ (Fig. 3b) and PQS (Fig. 3c) was investigated over the pH range 4.0 to 7.0. As a result of increasing pH, the peak potential shifted to less positive values. Figure 4 shows the DPVs for a ternary solution mixture of PYO, PQS and HHQ obtained with varying buffer pH. All the oxidation peaks of PQS and HHQ shifted to less positive values in the potential range of + 0.5 V to + 1.5 V and the oxidation peak of PYO shifted to more negative values in the potential range of − 0.8 V to − 0.25 V when increasing the pH from 4.0 to 7.0. The DPVs obtained at pH 5.0 and 6.0 exhibited five well-defined anodic peaks, whereas partially overlapped peaks of the biomarkers were observed in the voltammograms obtained at pH 4.0 and 7.0. The peak separations of PYO and PQS in the potential range of + 0.65 V to + 1.15 V were 138.9 mV, 176.2 mV, 170.5 mV and 161.1 mV for pH 4.0, 5.0, 6.0 and 7.0, respectively. The peak positions of the analytes might be attributed to different electrochemical activities of their functional groups on the electrode surface 33 . Similar to PYO, the oxidation peak currents of PQS and HHQ decreased with elevated pH. Therefore, considering the peak potential resolution and detection sensitivity, pH 5.0 acetate buffer (50 mM) was chosen as the optimal medium. ## Individual determination of PYO, PQS and HHQ. Considering sufficient peak potential difference and superb detection sensitivity in individual determinations, a series of studies was then performed to verify the feasibility of selective detection of PQS, HHQ and PYO by DPV using the BDD electrode. Experiments were carried out by changing the concentration of the target analyte whilst maintaining those of the other two species constant. Figure 5a shows the DPV response of PYO at different concentrations (2-100 μ M) while PQS and HHQ were constant at 20 μ M and 10 μ M, respectively. The peak potential of PYO at − 0.14 V remained unchanged when varying the concentrations of the compound. The current intensities of the oxidation peaks raised proportionally to increasing PYO concentration, in the range of 2-100 μ M with good linearity (R 2 = 0.991, Fig. 5b). The peak currents of PQS and HHQ decreased slightly without significant loss of electrocatalytic activities, due to the irreversible adsorption of the oxidation products on the electrode surface which hindered the further oxidation of PQS and HHQ on the electrode surface 34 . Similarly, Fig. 5c displays the DPV calibration of HHQ in the presence of PYO and PQS exhibiting satisfactory linearity in the range of 2-75 μ M (R 2 = 0.997, Fig. 5d). Also, Fig. 5e depicts the DPV calibration of PQS in the presence of PYO and HHQ demonstrating acceptable linearity in the range of 2-100 μ M (R 2 = 0.996, Fig. 5f). The oxidation peak of PQS ~+ 1.0 V was selected for calibration. The analytical parameters for simultaneous determination of PYO, HHQ and PQS are present in Table 1. The analytical performance of the BDD electrode using DPV is compared with the literature methods in Table 2. The effect of the pH and potential to obtain a lower limit of detection of PYO, HHQ and PQS are also investigated using amperometric measurement (I/t) and compared to DPV on the BDD electrode in Table 3. In the case of PQS and HHQ, there are two parameters that can result in lower LOD; the pH and the oxidation potential, for example, pH 2.0 gives sensitive LOD at + 0.8 V and + 1.2 V for PQS and HHQ, respectively. While in the PYO case, the oxidation potential at + 1.0 V can result in lower LOD at either pH. ## Analysis of P. aeruginosa PA14 cultures, growth curves and clinical sample analysis. The optimized method on DPV was initially applied to the analysis of cell-free culture (supernatant) extracts of P. aeruginosa PA14 and pqsA mutant. DPV of the biological sample extracts were recorded by diluting the samples into 50 mM acetate buffer (pH 5.0) containing 20% ACN. As expected, no analytes were detected from the supernatant of P. aeruginosa pqsA mutant. Nevertheless, P. aeruginosa PA14 produced a significant amount of PYO, in addition to the presence of PQS and HHQ (Fig. 6a). The concentration of PYO, HHQ and PQS measured in the cell-free extract of microbial strain PA14 were 37.03 ± 0.76 μ M, 4.48 ± 0.43 μ M, and 11.17 ± 0.15 μ M, respectively (n = 3). Subsequently, a time-course analysis was performed using the BDD electrode to monitor the real-time concentration profiles of the three target molecules from early log phase into the stationary phase of growth. Cultures were sampled at 1 h intervals from the mid-log phase and monitored as before. Consistent with the established kinetics of signal production in P. aeruginosa, HHQ was initially identified at the highest concentration 35,36 , not surprising given that it is the precursor to PQS (Fig. 6b) 37 . As the cells entered stationary phase, both PQS and PYO 38 become more abundant, with HHQ levels markedly reduced at 8 h (Fig. 6c,d). CF sputum is a complex mixture of airway mucus glycoproteins, serum, proteins, DNA, alginate, and rigidifying lipids as well as inflammatory substances including polymorphonuclear leukocytes, antibodies, antimicrobial peptides, and dead host cells. Also, CF sputum samples are more viscous than normal samples as a result of negatively charged biopolymers (mucin, DNA, and alginate) which are connected through noncovalent interactions such as electrostatic, hydrogen, and hydrophobic bonding. All these matrix constituents protect the epithelial cells and form a diffusion barrier for pathogen and harmful particles 39,40 . In order to evaluate the extent of matrix interference, a sputum sample obtained from a paediatric CF patient who was not infected with P. aeruginosa was spiked with known amounts of PYO (10, 20, and 40 μ M), HHQ (20, 40, and 80 μ M), and PQS (40, 80, and 160 μ M). All blank and spiked sputum samples were then extracted twice with chloroform (1:2, v/v sputum sample: chloroform). Importantly, the blank sputum sample did not produce any signature signal, consistent with the absence of P. aeruginosa signaling molecules. Therefore, the corresponding DPV exhibited no oxidation peaks. In the case of spiked sputum samples, all the target analytes were detected and recovery values measured were up to 32%, 43%, and 58% for PYO, HHQ, and PQS, respectively. The LOD of PYO, HHQ, and PQS in CF sputum sample using DPV on the bare BDD electrode is 0.15 μ M, 0.62 μ M, and 1.25 μ M, respectively (S/N = 3). In addition, the sputum sample was spiked with 1 × 10 5 cells of P. aeruginosa and incubated at 37 °C for several days to promote bacterial growth. The sputum sample was then extracted with chloroform, dried with a rotary evaporator and re-dispersed with ACN. The DPVs were recorded by adding a certain amount of the reconstituted sample into 50 mM acetate buffer (pH 5.0) containing 20% ACN. Figure 7 compares the DPV profiles of the blank and P. aeruginosa spiked sputum samples incubated for 3 d and 11 d. No peak obtained at the negative potential on the DPV curve of the sputum sample that was incubated for 3 d indicates the absence or a non-detectable amount of PYO. However, a clear peak at − 0.14 V was exhibited on the DPV curve of the sample incubated for 11 d. A concentration of 2.8 μ M, 7.2 μ M, and 10.4 μ M was calculated for PYO, HHQ, and PQS, respectively in spiked CF sputum sample incubated for 11 d. The target analytes were not detected in the sputum sample incubated for 3 d, likely due to a lower concentration of HHQ and PQS at this stage 41 and the matrix effect. ## Conclusions The BDD electrode was successfully used for simultaneous determination of PYO, HHQ and PQS in both standard mixtures and microbial P. aeruginosa strain PA14 using a conventional and rapid extraction method. The production of three target signals in P. aeruginosa PA14 over 8 h was monitored, illustrating the selectivity of BDD electrode with the obtained results in agreement with previous reports 35,36,38 . Most importantly, the application of the optimized method was extended to the CF sputum sample cultured with P. aeruginosa. In this regard, the importance of including the three biomarkers is highlighted by the recent suggestion that HHQ predominates over PQS in CF sputum samples 42 . Synthesis of PQS and HHQ. PQS and HHQ were synthesized as described by Pesci et al. 43 ## and McGlacken et al. 44 . The resulting products and their purity were confirmed by 1 H NMR (300 MHz) and spectra are shown in a previous publication 45 . Electrode preparation. The BDD electrode was polished with polishing papers (Buehler, UK) and subsequently with alumina (Buehler, UK) until a mirror finish was obtained. After cleaning the electrode with deionized water, the electrode was sonicated in 2-propanol and deionized water for 5 min and 10 min, respectively. Subsequently, the electrode was cleaned by CV between − 1.0 V and + 2.0 V versus Ag/AgCl (3 M KCl) at 100 mV s −1 in 50 mM acetate buffer (pH 5.0) until a stable CV profile was obtained. Bacterial strains, growth conditions and clinical samples. Bacterial supernatant extracts were obtained using a modified version of the Fletcher protocol 2 . Briefly, overnight cultures of P. aeruginosa PA14 were transferred into fresh Luria-Bertani (LB) broth (OD 600nm 0.01) and incubated for 7 h at 37 °C (total 20 mL). Culture supernatants were obtained by centrifugation at 4000 revolutions per minute (rpm) for 15 min and subsequently filter sterilized using Minisart (Sartorius) 0.2 μ m filter into a clean centrifuge tube. Extractions using acidified ethyl acetate (0.01% (v/v) glacial acetic acid) were performed twice on bacterial cell-free cultures (end-point assays) (1:2, v/v supernatants: acidified ethyl acetate) or whole cell cultures (time-point assays) (1:1, v/v whole cell culture: acidified ethyl acetate). In both assays, the organic phase was evaporated using a rotary evaporator and the residue was dissolved in ACN (end-point assay) and in 0.5 M formic acid pH 2.0 (time-point assay). Solid-phase extraction using MCX SPE was carried out after liquid-liquid extraction for time-point assays. The reconstituted samples were subjected to MCX SPE cartridges pre-conditioned with methanol and deionized water. The cartridges were washed then with 0.5 M formate buffer, pH 2.0 to increase the retention of analytes and washed with 100% methanol to remove neutral interferences. Subsequently, the analytes were eluted using 5% 4.5 M ammonium formate in methanol. Ethics statement and sputum processing. Sputum samples were collected from paediatric patients attending the CF clinic at Cork University Hospital, Ireland. Ethical approval was granted by the Clinical Research Ethics Committee (CREC) for sputum collection and samples were handled in accordance with the approved guidelines. All methods were carried out in accordance with the approved guidelines. Written informed consent from all patients/guardians was obtained for acquisition and analysis outlined in this study. Patient sputum samples were inoculated with 1 × 10 5 cells of P. aeruginosa and incubated at 37 °C for 3 and 11 d, respectively. CF patient samples were extracted twice with chloroform (1:2, v/v sputum sample: chloroform). As mentioned above, the organic phase was evaporated using a rotary evaporator and the residue was dissolved in ACN.

chemsum

{"title": "Molecular Signature of Pseudomonas aeruginosa with Simultaneous Nanomolar Detection of Quorum Sensing Signaling Molecules at a Boron-Doped Diamond Electrode", "journal": "Scientific Reports - Nature"}

incorporation_of_eu(iii)_into_calcite_under_recrystallization_conditions

## Abstract: The interaction of calcite with trivalent europium under recrystallization conditions was studied on the molecular level using site-selective time-resolved laser fluorescence spectroscopy (TRLFS). We conducted batch studies with a reaction time from seven days up to three years with three calcite powders, which differed in their specific surface area, recrystallization rates and impurities content. With increase of the recrystallization rate incorporation of Eu 3+ occurs faster and its speciation comes to be dominated by one species with its excitation maximum at 578.8 nm, so far not identified during previous investigations of this process under growth and phase transformation conditions. A long lifetime of 3750 μs demonstrates complete loss of hydration, consequently Eu must have been incorporated into the bulk crystal. The results show a strong dependence of the incorporation kinetics on the recrystallization rate of the different calcites. Furthermore the investigation of the effect of different background electrolytes (NaCl and KCl) demonstrate that the incorporation process under recrystallization conditions strongly depends on the availability of Na + . These findings emphasize the different retention potential of calcite as a primary and secondary mineral e.g. in a nuclear waste disposal site. The long-term safety of geological nuclear waste disposals has to rely on a detailed and thorough understanding of the processes governing the migration of radionuclides through the geo/biosphere. These reactions can include adsorption and surface precipitation as well as incorporation within the bulk of a material. Calcite, one of earth's most abundant minerals, has long been known for its potential to incorporate other cations than calcium and is present in significant amounts in clay rock formations 4 as well as as a degradation products of cementitious materials 5 . Consequently, calcite may play a significant role as a retention barrier for the transport of contaminants in the environment. Its performance as such a barrier will depend on the specific conditions of the interaction. Calcite is also of high relevance for many technical processes, where it occurs as scales, which have been shown to form solid solutions with certain contaminants, e.g. radium and other naturally occurring radioactive materials 6,7 . In many geological and geotechnical scenarios, such as nuclear waste storage in deep geological formations, the contaminants will interact with pre-existing calcite close to thermodynamic equilibrium with its contacting solution, under conditions where the incorporation process can be expected to be controlled by the mineral's recrystallization rate. Plutonium and the minor actinides neptunium, americium, and curium are expected to contribute most significantly to the long-term radiotoxicity according to performance safety assessments 8,9 . Under the reducing conditions expected for a deep geological disposal site the preferred oxidation state for Am and Cm, but also possibly for Pu is + 3. Europium has often been chosen as a homologue for trivalent actinides because of its similar ionic charge and radius 10,11 , as well as for its remarkable luminescent properties 12 . The capacity of calcite to incorporate guest ions with similar ionic radius compared to Ca 2+ (e.g. Eu 3+ , Am 3+ , and Cm 3+ ) under growth, phase transformation and even dissolution conditions was already evidenced in the past 1,3, . These processes represent an important retention mechanism for many contaminants, such as Pb 2+ or Cd 2+ , but also trivalent actinides (Pu 3+ , Am 3+ , Cm 3+ ) relevant to long term nuclear waste storage. In these previous studies it was shown that trivalent rare earth elements and actinides can be taken up by the calcite host lattice structurally. Natural mineral specimen contain considerable amounts of rare earths depending strongly on their origin. These findings imply that under natural conditions, mechanisms must be at work by which dissolved metal ions in contact with calcite at equilibrium incorporate into the material over time. Moreover, the dissolution and precipitation behavior of abundant minerals like calcite under different fluid compositions is important for the evaluation of geothermal energy production sites 26,27 . Also for the long-term storage of atmospheric CO 2 in geological formations calcite and other carbonates with their slow kinetics of mineral-fluid reactions play a significant role 28 . The interaction of Eu 3+ as well as Cm 3+ with calcite synthesized in mixed flow reactors 3,13,29 or formed by mineral phase transition 14 has previously been investigated at the molecular level by site-selective time-resolved laser-induced fluorescence spectroscopy (TRLFS). Here, three main species were identified. Species A corresponds to a surface species with ~2 water molecules left in its hydration shell and for Eu showed a peak maximum for the 7 F 0 → 5 D 0 transition at 578.1 nm. Species B (λ exc (Eu) = 578.4 nm) was found to be incorporated within the strongly distorted calcite structure. Finally, species C (λ exc (Eu) = 579.6 nm) was found to be located on the Ca 2+ site on the calcite lattice, with an almost undisturbed octahedral symmetry, identifiable by its characteristic splitting pattern. For both species B and C, Na + was required for charge compensation during the incorporation, suggesting a mechanism where two Ca 2+ ions were replaced by one Eu 3+ atom and one Na + atom. Atomistic simulations by Vinograd et al. 30 find a dolomite-like layered structure for the hypothetical NaEu(CO 3 ) 2 end member, and they suggest the formation of such domains in the solid solution. The Ca 2+ lattice site in dolomite has a D 3d point symmetry, as opposed to the C 3i symmetry of the same lattice site in calcite. Therefore, Eu in such a crystallographic environment should exhibit a splitting pattern similar to that of species C in calcite. Where calcite occurs as a primary phase, interaction will occur close to equilibrium. Studies dedicated to the interaction of Eu 3+ or Cm 3+ with calcite under recrystallization conditions are scarce. Our previous work 29 studied the sorption of Eu 3+ on a single calcite powder in 0.01 mol L −1 NaClO 4 at different contact times, i.e. one day, two weeks and one month at recrystallization conditions. In the excitation spectra, a broad peak centered at 579.3 nm with at least two species exhibiting partially preserved hydration shells were observed. Interaction was found to be dominated by a continuum of chemically similar adsorbed species as well as beginning incorporation into the bulk structure. An impact of the reaction time on the number of species and their lifetimes was noticed, but the transition was too slow to be probed meaningfully in the allotted time. Earlier Stumpf and Fanghänel 15 described a similar process of Cm 3+ interaction with calcite under recrystallization conditions. Via TRLFS of curium's 6 D 7/2 → 8 S 7/2 transition they observed a sorption species with two and an incorporation species with no water molecules bound to the Cm 3+ , respectively. Over a contact time of 6 months the incorporation of Cm 3+ in to the crystal bulk progressed at the cost of the sorption species. The incorporation species identified by Stumpf and Fanghänel 15 has its emission maximum at 618.0 nm, clearly distinct from the two incorporation species identified in co-precipitation experiments, species B at 616.3 nm and species C at 624.3 nm. Piriou et al. 31 also investigated the interaction of Eu 3+ with calcite. Batch sorption experiments were conducted at 50 °C and samples subsequently analyzed by site-selective TRLFS. Eu 3+ was found to be incorporated into a hydrated and/or hydroxylated surface layer, with two distinct site families. Substitution of Eu 3+ for Ca 2+ in the lattice framework was also in evidence. Nevertheless, sites A and B revealed by Schmidt et al. 3 and Marques Fernandes et al. 13 under growth conditions were not observed by Piriou et al. 31 Consequently, there are already clear hints in the literature highlighting the trend that speciation and interaction of Eu 3+ and Cm 3+ with calcite evidently is strongly affected by experimental conditions. To further investigate the impact of surface reactivity and recrystallization kinetics of calcite on the incorporation of Eu 3+ three calcite powders differing in their specific surface area (SSA) and impurities content were chosen and their reactions at near-equilibrium conditions studied. The evolution of the speciation was followed as a function of reaction time, starting from one week up to two months, and for some samples up to three years. In order to probe the relevance of the coupled substitution mechanism identified by Schmidt et al. 3 , some experiments were additionally conducted using KCl instead of NaCl as background electrolyte. ## Results Characterization of the solid phases. Three calcite powders (C1, C2, and C3) were chosen for this study. C1 and C3 were obtained as powders from Merck, Germany and Solvay, Germany, respectively. C2 was obtained from Ward's Science, USA as a single crystal and ground to a grain size < 63 μ m. All solids were characterized by powder XRD, BET, and SEM. In addition the impurity inventory was characterized by ICP-MS after dissolution of the minerals (for more details see Table S1 in the supplementary information (SI)). The sample characteristics are summarized in Table 1. The powder X-ray diffraction patterns of C1, C2, and C3 showed the presence of well-crystallized calcite, with no trace of other CaCO 3 allotropes such as aragonite or vaterite (for more details see Figure S1 in SI). The specific surface area (SSA) of the powders was found to be 0.6 ± 0.1, 1.1 ± 0.1 and 18.0 ± 0.1 m 2 g −1 for calcite C1, C2, and C3, respectively. The total organic content of calcites C1 and C2 were lower than 0.1 mg g −1 , while that of calcite C3 was 0.13 ± 0.04 mg g −1 . Representative micrographs showing the decreasing crystallite size from C1 to C3 are presented in Fig. 1. Sample C1 exhibited the overall lowest amount of impurities, the highest being Mg, Sr and Fe (7.5 to 14.5 μ g g −1 ). Concerning C2, impurities in Na, Mg, Si, K, Sr, Cs, Fe and Zn (10.4 to 142 μ g g −1 ) were present. In addition to the aforementioned elements for C2, material C3 had also minor impurities in Ba and Mn. The highest concentrations were found for Na (109 μ g g −1 ), Mg (1390 μ g g −1 ) and Si (389 μ g g −1 ). ## Calcite recrystallization rates. The exchange rate of calcium ions between the solid and liquid phase was determined by tracer experiments using the beta emitter 45 Ca. It is commonly assumed that the kinetics of this exchange is directly linked to the recrystallization rate and therefore the reactivity of calcite. The amount of recrystallized calcite material calculated from 45 Ca activity in solution as a function of reaction time is presented in Fig. 2. As also seen in the study by Berner et al. 32 , three stages of recrystallization kinetics can be identified. The highest rate is observed within the first hours of the reaction, followed by a decrease of the rate and eventually equilibrium which is reached when the concentration of the radioactive tracer in solution remains constant. The first two phases of this reaction are caused by the different availability of tracer ions in solution and solid. First, an excess supply of 45 Ca in solution leads to more incorporation than dissolution. With time, significant amounts of the tracer have become incorporated into the bulk and are then subject to dissolution and resolvation decreasing the overall recrystallization rate. Once chemical equilibrium is reached between dissolution and incorporation of 45 Ca, i.e. the same molar amount of ions is incorporated as is dissolved, no more information can be drawn from the data. When the second phase of the process (1.7-25.7 h) is used for data fitting, both calcites C2 and C3 exhibit significantly higher rates compared to C1. Material C1 recrystallizes much slower and equilibrium is not reached within the reaction time. Clearly, C1's recrystallization rate is significantly slower compared to the other materials, and we would expect this material to show the slowest surface reaction kinetics. Rate determination here is only slightly above detection limit. The rates of calcites C2 and C3 are rather high and equilibrium is reached rapidly. The specific rates, however, differ from each other with C3 being the most reactive. ## Species identification and reaction kinetics. The 3 different calcites (C1, C2, and C3) sorbed nearly all Eu 3+ of the solution independent of the reaction time (see Table S3 in SI). The excitation spectra of the ( 5 D 0 → 7 F 0 ) transition for all samples recorded at various reaction times are presented in Fig. 3. In total, four species named α -δ were identified. Their occurrence and relative distribution depends strongly on the material used, and hence the recrystallization rate. The characteristics of these species are summarized in Table 2. Seven to ten days. Already after ca. one week the spectra show significant differences between C1, C2, and C3, while all spectra show a broad distribution. For the calcite C1 with the lowest recrystallization rate, only one broad peak (Full Width at Half Maximum (FWHM) ~1 nm) is found at ~579.4 nm, which was assigned to species δ . The spectrum of sample C2 is dominated by the same species δ , but is slightly shifted to lower excitation wavelengths. The sample with the highest recrystallization rate C3, however, shows a significantly broader peak (FWHM ~ 1.6 nm), as well as a more strongly blue-shifted spectrum. A weak peak at 578.8 nm is assigned to species γ . The shoulder with higher excitation wavelength shows the presence of species δ , and the intensity at lower excitation wavelength points to one or more species present, which are not clearly resolved here. A general trend of more strongly blue-shifted spectra for the calcites with higher recrystallization rate is visible. One month. After one month the different progress in the reactions becomes more apparent. For C1 the excitation spectrum shows only minor changes. A slight shoulder is now visible at lower excitation wavelength which may indicate the presence of small quantities of species γ , but the speciation remains dominated by species δ . The same applies to C2, but weak peaks indicate contributions by species γ and species β , visible at 578.4 nm. The spectrum of C3 shows more clearly resolved peaks now, dominated by species γ . Both, the shoulder at the spectral range of species δ , and the shoulder at lower excitation wavelength decrease. At 578.2 nm a small peak becomes visible, which is attributed to a fourth species α . Two months. Due to the slow progress in the reaction of C1 from one week to one month, only C2 and C3 were probed after two months. The excitation spectra exhibit significant differences between C2 and C3. The excitation spectrum of C2 again shows a broad distribution with the maximum at species δ . Species γ is present, but because of the stronger blueshift of the whole spectrum species γ is not predominant. In addition, species β is recognizable as a small, but well-defined peak. The different reaction kinetics comes out clearly at this point. The excitation spectrum of C3 now shows well defined peaks for species α and γ , as well as a shoulder corresponding to species δ . The shoulder is decreased in intensity compared to one month earlier and appears to be divided into two peaks. Long-term experiments. Additional long-term experiments were conducted with C1 and C2, after 450 d and 1150 d, and 380 d, respectively to probe their slower reaction kinetics. The excitation spectrum of C1 is still broad and essentially featureless after 1.5 years (450 d), but starts to exhibit more clearly resolved features after ~3 years (1150 d). The speciation is still dominated by species δ , but the shoulder corresponding to species γ is now clearly visible. Also a small shoulder may be present, which would indicate the formation of species α . As expected the reaction proceeds quicker for C2. After one year (380 d) the excitation spectrum of C2 shows four distinguishable peaks. The dominating species is now γ , similar to C3 at the previous time steps. Species β is characterized by a sharp, narrow peak, while species δ shows a differentiation into two features similar to the same species observed for C3 after 2 months. ## Species characterization. Four unique species were identified throughout the reaction, which can be characterized by their emission spectra and fluorescence lifetimes. The emission spectra after direct excitation of the four different species are shown in Fig. 4, fluorescence lifetimes and additional characteristics are given in Table 2 (for fluorescence decay profiles see Figure S2 in SI). Species α was identified in C3 after 30 and 60 days, and may also be present in C1 after ~3 years. The 7 F 1 band shows a 3-fold splitting indicating a low symmetry 12 , despite the fact, that the 7 F 2 band exhibits only a 3-fold splitting. Ligand field considerations dictate that a full splitting of the 7 F 1 band must be accompanied by a full splitting (5-fold) of the 7 F 2 band, which is likely not sufficiently resolved in our spectra, as the peaks are noticeably broadened. The long lifetime of 3700 ± 350 μ s corresponds to the complete loss of the hydration of Eu 3+ 33 and therefore with an incorporation of Eu 3+ into the calcite crystal bulk. Species β has its excitation maximum at 578.4 nm and appears only in material C2 as a shoulder after one and two month and as a sharp and narrow peak after more than one year of reaction time. The corresponding emission spectrum reveals a 3-fold splitting in the 7 F 1 band and a 5-fold split 7 F 2 band, this maximum splitting pattern implies a low symmetry of the ligand field surrounding Eu 3+ . The lifetime of 2450 ± 150 μ s here also points at complete loss of hydration in the first coordination sphere of Eu 3+ and therefore an incorporation into the crystal bulk. Comparison of emission spectrum and lifetime clearly identifies β as species B, which had been previously identified in coprecipitation experiments 3,29 . Species γ is unique to the incorporation process close to equilibrium, and was identified in all 3 calcite materials. Its excitation maximum was found at 578.8 nm. To interpret its emission spectrum we have to consider an apparent overlap with species δ . This results from the broadness of the excitation peak of species δ and the small distance between the two excitation maxima. While no single species emission spectrum could be obtained, it appears a 3-fold 7 F 1 and a 5-fold 7 F 2 band can be determined, so site symmetry of species γ is low as well. Due to the inherent overlap of the emission spectra, additional interpretation of the observed splitting pattern is however not expedient. The same problem occurs in determination of the lifetimes of both species. We determine the respective lifetimes, where either species is most dominant, and we can be most confident in the determined value, but overlap remains an issue (see also discussion of lifetimes of species δ ). Despite these concerns, a lifetime of 3750 ± 450 μ s clearly shows a water free coordination shell of Eu 3+ , and hence bulk incorporation. Species δ also occurs in all three calcites and dominates speciation of all samples early on, before being successively replaced by species γ , as well as α or β . The excitation maximum is always broad within the range of 579.2-579.5 nm. In the last spectrum of C2 (380 d) and C3 (60 d), respectively, the excitation peak is divided into two distinct peaks, but neither the emission spectra nor the lifetimes exhibit significant differences, upon excitation at either of the two maxima. The full splitting of 7 F 1 and 7 F 2 indicates a low symmetry. As mentioned above determination of the lifetime of δ is difficult, due to its broad excitation range, and thus overlap with γ , and therefore is most reliable for early samples. Here we find lifetimes of ~700 ± 50 μ s, corresponding to ~0.9 ± 0.5 H 2 O in Eu's first coordination sphere. Species δ therefore can be identified as an Eu inner-sphere sorption complex on the calcite surface, the expected starting point of the metal/mineral interaction. However, with increased reaction time we determine continuously longer lifetimes up to a value of ~ 2300 ± 150 μ s, which would imply no water molecules surrounding Eu 3+ , with no apparent changes in the emission spectra. It seems unlikely that an incorporation species would form so rapidly, and with such predominance even in the otherwise slowly reacting calcites C1 and C2. Therefore, we suggest the long lifetime occurs for different reasons. First of all, it is likely that strong overlap with the long-lived species γ contributes significantly to an apparently longer lifetime of δ , especially as γ becomes more dominant with longer reaction times. It is also possible that with the advancing recrystallization reaction more CO 3 2− becomes available in the near-surface solution, which could potentially substitute the coordinating water molecules to form a ternary surface complex as an intermediate step to full structural incorporation. Such a substitution could also potentially cause the observed splitting of species δ 's broad excitation peak into two narrower (yet still strongly overlapping) peaks. Despite the uncertainties regarding species δ , it is evident that this species is not identical to the incorporation species C on the symmetrical Ca 2+ lattice site 3 , which had been previously identified in our co-precipitation experiments with a similar excitation wavelength (λ exc (C) = 579.6 nm). Charge compensation mechanism. Incorporation species γ had not been previously identified in our studies under growth conditions. However, those studies had shown that structural incorporation of Eu 3+ into calcite requires co-substitution with Na + for charge compensation. With the fastest reacting calcite C3 we conducted a recrystallization experiment in which we replace NaCl by KCl as background electrolyte to test whether the same charge compensation mechanism plays a role under recrystallization conditions. The excitation spectra of C3 reacted with both electrolytes are displayed in Fig. 5. The speciation of both systems is dominated by species γ , but the spectrum of the NaCl system is strongly blue-shifted compared to the KCl system. As the most strongly red-shifted species δ was identified as a sorption species, and all three more blue-shifted species (α -γ ) are incorporated into the bulk, this indicates enhanced incorporation in the NaCl system relative to the KCl system. This may indicate that also under recrystallization conditions coupled substitution of Eu 3+ and Na + for two Ca 2+ is required for incorporation. Nevertheless, incorporation remains the most significant interaction mechanism in the KCl system, likely due to a considerable amount of naturally occurring Na in calcite C3 (for impurity contents see Table S1 in SI). Also results of ICP-MS measurements show similar Eu content in both systems, NaCl and KCl (see Table S4 in SI). The suggestion of Vinograd et al. 30 , that a dolomite like structure forms through coupled substitution cannot be verified here, as no species with a point symmetry of D 3d was found in this study. ## Discussion Our study demonstrates that interaction of Eu and calcite will result in significantly different speciation depending on whether the interaction occurs under growth conditions, or via recrystallization. Moreover, in the latter case speciation will also be impacted by the recrystallization rate, with a clear trend towards increased incorporation at faster recrystallization. The initial step of the reaction is a Eu 3+ inner sphere sorption species (species δ ) on the calcite surface. Eu 3+ is initially still in contact with water, which presumably is substituted by CO 3 2− in a ternary surface complex, before Eu is buried under more crystal layers to become structurally incorporated. The main product of the incorporation reaction is none of the species A, B, or C that were previously identified after co-precipitation of calcite in the presence of Eu 3 , but instead the new species γ was identified and characterized. The site occupied by species γ is of low symmetry and bears little resemblance to the trigonal Ca site. The newly identified incorporation mechanism must then result either in a significantly stronger distortion induced in the host crystal's lattice, or Eu occupies a non-crystallographic site. We can exclude that the newly identified species is incorporated on a crystallographic site of a NaEu(CO 3 ) 2 -dolomite domain, based on symmetry considerations. One could speculate that the structural control of a local disturbance in the lattice is stronger in the slow process close to equilibrium, than in the faster reaction upon mineral growth. For example, Eu's preference for specific sorption sites on the calcite surface could direct its final location in the lattice. During growth of calcite preferred sorption of cations would be expected to occur in crystallographic locations (otherwise an amorphous material would form), while in the case of recrystallization sorption may be more likely at e.g. kink sites, step edges, or other similar surface defects. Thus, Eu would be found in a distorted or non-crystallographic site from the initial sorption step onwards. We observed two other low symmetry incorporation species, species α (C3) and species β (in C2). Species β has been observed in our previous co-precipitation experiments (as species B), but is less defined here than in the co-precipitation experiments. Species α was hitherto unknown and is always a minor species where it occurs at all. Species β only occurs for sample C2, and only after more than one year of reaction. Here, its corresponding peak is relatively strong and well defined, however. It remains unclear what the mechanism of formation of these minor species is, and whether they are transitory species, or will remain stable. At least species β only occurs in significant amounts after major contributions of γ have been found, and γ is not significantly reduced where it occurs, so evidently the two species form independently. The effect of coupled substitution (Eu 3+ + Na + ↔ 2 Ca 2+ ) has been described before 3 and is to some extent also observed for our incorporation species. Our results comparing NaCl and KCl as background electrolyte show a strong blue shift (i.e. towards incorporation species) in the excitation spectra of the NaCl system in comparison to the KCl system. C3, the calcite used for this experiment, contains a considerable amount of Na, so coupled substitution can still proceed here as well, but to a reduced extent. It is apparent that the formed incorporation species γ is reproducible throughout our experimental series, though the kinetics of its formation vary significantly. The higher the recrystallization rate, the faster the process of incorporation proceeds, and the more dominant species γ appears. Based on ICP-MS analysis of C1 − 3 it is also possible that the increased concentration of impurities plays a role in the incorporation process. Yet, it appears counter-intuitive to assume a high content of impurities would directly increase the potential for uptake of other contaminants by calcite. There may well be an indirect effect of the impurities though, either by providing a charge compensation pathway, or by increasing the recrystallization rate, in agreement with the overproportional release of Na from C3 in a dissolution experiment (see Table S2 in SI). ## Conclusion Our findings conclusively show that Eu 3+ is incorporated into calcite under recrystallization conditions. The observed speciation is different from that found in co-precipitation experiments and will depend on calcite's recrystallization rate or reaction kinetics. After sufficiently long time speciation is dominated by species γ , independent of the time required to reach this point. Species γ is incorporated, but not on a crystallographic site. It is likely that the high content of rare earth elements, which is found in natural calcites is present predominantly in form of species γ , where incorporation occurred close to equilibrium. In addition, this means that calcite will act differently as a retention barrier, e.g. in a nuclear waste disposal site, depending on whether it is present as a component of the host rock formation and thus close to thermodynamic equilibrium, or formed later on as a secondary phase. While a driving force for incorporation clearly exists under the recrystallization conditions studied here, it remains unclear whether species γ is more or less stable than species C, which is described as incorporation on the Ca lattice site. A conclusive answer to this question will likely have to rely on theoretical calculations. The charge compensation mechanism appears to be unaffected by the type of incorporation species formed. As previously found for species B and C, species γ as well is influenced by the background electrolyte. This suggests, that even though we cannot assign a crystallographic site to this species, its formation must involve substitution of Ca 2+ by Eu 3+ . ## Experimental Characterization of the solid phases. Calcite from Merck (calcite C1), natural calcite single crystals purchased from Ward's Science USA (calcite C2) and industrial calcite from Solvay (calcite C3) were used in this study. Materials C1 and C3 were used as delivered, while C2 was ground and sieved to grain size < 63 μ m with a ceramic mortar. The three calcites were characterized by powder X-ray diffraction (XRD) (see Figure S1 in SI). The specific surface areas (SSA) were determined by applying the Brunauer-Emmett-Teller (BET) equation with nitrogen adsorption isotherms at 77 K (Multi-point Beckman Coulter surface analyzer SA 3100). The total organic carbon content was determined by a multi N/C 2100 (Analytik Jena AG). To characterize the surface morphology of the calcite samples, scanning electron microscopy (SEM) was performed using a S-4800 microscope (Hitachi) operated at an accelerating voltage of 10 kV. Samples were spread homogenously over a silicon wafer and measured at room temperature. The amount of impurities was determined by ICP-MS (inductively coupled plasma mass spectrometry, ELAN9000 Perkin Elmer) after dissolution of the material in nitric acid. Reagents. The Eu 3+ stock solution was prepared by dissolving EuCl 3 ⋅6H 2 O (Sigma Aldrich p.a.) in Milli-Q water (18.2 MΩ cm −1 ). The concentration of this stock solution, 1 × 10 −4 mol L −1 , was confirmed by ICP-MS. All experiments were then carried out using diluted fractions of this solution. 0.01 mol L −1 NaCl and KCl solutions were prepared from a Merck powder (p.a.). ## Calcite recrystallization rates. For the determination of recrystallization rates of calcite we applied a method previously shown by Berner et al. 32 . For this, the radioactive tracer 45 Ca was used to measure the uptake of calcium ions and calculate the amount of calcite recrystallized. The formula developed by Berner et al. 32 was also used for this study: ) tot t 45 0 45 with n(t) as the time-dependent molar amount of recrystallized calcite, the total solution volume V, 45 Ca concentrations at t = 0 and t, the decay constant λ = ln(2)/t 1/2 of 45 Ca radioactive decay and reaction time t. Two series of experiments with different solid-to-liquid ratios were conducted to determine recrystallization rates of the three calcite powders considered in this study. Firstly, a high ratio of 20 g L −1 was chosen for better comparability with Berner's data. Due to the high recrystallization rates of calcites C2 and C3, equilibrium of the exchange reaction was reached after several hours and no rate determination could be conducted. Therefore, we performed a second experimental series with a lower solid-to-liquid ratio of 2.5 g L −1 . The suspensions for both series were pre-equilibrated for one day and spiked with 15.5 MBq 45 Ca as CaCl 2 in 600 μ L deionized water and shaken constantly in an overhead shaker. Sampling was performed within the first five days by removing 50 μ L of solution and measuring the activity of 45 Ca by liquid scintillation counting (Wallac WinSpectral, Ultima Gold LSC Cocktail, Perkin Elmer). The rates were determined in the linear regime by fitting the slope of the amount of recrystallized calcite as a function of time. Since the exchange reaction exhibits two time-dependent phases with distinct rates, only the first rate was determined in this study, being more sensitive to experimental conditions. ## Sorption experiments. Prior to the sorption experiments, all three calcites were equilibrated in 0.01 mol L −1 NaCl under atmospheric CO 2 until a pH of 8.3 ± 0.1 was reached. Afterwards, supernatants were collected by filtration or centrifugation at 6,800 g for 1.5 h (Avanti J-20 XP Beckman Coulter). 125 mg of each fresh calcite were suspended in 50 mL of these pre-equilibrated supernatants (in polypropylene tubes) to obtain a final solid-to-solution ratio of 2.5 g L −1 . Required amounts of the Eu 3+ stock solution were added to reach a final concentration of 10 −6 or 5 × 10 −7 mol L −1 . The pH of the calcite suspensions was regularly checked and was found to be constant at 8.4 ± 0.1. All pH measurements (pH-meter Inolab WTW series pH720) were performed using a combination glass electrode (BlueLine 16 pH from Schott Instruments) in which an Ag/AgCl reference electrode was incorporated. The pH values were measured to an accuracy of ±0.05. Electrodes were calibrated using two NIST-traceable buffer solutions (pH 6.87 and pH 9.18 from WTW). All experiments were performed at room temperature with a contact time of up to 3 years. After varying intervals, samples were centrifuged during 2 hours at 6,800 g and the remaining europium concentration in the supernatant was determined by ICP-MS. The difference to the initial europium content provided the amount of sorbed Eu 3+ . The calcite samples were dried at room temperature and subsequently analyzed by site-selective TRLFS. Their content in Ca and Eu at the end of the sorption experiments was determined by ICP-MS after dissolution. For the calcite C3, additional similar experiments were performed using calcite saturated solutions in 0.01 mol L −1 KCl. Site-selective TRLFS. The solid samples collected at the end of the sorption experiments were cooled down in a cryostat chamber at ultra-high vacuum (10 −6 -10 −7 mbar). Three laser systems were used for these studies. The description of two systems can be found elsewhere 29 . For the third one, a Nd-YAG (Continuum SL I-20, Continuum, San Jose, USA) pumped dye laser (NarrowScan K, Radiant Dyes Laser & Accessories GmbH, Wermelskirchen, Germany) with Pyrromethene 580 (Radiant Dyes) as lasing dye was used. Eu speciation was determined by adjusting the laser's emission wavelength in steps of down to 0.01 nm through the spectral range of the non-degenerate ( 5 D 0 → 7 F 0 ) transition (575-582 nm). For each step the integrated fluorescence intensity is recorded. As the transition is non degenerate (J = 0 for both states), the number of peaks in such an excitation spectrum corresponds directly to the number of Eu 3+ species in the system. Subsequently, these species can be excited directly to obtain single species, emission spectra and fluorescence lifetimes so far as spectral overlap permits. All site-selective TRLFS measurements were carried out at low temperatures (< 10 K), in order to increase the spectral resolution. A spectrograph (Shamrock 303i, Andor Technology Ltd., Belfast, UK) with polychromator grids (300, 600 and 1200 lines per mm) and intensified CCD detector (ANDOR iStar 734 cooled to −20 °C to reduce thermal noise effects) were used to record the luminescence emission. For lifetime measurements, the camera delay was increased gradually up to several milliseconds, until fluorescence intensity dropped to a minimum of 1/e.

chemsum

{"title": "Incorporation of Eu(III) into Calcite under Recrystallization conditions", "journal": "Scientific Reports - Nature"}

vibrational_coherences_in_manganese_single-molecule_magnets_after_ultrafast_photoexcitation

## Abstract: Single-Molecule Magnets (SMMs) are metal complexes with two degenerate magnetic ground states arising from a non-zero spin ground state and a zero-field splitting. SMMs are promising for future applications in data storage, however, to date the ability to manipulate the spins using optical stimulus is lacking. Here, we have explored the ultrafast dynamics occurring after photoexcitation of two structurally related Mn(III)-based SMMs, whose magnetic anisotropy is closely related to the Jahn-Teller distortion, and demonstrate coherent modulation of the axial anisotropy on a femtosecond timescale. Ultrafast transient absorption spectroscopy in solution reveals oscillations superimposed on the decay traces with corresponding energies around 200 cm −1 , coinciding with a vibrational mode along the Jahn-Teller axis. Our results provide a non-thermal, coherent mechanism to dynamically control the magnetisation in SMMs and open up new molecular design challenges to enhance the change in anisotropy in the excited state, which is essential for future ultrafast magneto-optical data storage devices. Single-Molecule Magnets (SMMs), molecules that show magnetic hysteresis below a certain blocking temperature 1 , show great promise for future applications in data storage devices because their small size and well-defined magnetic properties can reduce the size of data bits and therefore increase storage density. The recent observations of hysteresis loops close to 5,6 , or above 7 , liquid nitrogen temperatures (77 K) in lanthanide-based SMMs provide an important step forward, but to implement these molecules in devices methods to control the spins need to be developed. Being able to switch the magnetisation direction in SMMs using femtosecond laser pulses could provide the technology for future ultradense memory devices operating on unprecedented timescales. Because the slow magnetic relaxation in SMMs means that they can retain the magnetisation direction for months 1 , spin-switching in SMMs is advantageous over spin-crossover (SCO) switching in paramagnetic Fe(II) complexes, even though SCO is well studied and known to occur on ultrashort timescales . Light-induced SCO can result in SMM properties in some crystals but on timescales from minutes to hours 13,14 , which is too slow for applications. Despite this potential, there are a limited number of studies of ultrafast dynamics in molecule-based magnets. For example, ultrafast charge-transfer dynamics 15 , phase-transitions 16 and intersystem crossing 17 have been studied in magnetic Prussian blue analogues, and spin state switching has been observed in Cu(II)-based breathing crystals 18 . Magnetic nanotoruses have been studied using transient absorption (TA), identifying lanthanides as trap states for excitons 19 . In this work, we have developed a detailed understanding of the ultrafast photophysics of Mn(III)-based SMMs and prove that coherent vibrational wavepackets modulate the Jahn-Teller (JT) axis, and therefore the magnetic anisotropy, on an ultrafast timescale. Manganese-based coordination complexes, such as Mn12 20,21 , have been instrumental in the development of SMMs and are typically magnetically bistable due to a large, negative zerofield splitting caused by the magnetic anisotropy of individual Mn(III) ions. The 3d 4 electron configuration leads to an elongation or compression of the axial bonds via the JT distortion. Together with the spin-orbit interaction, this leads to two degenerate magnetic ground states where the ground state spin is saturated either parallel or anti-parallel to the magnetic easy axis. Transiently reducing the anisotropy in SMMs, using femtosecond laser pulses, could provide a method towards achieving optical control of their magnetisation, making use of quantum coherences that can be created using ultrashort laser pulses 22 . The prototype Mnbased SMM Mn12Ac (Ac = acetate) is promising for exploring optical modulation of the anisotropy because it has been shown that reorienting the JT axis with high pressure can strongly influence the molecules magnetic properties 23 . Similarly, there have been reports of ultrafast spin-switching of Cu(II) molecule-based magnets due to optical modulation of the JT axis 18,24 . However, said species are large and structurally complex, with smaller molecules being better suited to achieving a more detailed understanding of the photophysics. One such example is a family of oxime-based SMMs containing three or six Mn(III) ions whose magnetostructural relationship has been extensively investigated . Herein, we present ultrafast transient absorption spectroscopy of [Mn(III)3O(Et-sao)3(b-pic)3(ClO4)], or "Mn3", which has three high-spin Mn(III) ions arranged in a simple triangle 26 (Figure 1A) and [Mn(III)6O2(Etsao)6(O2CPh(Me)2)2(EtOH)6], or "Mn6", which contains six high-spin Mn(III) ions arranged in two triangles 30 (Figure 1B). We show that photoexcitation leads to a change in the JT distortion and that a coherent wavepacket is launched towards the new equilibrium bond length on the excited state potential energy surface for both Mn3 and Mn6. The structural rearrangements and a dephasing of the vibrational wavepacket take place on a sub-ps timescale. The excited electronic state decays back to the ground electronic state with a decay constant of 8 -9 ps for both molecules studied. These measurements reveal, for the first time, the possibility to coherently control both the anisotropy and the exchange interaction in SMMs on the femtosecond timescale. ## Static UV/ Vis absorption spectroscopy Static UV/Vis absorption spectra of both manganese complexes (Figure 2A) show almost identical absorption bands (which are also similar to the Mn12 spectrum ). In the visible region, the spectra are rather unstructured although one weak shoulder at 375 nm can be discerned. MCD measurements by Bradley et al. 28 showed that metal-centred transitions are responsible for the absorption above 410 nm. In the ultrafast TA measurements, 400 nm light was used for excitation. This spectral region is tentatively assigned to charge-transfer (CT) transitions due to the large molar absorptivity observed (Figure 2A). Ligand-centred ππ* transitions on the salicylaldoxime ligand show no absorption below 400 nm 35 . The CT transitions are most likely due to ligand-to-metal CT (LMCT), transiently reducing the Mn(III) to Mn(II). However, due to the low symmetry of the molecules, the degeneracy is completely lifted and metal-centred transitions with non-zero extinction coefficients are also expected 36 28 32 . We therefore assign the 375 nm shoulder to a mixture of LMCT transitions and spin-allowed d-d transitions, although CT transitions will most likely dominate due to the larger oscillator strength. ## Transient absorption In Figures 2B -D, the difference spectra for selected time delays of the TA are shown. The difference spectra for both Mn complexes are similar. Two positive bands, due to excited state absorption (ESA), are observed around 350 nm ("UV-band") and 430 nm ("Vis-band"). The Vis-and UV-bands are broad and are probably comprised of several unresolved peaks. In fact, the Vis band can be seen to separate into two bands at longer time delays with maxima at ca. 420 and 500 nm. However, the decay kinetics are very similar for the two sub-bands and because they cannot be resolved at earlier times, we treat them as one ("Vis band"). The maxima of both the UV-and Vis-bands are shifted towards shorter wavelengths for longer time delays for both molecules. Furthermore, the maximum of the Vis-band of Mn6 is blue shifted in relation to Mn3 for pump-probe delays longer than 0.5 ps. Kinetic traces of the TA measurements at two different wavelengths for both Mn complexes are shown in Figure 3 (further kinetic traces can be found in Figure S2). The 440 nm kinetic trace shows a fast decay during the first few ps and a slower decay back to zero in ~20 ps. This is observed for both Mn3 and Mn6. For the 345 nm trace, a very fast initial rise of the absorption change is observed for Mn3 (Figure 3E). The rise is not observed at 345 nm for Mn6 (Figure 3G), although can be discerned at shorter wavelengths (Figure S1B and S2E). After reaching the maximum absorption change, the UV-bands for both molecules decay back to zero on a similar timescale as the Vis-bands. The Glotaran software 37 ## Wave packet dynamics We clearly observe oscillations superimposed on the transient absorption signals during the first picosecond across the two ESA bands (Figure 3E -H). In Figure 4, averages over a small wavelength region, corresponding to ca. 1000 cm −1 , of the residues from the global triexponential fit of the Vis-and UV-bands are shown. The Fourier analysis of the fit residues are shown in Figure 5A. We found that the frequency spectrum is mainly composed of one dominant peak for both Mn complexes, with peaks centred at 172 cm -1 (177 cm -1 ) for the Mn3 Vis-band (UV-band) and 189 cm -1 (201 cm -1 ) for Mn6 Vis-band (UV-band). We have also performed Raman spectroscopy on crystal grains of the molecular solids to compare the vibrational spectra to the oscillations observed in the transient absorption. The resulting spectra are shown in Figure 5B. The Raman spectra for the two molecules are rather similar, although the peaks are somewhat broader for Mn6. However, some of the peaks are shifted, and interestingly, the largest shift is for the intense peaks at 213 cm −1 and 191 cm −1 for Mn3 and Mn6, respectively. These are in the same region as the ones found in the Fourier analysis of the TA data. We also performed a DFT analysis to calculate the Raman spectrum of Mn3 (Figure 5C). The calculated Raman spectrum agrees well with the measured one. In particular, the peak position in the calculated spectrum at 210 cm −1 fits very well to the measured peak at 213 cm −1 . This mode can be described as a collective in-phase asymmetric stretch mode along the JT axis for all three Mn ions (a video can be found in the SI), where the Mn-O bonds are contracted while the Mn-N bonds are extended and there is a flattening of the Mn triangle. ## Discussion The static absorption spectra of the two complexes show similar bands, which indicates that the optical transitions are to some extent localised, such as LMCT between O/N and Mn or d−d transitions, as previously discussed. The TA results are also similar, which is perhaps not surprising if the excitation is dominated by CT between specific Mn−O/N sites in the molecules. From the transient absorption data for both complexes we extract three time-constants in a sequential decay model. The first time constant is faster than 200 fs for both molecules and the corresponding DAS for this timescale are significantly different to the DAS for the two other timescales (ca. 1-2 and 8-9 ps), both of which show similar spectra with the characteristic UV-and Vis-bands. The A1 spectrum in Figure 2E therefore corresponds to the absorption from the initially excited LMCT state, which absorbs predominantly in the visible spectral region. The lack of absorption in the UV region allows for the GSB to be observed in this short timescale. The A2 and A3 DAS are different to A1, which implies that there is a change in the electronic character of the excited state 38 and therefore an intermediate state is populated on a <200 fs timescale through internal conversion (IC) either within the LMCT manifold or via back-electron transfer into a metal-centred excited state. We argue below that key to interpreting our results comes from realising that the molecules have a strong JT distortion, which introduces axial anisotropy in the system, perpendicular to the plane of the triangle formed by the three Mn(III) ions (Figure 1). One can therefore envisage ESA transitions whose transition dipole moments lie either equatorial (in-plane) or axial (out-of-plane) with respect to the triangle (or triangles in the case of Mn6). The bond length between the Mn and the axial oxygen, connecting the two triangles, is 243 pm. This is much longer than the equatorial Mn−O bonds of 190 pm. The bond strength is therefore considerably weaker along the JT axis than the in-plane bonds. A CT transition into the dorbitals will induce a change in the charge distribution in the molecule and consequently the JT distortion will be released, which was recently observed in perovskite manganites . The release of the JT distortion via the mode identified by the DFT calculations causes a contraction of the axial Mn-O bonds. However, because the equatorial bonds are more restricted due to the larger bond strengths, we observe a larger change in the bond length along the JT axis. Because of the change in the bond lengths, ESA transitions will be affected by the different ligand-proximity to the Mn centres, and therefore the oscillator strength of the ESA will be modulated. Importantly, this effect will be larger for axial transitions than for equatorial transitions, since the bond length change is smaller in the latter case. There are several experimental observations that point towards a large motion along the JT axis. By comparing the Mn3 and Mn6 results, we can gain some insights into any out-of-plane dynamics because the second triangle in Mn6 enforces some constraints on the axial motion in either triangle (Figure 1). The Vis-band peak maximum in the DAS for Mn6, for both A2 and A3 (Figure S1B), is blue-shifted with respect to the peak positions in the Mn3 DAS (Figure 2E) by 290 cm −1 for A2 and 260 cm −1 for A3. In contrast, there is no observable change in the peak position for the UV-bands. We therefore argue that the Vis-band arises from ESA transitions with axial transition dipole moments whereas the UV-band arises from transitions with equatorial transition dipole moments. We observe in the kinetics for both molecules that the magnitude of the change is larger for the Vis-band than the UV-band, as seen in Figure 3. This therefore also points towards a much larger change in bond lengths for the axial bonds than the equatorial bonds, due to the dependence of the oscillator strength on ligand-proximity. We therefore conclude that the JT distortion is indeed relaxed after exciting at 400 nm. After the sudden release of the JT distortion, a fast motion out of the Franck-Condon region takes place towards the new bond length in the excited state. The fast motion occurs via the formation of a wavepacket involving quanta of a specific vibrational mode. The wavenumber of the wavepacket coincides with the observed strong Raman modes around 200 cm −1 in Figure 5B, although it is somewhat smaller because of weaker bonds in the excited state. The peaks at 200 cm −1 in the Raman spectra are strongly shifted for the two molecules, in contrast to the other Raman peaks in the spectra (Figure 5B). There is therefore a difference in vibrational energy when there is one (Mn3) or two (Mn6) triangles in contact with each other (Figure 1). Therefore, it can be concluded that the vibrational motion involves the out-of-plane JT axis. The other peaks in the Raman spectrum are localised to Mn−O or Mn-N bonds, where it does not matter if there is another Mn triangle in the proximity. The DFT calculations do indeed demonstrate a collective in-phase motion of the Mn ions along the JT axis (Figure 6B and supplementary video). The dephasing of the wavepacket takes place on a longer timescale (ca. 300 fs) than the lifetime of the initially populated LMCT state and therefore survives the IC into the intermediate electronic state. The 1-2 ps decay constants can be assigned to vibrational cooling via intramolecular vibrational redistribution (IVR), which leads to the observed dephasing of the wavepacket. The vibrational cooling can be inferred from the spectral narrowing and blueshift in the absorption peak, as observed in Figure 3. The faster LMCT state decay and IVR in Mn6 is due to a higher density of states than in Mn3. Since the DAS for the 8-9 ps decay is similar to the 1-2 ps DAS, we argue that this corresponds to the decay of the vibrationally cool intermediate state back to the ground electronic state via IC. The photophysics model is summarised in Figure 6. ## Conclusions In summary, we have performed transient absorption spectroscopy of Mn3 and Mn6 SMMs in ethanol at room temperature. We observed two characteristic ESA bands at ca. 345 and 430 nm after exciting with 400 nm light. We found oscillations superimposed on the decay signals, which were attributed to a vibrational wavepacket. By comparing the peak position of the ESA, decay dynamics, and experimental and calculated Raman spectra of the two molecules, we concluded that ESA transitions with either axial or equatorial transition dipole moments give rise to the Vis-and UV-bands, respectively. Two ESA bands could be fitted with the same time constants but showed different sensitivity to the ensuing structural dynamics in the excited state. The observed differences in the results for Mn3 and Mn6 allowed us to conclude that there is a sudden release of the JT distortion after LMCT excitation and this change gives rise to the vibrational wavepacket. The wavepacket dephases on a 300 fs timescale by coupling to other vibrational modes via IVR. Based on the global fit analysis, we conclude that the LMCT state decays into an intermediate state in less than 200 fs. This state in turn decays back to the ground electronic state with a 8 -9 ps time constant. Our experimental strategy of comparing molecules with either one or two Mn triangles allowed us to measure the JT distortion dynamics after photoexcitation of two Mn-based SMMs and find that the motion is governed by a coherent vibrational wavepacket. Design strategies for achieving ultrafast coherences in complex chemical systems for control and functionenhancement are actively being developed 22 and there have recently been several interesting observations in metal complexes, such as retention of vibrational coherence during intersystem crossing in Cr(acac)3 42 , metal-metal bond modulations in di-Pt(II) complexes 43 and in particular it has been shown that JT distortions, such as in Cu(I) complexes 44 , are important. Our work contributes to this field by demonstrating vibrational coherences in a molecule with as many as six metal ions. There is a peak in the Raman spectrum of the Mn12Ac 45 SMM at 209 cm −1 and so it is possible that our approach is more general, which is supported by the observation of light-induced magnetisation changes in another Mn12 complex using continuous irradiation 46 . Our results therefore open up the possibility to study and control coherent magnetic interactions on the femtosecond timescale in a large range of SMMs. ## Materials and Methods The synthesis of the complexes has been described previously in Ref. 25 for Mn6 and in Ref. 26 for Mn3. For the transient absorption measurement, the Mn6 and Mn3 complexes were dissolved in ethanol. The concentration of the Mn6 solution was 1.8 x 10 -3 mol/l and the concentration of the Mn3 complex in ethanol was 1.88 x 10 -3 mol/l. A Starna flow cuvette with 0.2 mm pathlength was used for the TA measurements with a flow of 8 μl/min. The TA setup is based on the apparatus described in ref. 47 . As the pump beam, the second harmonic (400 nm) of a Coherent Legend Elite laser was used (pulse duration 120 fs and output 800 nm wavelength). The pump pulses were focused into the sample by a f = −500 mm concave mirror producing a spot size of 226 microns (1/e 2 ). The laser fluence was 3.3 mJ/cm 2 . For the probe and reference beams 1.4 μJ/pulse of the 800 nm fundamental was focused with an f = 100 mm fused silica lens in a 5 mm thick CaF2 plate, which was continuously moved in two dimensions, to produce a broadband white light continuum. The white light was collimated with an f = −100 mm concave mirror and the 800 nm fundamental was removed with a 720 nm cut off filter. The detected probe spectrum ranged from 320 to 720 nm. The white light beam was divided with a reflective metallic neutral density filter for probe and reference. The probe light was focused into the sample with an f = −500 mm concave mirror. The probe beam diameter in the sample was 105 μm (1/e 2 ). To avoid anisotropic signals, the pump-probe polarization angle was set to 54.7° ("magic angle"). For controlling the time delay between pump and probe, a delay stage with mounted retroreflector was used. For each time delay 1000 spectra were collected. The whole procedure was repeated five times to get 5000 spectra in total for each delay position. A prism was used to disperse the white-light beams onto two fast CCD cameras from Entwicklungsbuero Stresing equipped with Hamamatsu S7031-0906 sensors with 512x58 active pixels. Full binning was used, where the 58 vertical pixel were binned, which allowed a synchronous read-out at 1 kHz for both probe and reference beams. Raman spectroscopy of crystal grains of the Mn3 and the Mn6 complexes was performed on a Renishaw Raman microscope with a laser wavelength of 785 nm. The geometry of [Mn(III)3O(Et-sao)3(b-pic)3(ClO4)] has been optimised using the unrestricted DFT approach with the Perdew-Burke-Ernzerhof (PBE) functional 48 and the DKH-recontracted def2-SV(P) basis set 49 in the ground state with the Orca package 50 . Relativistic effects were taken into account within the Douglas Kroll Hess at second order (DKH2) approach as implemented in Orca 54 . The frequency analysis of the optimised structure in the ground state of spin multiplicity Ms=13 showed one imaginary frequency (−3.93 cm −1 ). The corresponding normal mode involves the rotation of the phenyl groups around the N-Mn bond which is expected to occur along a flat potential energy curve. We therefore consider the geometry to be close enough from its minimum to conduct further studies. FL performed the optical experiments and analysed the data and RMcN synthesized and characterised the samples under the supervision of RI. JE and TJP carried out the DFT calculations. FL, EKB and JOJ conceived the experiments and interpreted the results. FL and JOJ co-wrote the paper. All authors discussed the results and commented on the manuscript.

chemsum

{"title": "Vibrational coherences in manganese single-molecule magnets after ultrafast photoexcitation", "journal": "ChemRxiv"}

highly_luminescent_hetero-ligand_mof_nanocrystals_with_engineered_massive_stokes_shift_for_photonic_

## Abstract: An high efficiency emission with a massive Stokes shift is obtained by fluorescent conjugated acene building blocks arranged in nanocrystals. The two ligands of equal molecular length and connectivity, yet complementary electronic properties, are co-assembled by zirconium oxy-hydroxy clusters, generating highly crystalline hetero-MOF nanoparticles The fast diffusion of singlet molecular excitons in the framework, coupled with the fine matching of ligands absorption and emission properties, enables to achieve an ultrafast activation of the low energy emission by diffusion-mediated non-radiative energy transfer in the 100 ps time scale, by using a low amount of co-ligands. This allow to obtain MOF nanocrystals with a fluorescence quantum efficiency of ̴ 70% and an actual Stokes shift as large as 750 meV. This large Stokes shift suppresses the reabsorption of fast emission issues in bulk devices, pivotal for a plethora of applications in photonics and photon managing spacing from solar technologies, imaging, and detection of high energy radiation. These features allowed to realize a prototypal fast nanocomposite scintillator that shows an enhanced performance with respect to the homo-ligand nanocrystals, achieving benchmark. values which compete with those of some inorganic and organic commercial systems. The Stokes shift is an important property of luminescent materials, defined as the energy difference (E) between the absorption band maximum and the emission spectrum maximum frequencies. 1 The value of E is a key parameter in photonic devices and applications because, at a first approximation, it enables to estimate if a specific emitter would be affected by significant reabsorption of the generated light. For example, if the E value is lower or similar to the bandwidth of the absorption and emission spectra, the consequent intrinsic extensive 'inner-filter' effect can heavily limit the lighting performance of bulk photonic devices, and, in the worst cases, it can also affect the kinetics of the luminescence generation. Conversely, if E is larger than the spectral bandwidths the system can be considered a large Stokes shift emitter, with no inner filter effects (Fig. 1a). These reabsorptionfree materials are highly desirable for several applications. For example, in fluorescence imaging large Stokes shift optical probes allow to obtain high contrast images with limited excitation stray light, avoiding the use of expensive filtering component or time-consuming image post-processing. 5,6 For solar applications, large Stokes shift emitters are undoubtedly the most promising materials to realize luminescent solar concentrators without reabsorption of the condensed radiation. 7 Similarly, the sensitivity of scintillating detectors for ionizing radiation would greatly benefit from the use of fast emitters with no reabsorption 8 showing good light output intensity without effects on the scintillation pulse timing, as required by the most advanced medical imaging techniques such as timeof-flight positron emission tomography (TOF-PET) 9 and high-rate high-energy physics (HEP) experiments. The extensive recent literature in the field of semiconductor nanocrystals testifies this very actual interest on large Stokes shift emitters. In these materials, for example, the E can be tuned by doping of semiconductor with electronic impurities, 11 resulting in the appearance of intragap states from which red shifted luminescence is produced. A notable E value as large as 1 eV can be achieved, 12 but a current unsolved drawback is the slow luminescence kinetics that strongly limit their use for fast timing applications in nanosecond time scale and below. 10,13,14 Moreover, in photonic devices where fast timing is foreseen, traditional wavelength shifters exploiting radiative energy transfer cannot be employed, because of the consequent slowing down of emitted light pulse. In the search for fast emitters with remarkably large Stokes shift, we selected Metal-Organic Frameworks (MOFs), which constitute a solid platform to build materials wherein active struts perform tailored functions. Synthetic procedures based on self-assembly processes enable the controlled framing of struts in the porous crystalline architecture and the regulation of distances among linkers. 15,16 The impressive versatility of MOFs promoted several applications such as gas storage, 17,18 catalysis and dynamic materials, and triggered the most recent advances in the field of luminescent MOFs. This hot topic gave birth to a new class of optically active nanomaterials with tailorable electronic properties for photonics and optoelectronics, sensing, and biomedicine. 26,30,31 MOFs are also excellent candidates to be used in light-emitting devices, due to their structural diversity and tunable emission. A key advantage is the possibility to design their framework composition and structure which control both optical and energy-transport properties, such as those required for managing site specific photoreactions, 32 or multi-excitonic processes. 33 Therefore, optimized luminescent MOF nanocrystals can represent the next generation of luminescent materials with a potential impact comparable to their inorganic counterpart colloidal semiconductor nanocrystals. Among conjugated molecules, polycyclic aromatic hydrocarbons of acene family have attracted great interest for various application in photon managing such as photon upconversion and singlet fission because of their peculiar electronic properties. Here we present the fabrication of MOF nanocrystals with fine-tuned composition, wherein tetracene-bearing fluorescent moieties were co-assembled with anthracene-based linkers to engineer the system emission properties and obtain significant energy down conversion of the emitted photons with respect the absorption, thus maximizing the emission Stokes shift. MOFs containing linear tetracene linkers have not yet been realized so far. The strategy of increasing the number of fused aromatic rings in the ligand core, yet maintaining a constant spacing between the chelating groups, proved successful in providing a series of customized hetero-ligand Zr-MOFs, which exhibited benchmark-efficiency fluorescence accompanied with negligible reabsorption. MOF nanocrystals were obtained by co-assembling the green-fluorescent chromophore 5,12-diphenyl-tetracenedicarboxylate (DPT) and the blue-emitting ligand 9,10-diphenyl-anthracenedicarboxylate (DPA) with Zr oxy-hydroxy nodes (Fig. 1b). By exploiting the diffusion within the crystalline framework of singlet molecular excitons generated on DPA ligands, the incorporated DPT co-ligands are excited by means of non-radiative energy transfer and subsequently recombine radiatively producing photons with a E as large as 750 meV. The fine matching of frequency emission of antracene moieties with the absorption of tetracene units enables an efficient energy transfer (ET) of 97% and photoluminescence quantum yield (QY) of ~70% even with a low DPT loading of 8% with respect to DPA (denoted Zr-DPT:DPA-8%). Such a low loading enables to preserve the structural features of the parent homo-ligand nanocrystals. The potential technological transfer of the obtained hetero-ligand fluorescent nanomaterials is demonstrated by the realization of a prototypal fast polymeric nanocomposite scintillator that shows enhanced performances with respect to the homo-ligand nanocrystals, thus achieving benchmark values competing with those of several organic and inorganic commercial systems. We designed and prepared the new conjugated tetracene-containing ligand (DPT) to be coassembled with the antracene-based linker (DPA) by a solvothermal process (Methods and Supplementary Information): the two molecules DPA (QY=0.96) and DPT (QY=0.80) were chosen because of their complementary of absorption/emission properties that make them an ideal donor (DPA) and acceptor (DPT) pair for non-radiative energy transfer (Supplementary Information). The two rod-like ligands with identical end-to-end length and connectivity were co-assembled by zirconium oxy-hydroxy clusters, generating a series of isostructural hetero-ligand MOF nanocrystals with modulated composition ranging from 0.1% to 8% of DPT/DPA ligand ratio content (Zr-DPT:DPA-x%, Fig. 1b). For comparison, the homo-ligand MOFs were synthetized using separately the single ligands (Zr-DPA and Zr-DPT, respectively). The composition of hetero-ligand Zr-MOFs is in agreement with the feeding ratio, as shown by 1 H NMR of digested samples. Connectivity, purity, and thermal stability were demonstrated by FT-IR, 13 C MAS NMR and TGA analysis (Supplementary Figs. S1-S40). Scanning electron microscopy (SEM) images of the Zr-DPT:DPA specimens reveal a homogenous population of nanocrystals with octahedral morphology. Figure 2a depicts the SEM image of the Zr-DPT:DPA-8% sample, which consists in a nanocrystal ensemble with average size of 185±20 nm. Upon activation at 130°C under vacuum, the hetero-ligand Zr-MOF nanocrystals exhibit a high crystallinity and a cubic crystal structure (Fm-3m) with fcu topology, as established by PXRD Rietveld refinement, which corresponds to that of parent Zr-DPA and Zr-DPT MOFs (Supplementary Figs. S13-S17). Twelve ligands coordinate to each Zr-based node (Zr6(μ3-O)4(μ3-OH)4(CO2)12 cluster) and yield a framework containing interconnected octahedral and tetrahedral cavities (Fig. 2c). Thus, the ligands are arranged at a sufficiently short center-to-center distance of 11.7 that enables both fast exciton diffusion and non-radiative energy transfer (Supplementary Information). Consistently with the crystal structure, N2 adsorption isotherms at 77 K (Fig. 2b) showed remarkable surface areas up to 3000 m 2 /g and step-wise profiles due to the subsequent filling of the well-differentiated tetrahedral and octahedral cavities with pore size of 10.8 and 14.5 in the framework. The size, symmetry and homogeneity of the pores were probed by the highly sensitive hyperpolarized laser-assisted 129 Xe NMR (Fig. 2d). The homo-ligand Zr-DPA and Zr-DPT MOFs show a single sharp signal at =94.6 and =117.0 ppm, respectively, reflecting the patency of the cavities and the smaller size in Zr-DPT caused by the long flag-like conjugated tetracene moiety protruding into the nanochannels, with a steric encumbrance larger than that of antracene unit. The hetero-ligand Zr-DPT:DPA-8% MOF exhibits a chemical shift at =96.5 ppm which corresponds precisely to the expected weighted-average of the two chemical shifts of the homo-ligand nanocrystals. Remarkably, no residual signals of the homo-ligand MOFs are present, demonstrating the excellent structural homogeneity of the co-assembled nanocrystal ensemble. The photophysical properties of the obtained MOF nanocrystals are investigated by means of photoluminescence spectroscopy. Figure 3a shows the optical absorption and continuous wave photoluminescence spectra of the hetero-ligand Zr-DPT:DPA-1% in tetrahydrofuran dispersion (0.1, mg mL -1 ). We observe a main absorption band in the near-UV spectral matching the profile of the Zr-DPA reference sample, because the absorption of the few DPT substituents is negligible. Upon photoexcitation at 355 nm, the hetero-ligand nanocrystals show a broad luminescence in the visible spectrum. The most intense emission peaked at 515 nm and its vibronic replicas at 550 nm and 590 nm match the photoluminescence profile of the isolated DPT ligand (Fig. S7) and of the Zr-DPT MOF as a control sample. MOFs. This result suggests that the green luminescence is generated by the radiative recombination of singlet excitons on DPT co-ligands populated by energy transfer from directly excited DPAs ligands, as demonstrated by the excitation photoluminescence spectrum recorded at 540 nm that follows the DPA absorption profile (Fig. 3a). The weak residual blue luminescence peaked at 430 nm mirrors an energy transfer yield (𝜙 𝐸𝑇 ) lower than unity. Nevertheless, the presence of this residual emission is crucial to investigate the antenna effect sketched in Fig. 1b that occurs in the framework. Figure 3b depicts the normalized photoluminescence spectra of the Zr-DPT:DPA-1% dispersion at different dilution ratios. Notably, the relative intensity of the green and blue emission components is unchanged. This is a crucial result, because if DPAs and DPTs experienced radiative energy transfer as two separated entities, i.e. in a standard bicomponent solution, the 𝜙 𝐸𝑇 value should decrease following the solution dilution level that reduces the concentration of the energy acceptor DPT and therefore the transfer probability. 1,38 Consequently, the relative intensity of the green vs. blue component should be reduced as well. Conversely, the data clearly indicate that 𝜙 𝐸𝑇 is independent from the DPT dilution, thus demonstrating that each nanocrystal works as an individual emitter whose green luminescence is activated by intra-crystal energy transfer. This picture is confirmed by time resolved experiments. The top inset of Fig. 3b shows the photoluminescence intensity decay with time at 430 nm as a function of the dilution level. The signal decay in the hetero-ligand MOF is faster than in the Zr-DPA reference, indicating an efficient energy transfer (vide infra), 1 but still no change in the decay kinetic is observed at different dilutions, thus demonstrating that the DPA-DPT interaction is unaffected. Similarly, the bottom inset shows that the green photoluminescence intensity at 540 nm decays with time as a single exponential function with a characteristic lifetime of 𝜏 ~ 11 ns (Supporting Table 7) regardless of the dilution level. This is another key result, indeed, the observed lifetime is almost identical to that one of the DPT molecule (11.5 ns, Fig. S41), and is significantly longer with respect to the reference homo-ligand DPT-MOFs (7.7 ns, Supporting Table 7). These findings suggest that co-assembled DPT ligands are effectively incorporated and framed as non-interacting single molecules within the nanocrystal architecture and preserve their excellent luminescence properties pivotal for the fabrication of photonic devices. between DPA and DPT co-ligands as a function of their relative molar ratio in the MOF framework. The theoretical 𝜙 𝐸𝑇 (solid line) is calculated considering a diffusion-mediated Förster energy transfer with interaction radius of 2.8 nm derived from ligands properties. The fit of experimental data (dotted line) results a characteristic interaction radius of 2.9 nm. The top inset shows the PL spectra of the series of nanocrystals dispersions analyzed (0.1 mg mL -1 ). The bottom inset show their time-resolved PL spectra recorded at 430 nm. Solid lines are the fit of the data with multi-exponential decay function. d, Absorption (dashed line) and PL (solid line) spectrum of Zr-DPT:DPA-1% nanocrystals dispersion (0.1 mg mL -1 , optical path 1 cm). The excitation wavelength is 355 nm. The inset is a digital picture of the dispersion in the quartz cuvette under 355 nm excitation. Once assessed the validity of the synthetic strategy, we investigate quantitatively the energy transfer mechanism. As sketched in Fig. 1b, the activation of DPT luminescence occurs by energy transfer during the random diffusion within the framework of an excited DPA singlet exciton, which is created upon light absorption or free-charge recombination in scintillation. 39 Before spontaneous recombination, the singlet moves from the original position by an average diffusion length 𝐿. This implies that if a DPT moiety is placed at a distance shorter than 𝐿 from the position where the DPA exciton is created, the energy transfer can most likely occur before singlet recombination, thus without energy dispersion. Considering the MOF structure and a DPA-DPT Förster interaction radius of Rfs = 2.8 nm (Fig. S42), the theoretical energy transfer rate 𝑘 𝐸𝑇 and efficiency 𝜙 𝐸𝑇 are calculated as a function of the nanocrystal composition under the assumption of a diffusion-mediated energy process (Supplementary Material). 38 The solid line in Fig. 3c depicts the theoretical 𝜙 𝐸𝑇 vs. the DPT ligand fraction in the hetero-ligand MOFs, expressed as the nominal DPT:DPA relative molar ratio employed for the synthesis. The plot shows that an excellent 𝜙 𝐸𝑇 ̴ 0.9 (90%) is reached with a DPT content as low as 1%. This suggests that the proposed strategy to achieve large Stokes shift is already effective with low levels of DPT substitution in the DPA based MOF. In this way the risk to affect the MOF structural properties is minimized. The predicted 𝜙 𝐸𝑇 is compared with the one measured in the series of nanocrystals. The top inset of Figure 3c Given their excellent luminescence properties, we tested the hetero-ligand MOF nanocrystals as emitters in bulk scintillating devices typically employed as detector of ionizing radiation where large Stokes shift is usually required to maximize the extraction of scintillation light. 8 Figure 4a shows the radioluminescence spectrum under soft X-ray exposure of a composite scintillator (thickness 0.1 cm, diameter 1 cm) fabricated by loading Zr-DPT:DPA-8% nanocrystals in a polydimethylsiloxane (PDMS) matrix (0.5% weight, Methods, Figs. S43-S46). The radioluminescence spectrum (solid line) is dominated by a structured green emission peaked at 530 nm with a weak residual emission at 430 nm, suggesting that the scintillation light is produced by radiative recombination of DPT singlet excitons. The absence of reabsorption is demonstrated by the possibility to clearly observe the first vibronic replica in the emission spectrum at 515 nm, as in the diluted dispersion case, despite the high concentration of embedded nanocrystals (Fig. 3d). This result is in excellent agreement with the simulated emission spectrum (circles) calculated considering the light propagation in the device (Methods). Radioluminescence measurements under continuous irradiation up to around one hundred Gy demonstrate the emission stability and the absence of long-time phosphorescence due to delayed carrier recombination (inset of Fig. 4a). No significant variation of the radioluminescence intensity can be observed by heating the sample up to 50 °C, demonstrating a good thermal stability (Fig. S47). After 30 days of exposure to the atmospheric moisture, a mere 10% reduction of the emission intensity is observed, demonstrating a good resistance to molecular oxygen despite tendency of tetracene to photobleaching (Fig. S48). The nanocomposite shows a radioluminescence intensity five times greater than the reference composite made with the homo-ligand Zr-DPA (dotted line) which demonstrated a scintillation yield 𝜙 𝑠𝑐𝑖𝑛𝑡 , defined as the number of emitted photons for each MeV of deposited energy for ionizing radiation, of approximately 1000 ph MeV -1 . 39 The obtained data indicates therefore that the prototype 𝜙 𝑠𝑐𝑖𝑛𝑡 is assessed at around 5000 ph MeV -1 . This value demonstrates the success of the proposed strategy to enhance the scintillation performance of composite materials based on fluorescent MOF nanocrystals, which show a 𝜙 𝑠𝑐𝑖𝑛𝑡 comparable to that of commercial plastic an inorganic scintillators. 45 We further investigated the composite emission by time resolved photoluminescence and radioluminescence experiments as function of the temperature. At room temperature the composite photoluminescence intensity at 540 nm decays with a characteristic lifetime of 10.9 ns (Fig. 4c) matching that of the single molecules (Fig. 4b). This finding is crucial to point out an important feature of the hetero-ligand emitters. In the case of homo-ligand MOFs, the use of high loading levels induces a shortening of the photoluminescence lifetime and efficiency reduction, because of a partial aggregation of poorly dispersed crystals that can limit also the surface passivation effect of the host matrix. 39 Conversely, in hetero-ligand MOFs this effect appears absent, suggesting that DPT ligands are successfully incorporated as separated and protected units, whose emission ability is insensitive to the nanocrystal environment and aggregation-induced losses. 39,46 The absence of significant reabsorption is further highlighted by low temperature experiments. By cooling the composite down to 10 K, we still observe the first vibronic replica together with a simultaneous increment of the global green emission intensity (+19%, Fig. 4b). This increment is completely ascribed to the suppression of the intramolecular vibrational quenching mechanism at low temperature, as indicated by the emission lifetime that increases up to 12.5 ns at 10 K (+15%, Fig. 4c), thus demonstrating the absence of reabsorption-related losses. A more peculiar dynamic is observed for the radioluminescence blue component peaked at 430 nm. At 10 K, we observe a refinement of the vibronic structure in the residual DPA emission, as well as the expected slight lifetime increment (Fig. 4b, inset). 47 However, we observe the simultaneous growth of an overlapped component at 410 nm, which is the main responsible of the blue emission intensity increment, and a UV emission peaked at 280 nm (Fig. S50). Both these components are competitive channels ascribed to the host PDMS. We notice that these electronic transitions are completely dark at 300 K, thus we speculate that they can represent one of the main dissipative pathways that limit 𝜙 𝑠𝑐𝑖𝑛𝑡 due to a non complete energy transfer from the host matrix to the embedded nanocrystals. The scintillation of large Stoke shift nanocomposites is investigated in a bulky cylindrical specimen (diameter 1 cm, height 0.5 cm) loaded with Zr-DPT:DPA-8% (0.5% weight) irradiated with a pulsed X-ray beam (Methods). The expected emission output is reported in Fig. 4d These findings demonstrate the success of the designed strategy of mixed acenes in MOFs and, despite the scintillation rise time cannot be accurately quantified, they suggest that even faster activation kinetics can be achieved. For example, by employing complementary ligands with better energetic resonance or by developing high-diffusivity nanocrystals by the fine tuning of the intermolecular orientation in the MOF framework, it would be possible to further enhance the energy transfer rate, thus achieving the activation of the Stokes shifted luminescence in times below the 100 ps threshold. These emitters will be therefore the ideal building blocks to realize high-optical quality bulk composite systems exploiting optimized loading strategies, which will enable to increase the amount of embedded nanoscintillators avoiding aggregation and therefore limiting the scattering of the scintillation light also in large area devices. In conclusion, we successfully engineered the composition of co-assembled hetero-ligand ## Methods Synthesis of Zr-DPT. MOF nanocrystals were synthetized under solvothermal conditions modulated by acetic acid. The conditions were optimized to generate a highly crystalline sample. Briefly, ZrCl4, 5,12-bis(4-carboxyphenyl) tetracene (DPT) and acetic acid were dispersed in DMF (see SI for further details). The resulting mixture was heated at 120°C for 22 hours and the orange powder was filtered and washed with fresh solvent before activation at 130°C under high vacuum before characterization. The synthesis of DPT ligand is reported in Supporting Information. were dispersed in a mixture of DMF and acetic acid. The mixture was heated at 120°C for 22 hours and the yellowish powder was filtered and washed with fresh solvent before activation at 130°C under high vacuum. ## Synthesis of Synthesis of Zr-MOF:PDMS composites. PDMS nanocomposites were prepared by dispersing MOF nanocrystals in a prepolymer mixture that was poured in a proper mould and cured at 60°C to obtain self-standing nanocomposites. The nanocomposites were obtained by the reaction of vinyl-terminated polydimethylsiloxane with polydimethylsiloxane-co-methylhydrosiloxane by thermal curing. The cross-linking reaction starting from the polymer terminals preserved the flexibility of the polymer chains and produced very low glass transition. 48 Structure analysis and microscopy. The structure and composition of Zr-MOF nanocrystals and nanocomposites were determined by means of powder X-ray diffraction (PXRD) structure refinement, nuclear magnetic resonance (NMR) spectroscopy, Fourier -transform infrared (FT-IR) spectroscopy, thermogravimetric analysis (TGA), adsorption properties, helium picnometry, scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Details on the instrumental setup and the measurement protocols are reported in the supplementary material file. The crystal structures were refined by the Rietveld method combined with molecular mechanics and plane-wave DFT calculations (see Supplementary Material). Photoluminescence studies. Absorption spectra has been recorded with a Cary Lambda 900 spectrophotometer at normal incidence using quartz Suprasil cuvettes with 0. Measurements on composites as a function of the temperature were performed on films 0.1 cm thickness and diameter 1 cm, mounting the sample in closed circle He cryostat with direct optical access. Radioluminescence studies. Steady state RL measurements were carried out at room temperature using a homemade apparatus featuring, as a detection system, a liquid nitrogen-cooled, back-illuminated, and UV-enhanced charge coupled device (CCD) Jobin-Yvon Symphony II, combined with a monochromator Jobin-Yvon Triax 180 equipped with a 100 lines/mm grating. All spectra are corrected for the spectral response of the detection system. RL excitation was obtained by unfiltered X-ray irradiation through a Be window, using a Philips 2274 X-ray tube with tungsten target operated at 20 kV. At this operating voltage, a continuous X-ray spectrum is produced by a Bremsstrahlung mechanism superimposed to the L and M transition lines of tungsten, due to the impact of electrons generated through thermionic effect and accelerated onto a tungsten target. The dose rate was 0.2 Gy/s, evaluated by comparison with a calibrated 90 Sr-90 Y beta radioactive source and using optically stimulated luminescence emission from quartz crystalline powder (100 -200 μm grains). In order to record the PL measurements, the same acquisition system of RL measurements has been coupled to a 405 nm pulsed diode laser (EPL-405 Edinburgh Instruments) through a quartz optical fibre bundle allowing the illumination of the sample in the X-ray chamber. Scintillation studies. Pulsed X-rays with energies up to 25keV were generated with a repetition rate of 1MHz by a picosecond diode laser at 405nm (Delta diode from Horiba) focused on a X-ray tube (model N5084 from from hamamatsu). In the case of optical excitation, the same laser (405nm) was Discriminator (model 9237, ORTEC). This processed HPM output signal was used as stop signal for a Time to Digital Converter (TDC xTDC4, chronologic), while the start signal was given by the external trigger of the PDL. An optical band-pass filter (450 nm with a FWHM of 40 nm) was used, chosen accordingly to the emission spectrum of the samples, to cut observed air excitation by X-Ray. The scintillation pulse was fitted with a convolution between the Impulse Response Function (IRF) of the whole system with a full width at half maximum (FWHM) of 180 ps and the intrinsic scintillation rate. 9 Light propagation modelling. Simulations of the scintillating nanocomposite performances were carried out using a Monte Carlo ray-tracing method previously presented. 39 The photon propagation follows geometrical optics laws where the interference is neglected. Each photon can be absorbed and re-emitted by a chromophore, isotropically scattered, and reflected or transmitted at the interfaces, where the Fresnel coefficients have been used to compute the reflection probability. The simulated scintillator contains the same number of nanocrystals employed to fabricate the described sample. The absorption, scattering, transmission, or reflection events are chosen according to random Monte Carlo drawing. The simulations were performed using the experimental absorption/luminescence spectrum and emission efficiency of nanocrystals (ΦPL= 67% for Zr-DPT:DPA-8% and ΦPL= 27% for Zr-DPA). The scattering is supposed to induce light attenuation corresponding to an absorption coefficient ranging from 0 to 20 cm -1 .

chemsum

{"title": "Highly luminescent hetero-ligand MOF nanocrystals with engineered massive Stokes shift for photonic applications", "journal": "ChemRxiv"}

single_copy-sensitive_electrochemical_assay_for_circulating_methylated_dna_in_clinical_samples_with_

## Abstract: Tumor-related circulating methylated DNA represents only a small fraction of the total DNA in clinical samples (e.g. plasma), challenging the accurate analysis of specific DNA methylation patterns. Yet conventional assays based on the real-time quantitative methylation-specific PCR (qMSP) are generally limited in detection sensitivity and specificity due to its non-specific amplification interference including primer dimers and off-target amplification. Here we propose a single copy-sensitive electrochemical assay for circulating methylated DNA with ultrahigh specificity on the basis of a sequential discrimination-amplification strategy. Methylated DNA rather than unmethylated DNA in a bisulfitemodified sample is identified and amplified by the asymmetric MSP to generate abundant biotin-labeled single-stranded amplicons with reduced primer-dimer artifacts. Self-assembled tetrahedral DNA probes, which are readily decorated on an electrode surface as nanostructured probes with ordered orientation and well controlled spacing, enable the highly efficient hybridization of the specific single-stranded amplicons due to greatly increased target accessibility and significantly decreased noise. The interfacial hybridization event is quantitatively translated into electrochemical signals utilizing an enzymatic amplification. The proposed assay integrates dual sequence discrimination processes and cascade signal amplification processes, achieving the identification of as few as one methylated DNA molecule in the presence of a 1000-fold excess of unmethylated alleles. Furthermore, the excellent assay performance enables tumor related methylation detection in lung cancer patients with 200 microlitre plasma samples. The results are in good consistency with those of clinical diagnosis, whereas the conventional qMSP failed to detect the corresponding methylation pattern of these clinically confirmed positive patients in such trace amounts of samples. Single copy-sensitive electrochemical assay for circulating methylated DNA in clinical samples with ultrahigh specificity based on a sequential discrimination-amplification strategy † Introduction DNA methylation is widespread in mammals and observed mainly as the addition of a methyl group to cytosines at CpG dinucleotides. The methylation of the promoter region CpG islands in tumor suppressor genes is frequently associated with the progression of cancers such as lung cancer. 1,2 Accumulated evidence has indicated that methylated DNA could be released into circulation during different stages of the tumor, and this cell-free DNA can be regarded as a prognostic indicator for early cancer detection and behavior monitoring. 2,3 However, tumor related methylated DNA represents only a small fraction of the total DNA in complex clinical samples (e.g. plasma), posing persistent technical challenges in the accurate analysis of specifc DNA methylation patterns. Numerous available tools to assess DNA methylation have been successively developed over the past two decades. Generally, three different strategies including sodium bisulfte conversion, 9 restriction enzyme digestion and affinity enrichment 13 are employed to identify methylated cytosines ( m C) from unmethylated cytosines (C). Notably, the bisulfteassisted methylation-specifc PCR (MSP) has been one of the most common methylation analysis tools. This technique relies on the sodium bisulfte treatment of DNA, which converts C to uracils while leaving m C unaffected, thus turning the epigenetic difference into the sequence difference. The modifed DNAs are then amplifed by the PCR with methylation-specifc primers, the products of which are identifed via gel electrophoresis. However, this MSP approach offers only qualitative analysis and low assay sensitivity. The real-time quantitative MSP (qMSP) 14,15 takes advantage of the fluorescent dye reporter or TaqMan probe to achieve quantitative analysis, and improves the sensitivity. Nevertheless, it still often suffers from non-specifc amplifcation including primer dimer formation and off-target amplifcation of unmethylated alleles. 16 Besides the real-time qMSP, several novel end-point MSP strategies have been developed to solve these problems. Notably, the method (methylation-specifc quantum dot fluorescence response energy transfer, MS-qFRET) reduced the number of PCR cycles required to inhibit the formation of primer dimers and pushed the detection limit of methylated DNA down to 15 pg. 19 However, off-target byproducts from unmethylated DNA amplifcation still exist and can be detected by FRET as interfering signals. 19,22 Nie and coworkers employed a supercharged green fluorescent protein as a versatile probe for the detection of MSP products, achieving the highly sensitive detection of methylated DNA extracted from human colon carcinoma tissue samples. 21 Furthermore, this method used toehold strand displacement hybridization to improve the specifcity by singlebase mismatch between methylation and unmethylation sequences. In addition, Li's group has developed an alternative methylation assay based on the ligase chain reaction. This assay evaded the problems of primer dimers and off-target amplifcation in the MSP, achieving the determination of as low as 10 aM and 0.1% methylated DNA from a large excess of unmethylated DNA with multiplexed methylation sites. 23 Yet, mismatch ligation 24 and blunt-end ligation also exist in the ligase chain reaction, causing non-specifc signals. Thus, developing a highly specifc and sensitive methylated DNA detection strategy without non-specifc amplifcations still remains a challenge. Considering that the high concentration of the primers is always accompanied by the generation of primer dimers, 25 we speculate that decreasing the concentration of the primers can reduce and even eliminate the dimers, for example, in the asymmetric PCR. However, it conversely suffers from low amplifcation efficiency. Thus the integration of a downstream amplifcation strategy is required to further improve the sensitivity. 26,27 Additionally, as off-target amplifcation still remains in the asymmetric PCR, the combination of specifc downstream discrimination can further improve the specifcity. Recently, we developed a tetrahedral DNA nanostructure-based electrochemical detection strategy. 32,33 These tetrahedrons provide greatly increased target accessibility and signifcantly minimize non-specifc adsorption. By simultaneously utilizing enzymatic amplifcation, we realized the detection of nucleic acids in the aM-level. Collectively, by integrating the above solutions, we expect to overcome the problems of primer dimer formation and off-target amplifcation, realizing methylation detection with high sensitivity and high specifcity. Herein, we developed a sequential discrimination-amplifcation (SEDA) strategy for circulating methylated DNA detection with single-copy sensitivity and ultrahigh specifcity. In this method, methylated DNA rather than unmethylated alleles in a bisulfte-modifed genomic sample was identifed and amplifed by the asymmetric MSP (AMSP) using a biotin-labeled methylation-specifc primer, producing lots of biotin-labeled single-stranded amplicons with reduced primer-dimer artifacts. DNA nanostructured probes via decorating self-assembled tetrahedral DNA on a gold electrode surface were employed to capture these amplicons. These probes greatly increased target accessibility and minimized the non-specifc adsorption of amplifcation byproducts due to the rigid scaffold, ordered orientation and well controlled spacing, thus enabling the high efficiency and specifc hybridization and signifcantly decreasing noise. Horseradish peroxidase-conjugated avidin (avidin-HRP) was then captured via biotin-avidin binding to generate an electrochemical signal by the enzyme catalytic process. Dual sequence discrimination processes, including methylation-specifc annealing and specifc interface hybridization, as well as cascade signal amplifcation processes represented by the asymmetric MSP and HRP catalytic reaction, were well integrated in the proposed assay, realizing the detection of single-copy methylated alleles. Moreover, the DNA methylation at the p16INK4a gene promoter in trace amounts of plasma samples (200 microlitres) from eleven lung cancer patients was accurately determined by our proposed strategy, the results of which are in good consistency with those of clinical diagnosis. ## Results and discussion The design principle of the proposed SEDA strategy is illustrated in Fig. 1. Circulating DNA was frstly extracted and converted with sodium bisulfte treatment. Methylation-specifc primers were then applied to discriminate the methylated sequence. Through AMSP amplifcation, abundant specifc amplicons of single-stranded DNA (ssDNA) were produced, accompanied by possible non-specifc amplicons including offtarget amplicons and primer dimers. Subsequently, interfacial DNA nanostructured probes were employed to discriminate specifc amplicons from the possible non-specifc ones. These probes are complementary to the region near the primer binding site in the specifc amplicons. The hybridization sequence length is optimized from our previous work to guarantee the effectiveness as well as the specifcity. Non-specifc adsorption to the electrode surface is signifcantly minimized. Finally, avidin-HRP was introduced into the system, catalyzing the second signal amplifcation. The synergetic combination of dual sequence discrimination and cascade signal amplifcation decreased the interference from non-specifc amplifcation and adsorption, and increased the detection sensitivity and specifcity. According to the chronocoulometric quantitation method reported, 34 using a cationic redox marker, RuHex, we can prove the immobilization of the DNA tetrahedron on the electrode and further calculate the surface density (Fig. S1 † 10 12 molecules per cm 2 , which corresponds to around 0.2 pmol on the 2 mm-diameter gold electrode. Using 200 nM of reverse primer as the excess primer, we can obtain at most 3.6 pmol of the ssDNA target amplicon (excluding 0.4 pmol of the double stranded amplicon). A defnite amount of tetrahedral probe can theoretically hybridize with an equal amount of amplicon. Therefore, 1 ml of the reaction solution may be approximately saturated. To validate this, we further tested 3 ml of the solution and the result in Fig. S2 † shows that the current only increases by 10% of that of 1 ml. Thus, the AMSP product is substantially excessive and when optimizing the experimental conditions, we diluted the AMSP solution 10-fold. Considering that the probe is relatively long and that high ionic strength can help hybridization, we explored the effect of increasing the Na + concentration in the hybridization buffer. 35 As shown in Fig. 2A, the addition of Na + in the hybridization system induced rapid background growth, whereas relatively slow signal growth was observed. Therefore, in order to avoid the impact of an excessive background, we chose 200 mM of Na + for the following experiment. In addition, we tested the hybridization time and concluded that 30 min is enough for the reaction (Fig. 2B). According to the design of the proposed hybridization strategy, we used a 34 nt probe to capture the 150 nt amplicon, yielding an overly long 3 0 -tail (the tail in Fig. 1). Due to the possible steric effect and electrostatic repulsion between the long tail and the skeleton of the tetrahedral DNA, we systematically studied the effect of different tail lengths on hybridization. We designed a series of forward primers and synthesized a template strand. All fve single-stranded amplicons (inset of Fig. 3B, present in dsDNA) contain a region complementary to the tetrahedral probe, yielding a respective 93, 68, 45, 24 and 0 nt (non) tail. The effect of the primer sets in the qMSP is shown in Fig. S3. † We can see that when template DNA is present (positive), the Ct values are relatively lower, but in the absence of template DNA (negative), the curves still rise as the primer dimer forms, which is inevitable especially in the conventional MSP. However, when performing the AMSP, the background is much lower (Fig. 3A). As for the signal in the electrochemical measurements when diluting the PCR solution (stored as a stock solution) 10-fold, the difference of the current values caused by the different length of the tails responding to hybridization is negligible. Considering that the steric effect and electrostatic repulsion may be affected by the target concentration, we additionally tested the stock solution and the solution diluted 50-fold. The further diluted solution showed a similar phenomenon but the stock solution seemed slightly different. This may be thanks to the special construction of the tetrahedron that contains four sloping rigid triangles of DNA helices as the side face, with every two terminals of the oligonucleotides merging at each vertex. Under this circumstance, while this homogeneous self-assembled monolayer is neatly ordered in a relatively close arrangement at the bottom, the upper space is oppositely spacious. Thus, when target ssDNA is limited (dilution ratios of 1 : 10 and 1 : 50 in Fig. 3B), the dissociative tails in the relatively large buffer zone will not induce the steric effect and electrostatic repulsion between targets and targets with the tetrahedrons, which in turn does not interrupt or interfere with hybridization. However, when the targets are saturated (dilution ratio of 1 in Fig. 3B), longer tails will be affected more, but not enormously, and the factors of AMSP should also be taken into account as the presented electrochemical signal is a combination of the sequential discrimination-amplifcation processes. Furthermore, the nontail target generated a very large current signal in high concentration, but no signifcant increase in limited concentration. Consequently, the original set of the primer which generates a long 150 nt target can be used in this strategy with no more consideration, as long as the target for hybridization is not saturated (which can be judged from the current value). In order to meet the critical demand for the diagnosis of methylation-related diseases, a good S/B ratio and great sensitivity for the analysis of an amount of DNA input as low as possible should be achieved. Here, we investigated the variation of the chronoamperometry current with the concentration of methylated DNA input. We used 1 ml of the AMSP product to produce a strong enough current so as to maintain the potential discrimination when an extremely low amount of DNA was added. As can be seen from Fig. 4A, methylated DNA down to a single copy can be discriminated from the NTC (non-template control), which means that once a correspondingly lower amount of target ssDNA is amplifed, it can be captured by the stable DNA nanostructured probe system and presented as the further amplifed current signal by the TMB-HRP system. Meanwhile, a linear correction (inset of Fig. 4A, R 2 ¼ 0.983) can be observed when the DNA amount is relatively low. When the DNA amount was increased, an exponential curve was observed, which indicates that more and more tetrahedral probes were occupied and the second amplifcation of the signal continued. A plateau effect is nearly reached when a common ng-level of DNA was used as the template in the AMSP, which corresponds with the almost saturated probes. Compared to the conventional qMSP which usually has ng-level sensitivity (Fig. 4C), we expanded this method by analyzing a very large range of methylated DNA inputs (by arbitrarily diluting the AMSP product when a relatively large amount of DNA was added), and it has the potential to detect very low amounts of methylated DNA (down to a single copy that can be discriminated from the NTC). Furthermore, the methylation index across a target region can vary among cancer types and stages of the disease. 36 Clinical samples usually also contain a mixture of tumor and normal cells, leading to a mixture of methylated (M) and unmethylated (U) DNA, where the presence of unmethylated sequences can potentially influence the sensitivity and the specifcity of the diagnostic assay. 37 Particularly, for early cancer detection, it is essential to discriminate the low amount of methylated DNA from a high background of unmethylated DNA. Therefore, we investigated the ability of our approach to detect the DNA methylation index by mixing artifcially methylated DNA with unmethylated DNA (a converted sequence from healthy volunteers) in a total amount of 15 ng (assuming 100% recovery in the conversion process) at different proportions representing 0, 0.1, 0.2, 0.5, 1, 5, 20, 50 and 100%. As shown in Fig. 4B, the chronoamperometry current increased with the methylation index. In the case of 0%, there was no methylated sequence present in the solution, and consequently, no methylation-specifc primerbased extension occurred, and no sequence specifc hybridization formed. The low background is caused by the non-specifc binding between the biotin-labeled single strands (no extended reverse primers or primers extended in a non-specifc way, nonetheless lacking the ability to hybridize with the tetrahedral probes, i.e. primer dimers) and the exposed gold electrode surface (Fig. S9 †). Similar to the situation when only methylated DNA is present, the whole process successively consists of an exponential phase (the initial phase in the AMSP when both the forward and reverse primers exist), a linear phase (the latter phase in the AMSP when only the excess reverse primer exists) and a non-linear phase of target-probe hybridization and HRPcatalyzed TMB redox activity, resulting in an amperometric current. 38 Thus, according to the curve obtained from the situation when only methylated DNA was present, we calculated the methylation index derived by the measured current value (Yaxis). The inset of Fig. 4B shows the plot of the measured methylation index against the actual input methylation index. A good correlation coefficient (R 2 ¼ 0.996) was observed, proving that the presence of the unmethylated sequence does not interfere with the target methylated sequence detection in our approach. Meanwhile, the overly excessive unmethylated DNA interferes with the result obtained from the qMSP (e.g. the magenta curve compared to the green curve in Fig. 4C). Conversely, the high specifcity reaching a 0.1% methylation index in our approach thus guarantees the ability to analyze a real sample even from early cancer patients. At this point, we have achieved the identifcation of as few as one methylated DNA molecule in the presence of a 1000-fold excess of unmethylated alleles. We collectively compared our method with the conventional qMSP. As can be seen in Fig. 5, only a ng-level (300 copies) of the DNA and 5% (20-fold) of the methylated DNA can be detected in the qMSP. Additionally, a reduced number of PCR cycles is enabled when coupling with this electrochemical method (Fig. S10 †). These advantages are achieved by the synergetic combination of dual sequence discrimination and cascade signal amplifcation, which significantly decreased the interference from non-specifc amplifcation and adsorption, and increased the detection sensitivity and specifcity. To validate the application in trace amounts of clinical patient samples, circulating DNA extracted from 200 microlitres of NSCLC patients' plasma was tested. Gratifyingly, all the eleven patient samples produced relatively much higher current values than those of the healthy volunteer and the negative control (Fig. 6), confrming the high methylation level of circulating DNA in these patients, whereas the conventional qMSP failed to detect the corresponding methylation pattern of these patients in such trace amounts of samples. These results show that our method has the ability to sensitively analyze trace clinical patient samples and the potential for DNA methylation detection-based early cancer diagnosis. ## Conclusions In this work, a single molecule-sensitive electrochemical assay for tumor-specifc circulating methylated DNA with ultrahigh specifcity was reported based on a SEDA strategy. Single-copy methylated DNA in a clinical sample was accurately identifed even in the presence of a 1000-fold excess of unmethylated alleles. The signifcant merits are ascribed to the integration of sequential discrimination-amplifcation processes, embodied in dual sequence discrimination events including methylation-specifc annealing and specifc interface hybridization, as well as cascade signal amplifcation processes represented as the AMSP and HRP catalytic reaction. The former eliminates the non-specifc amplicons to reduce the background and the latter increases the signal. Notably, the AMSP dramatically reduced primer-dimer artifacts and the interfacial nanostructured probes signifcantly resist the non-specifc adsorption of amplifcation byproducts. Additionally, the high S/B ratio enables the reduced use of PCR cycles, which allows for the further avoidance of non-specifc amplifcation byproducts with positive feedback. Finally, while the conventional qMSP lacks this ability, the proposed strategy achieved the robust analysis of DNA methylation in trace amounts (200 microlitres) of plasma from lung cancer patients. Therefore, our single-copy sensitive electrochemical assay is superior, and we believe that it has an immediate and promising impact on fundamental research and clinical applications.

chemsum

{"title": "Single copy-sensitive electrochemical assay for circulating methylated DNA in clinical samples with ultrahigh specificity based on a sequential discrimination\u2013amplification strategy", "journal": "Royal Society of Chemistry (RSC)"}

a_novel_design_for_porphyrin_based_d–s–a_systems_as_molecular_rectifiers

## Abstract: Two Si-based hybrid self-assembled monolayers of porphyrin based on a D-s-A system were synthesized by electro-grafting. The monolayers showed a stable and reversible rectification at room temperature. The monolayer fabricated using a porphyrin with an eleven-carbon alkyl chain linker was comparatively more compact and exhibited a 10 5 times higher rectification ratio (RR) relative to another similar system that had a six-carbon alkyl chain linker, possibly because of the compact packing. ## Introduction Miniaturization is a vital need of the electronics industry, but it is limited by changes in the bulk properties of materials as they move to nanoscale dimensions. Most of the successes in this feld have focused on the electrical properties of organic molecules placed between metal electrodes. In particular, selfassembled monolayers (SAMs) of alkanes and aromatic thiols on gold substrates have been very popular for constructing metal-molecule-metal (MMM) junctions. Concurrently, efforts aimed at synthesizing metal-molecule-semiconductor (MMS) junctions by covalent linking of organic molecules to semiconductor surfaces are gaining momentum. 11 Such assemblies present opportunities for novel molecular electronic charge transport mechanisms, and are potentially compatible with conventional metal oxide-semiconductor (MOS) technology. To this end, there is a burgeoning interest in small organic molecules capable of switching their redox status, which, in association with semi-conductors such as Si, may scale down the size of the molecular electronic devices. Here, surface potential tailoring can be achieved by chemically-grafting organic molecules onto Si to develop improved hybrid molecular devices. For example, the p-n junction threshold voltage for rectifcation can be adjusted by changing the electronic nature of the organic p group molecules, instead of via the classical doping method. 14 Different techniques such as making Langmuir-Blodgett (LB) flms 15 or SAMs of organic molecules on solid substrates via MMM junctions 16 are most commonly used for this purpose. Compared to the LB flms, SAMs are easy to prepare and may be more robust as the organic molecules are sturdily anchored onto the metal substrates at fxed distances. Chemically bonded monolayers on Si surfaces can be prepared either on Si oxide (SiOx) surfaces or on oxidefree Si, 11,17 the latter being preferred due to better electronic coupling of the Si molecules and the lack of charging effect. Indepth reviews with excellent analyses of the different methods of fabrication and characterization of SAM junctions on Hterminated Si surfaces are available. 18,19 The protocols usually adopted for constructing densely packed Si-organic hybrids involve the deposition of functionalized alkenes/alkynes using heat, 20 light, 21,22 electrochemical techniques, radical initiators 23 or Lewis acids, 24 as well as alkylhalides via either a Grignard route or lithiation. 25 This is followed by attachment of the electro-active organic molecules to the terminal functionality of the resultant alkane/alkene-Si hybrids by esterifcation or amidation. However, due to steric factors, not all of the deposited alkane/alkene moieties can be modifed with organic molecules. This may produce non-uniform organic-Si hybrids. In our previous work, we found that cathodic electro-grafting of pre-synthesized alkenylated electro-active organic molecules onto a Si-H surface can conveniently provide SAM-based molecular electronics devices with the following advantages: 26 the process is simple; it can be monitored in situ to ensure completion of deposition; it can exclude oxidation and/or hydrolysis at the Si surface due to the negative potential bias of the Si wafers; and it can produce materials where the Si-H surface is modifed only by the chosen molecules. Amongst the many electron-rich organic molecules, porphyrins 27 are ideally suited for fabricating molecular devices because they: (i) can form stable p-cation radicals and exhibit two accessible cationic states in their monomeric forms; (ii) have long charge retention times, resulting in lower power consumption; (iii) are highly stable, 33 and (iv) can form selfassembled structures. 34 In view of these favourable attributes, porphyrins have been extensively used as p molecules for the construction of storage devices, molecular wires and memory devices. Reports on current rectifcation using C 60 -porphyrin combinations also exist. 35 Molecules exhibiting rectifcation behavior with a high rectifcation ratio (RR) are very useful for making diodes. According to Aviram and Ratner, a single molecule with a donor-spacer-acceptor (D-s-A) structure should behave as a rectifying diode when placed between two electrodes, where the s-bond bridge prevents the direct overlap of the donor (D) and acceptor (A) energy levels to allow unidirectional flow of current. 1 Several groups have experimentally verifed this model, but porphyrins have never been used for this purpose in silicon hybrid systems. 36,37 In the present investigation, two such single molecules (5a/5b) were synthesized, where porphyrin and aniline moieties served as the (A) and (D) units respectively, while a -CH 2 -NHmoiety was anticipated to be a suitable spacer. These molecules were electro-grafted onto Si-surfaces using the C-6/C-11 alkenyl chain of 5a/5b as the linker to construct the respective MMS heterostructures. Measurement of their I-V behavior revealed high current RRs for these assemblies. Moreover, a subtle change in the linker length signifcantly changed the monolayer packing on the Si-surface, resulting in a pronounced alteration in the current rectifcation properties. ## Synthesis of the porphyrins Porphyrin-based functional molecules are, by and large, synthesized via functionalization of the aryl moieties of unsymmetrical meso-tetraryl porphyrins. However, the synthesis of unsymmetrical porphyrins is fraught with limitations such as poor yields and tedious isolation procedures. Instead, functionalization of the pyrrole units of the porphyrins offers a better alternative to alter the porphyrin scaffold. However, this strategy is rarely used because the porphyrin pyrrole units are inert towards most electrophilic reactions such as Friedel-Crafts alkylation and acylation, while halogenation 38 and nitration 38,39 often lead to di-or higher substituted products. An exception to this is the Vilsmeier-Haack reaction, which can provide mono-formyl porphyrins in appreciable yields. 39 We reasoned that the resulting formyl group could subsequently be used to construct the desired D-s-A structure for the present studies. So, following Bonfantini's method, 39 tetraphenylporphyrin (TPP, 1) was converted to Cu(II)-TPP and then subjected to the Vilsmeier-Haack reaction to obtain bformyl-TPP (2). For the synthesis of the donor part of the molecule, p-nitrophenol was o-alkylated with either 1-bromohexene or 1-bromo-10-undecene to furnish compounds 3a and 3b, respectively. These were converted to aniline derivatives 4a and 4b by reduction with Zn/HCO 2 NH 4 . Next, aldehyde 2 was separately subjected to a reductive amination using 4a or 4b to obtain the target porphyrins 5a and 5b, respectively. Previously, Welch et al. 40 synthesized the Schiff's base of 2 in toluene after 72 hr, using a Dean-Stark apparatus for simultaneous removal of water. We performed the reductive amination in THF in the presence of 4 molecular sieves followed by a one-pot reduction of the intermediate imine to obtain 5a and 5b in improved yields ($78%) in only 6 h (Scheme 1). ## Device fabrication Preparation of the Si-hybrids. Molecules 5a and 5b were electrochemically deposited on H-terminated silicon via a twostep process, which is schematically shown in the ESI (Fig. SL1 †). In the frst step, application of a negative potential to the working electrode releases H free radicals from the Si-H surface. The newly generated nucleophilic Si atoms subsequently react with the alkene functionalities of 5a and 5b to form Si-C bonds, resulting in an irreversible oxidation peak at $0.3 V. A similar oxidation peak was observed with 1-undecene (Fig. SL2 †), but not with the blank Si sample (electrolyte only), confrming our interpretation. Cyclic voltammograms (CVs) (Fig. 1), recorded during this electrochemical deposition helped to monitor the extent of deposition. Disappearance of the oxidation peak indicated completion of the process. AFM analysis revealed the formation of homogeneous monolayers with both 5a and 5b after 25 scans. Characterization of the monolayers. To ensure monolayer deposition on Si, the electro-grafted materials were characterized by contact angle measurements, polarized FT-IR spectroscopy, ellipsometry, AFM, secondary ion mass spectrometry (SIMS) and electrochemistry. The contact angles of deionized water at the Si surface grafted with 5a and 5b were 55 and 64 , respectively. For the cleaned Si wafer and the C-11 alkyl-grafted Si surfaces, the angles were 84 and 112 , respectively. The value for the cleaned Si wafer is consistent with several previous reports. The low contact angles of the porphyrin monolayers suggested they were tilted on the Si-surface, exposing the pyrrole and amine nitrogen atoms for interaction with water droplets. The observed contact angles of the porphyrin monolayers are in close proximity to the reported contact values (66-74 ) for thiophene-terminated alkyl monolayers on Si-surfaces that were prepared by a late-stage attachment of the aryl moieties. 46 This established the suitability of our direct attachment protocol for the preparation of the monolayers. The average thicknesses of the monolayers, estimated by ellipsometry were found to be 2.4 AE 0.1 nm and 2.9 AE 0.2 nm in case of 5a and 5b, respectively. AFM analysis revealed that the monolayers formed after 25 scans were organized with the least number of voids and hillocks. The void depth and RMS roughness of the 5a monolayers were $2.5 nm and 0.91 nm, respectively, while for 5b they were 3 nm and 0.7 nm, respectively (Fig. 2). Compared to 5a, the monolayers of 5b were more compact and uniform with a larger grain size. Fast scan (10 V s 1 ) CVs (Fig. 3) of the respective porphyrin monolayers exhibited a reversible peak at +0.8 V, confrming attachment of the porphyrin moieties. This was absent in the blank Si sample and the C-11 alkyl monolayers. The net charge transferred during the oxidation process, calculated from the area under the oxidation peak divided by the scan rate were 8.6 10 7 C and 2.45 10 6 C for 5a and 5b, respectively. Using these values, the surface coverages for the monolayers were calculated using the formula: surface coverage ¼ total charge/(F area dipped in electrolyte). The surface coverages were 1.11 10 12 and 4.5 10 14 molecules per cm 2 for 5a and 5b, respectively. Thus, the areas occupied by each molecule in the 5a and 5b monolayers were 90 nm 2 and 22 2 , respectively, indicating that 5b formed more compact monolayers than 5a. The signifcantly higher value for 5a compared to that previously reported 19 for monolayers of simple C 18 -, C 16 -, and C 12alkanes on Si (100) revealed poor packing. This may be due to the edge-on orientation of the porphyrins. 47 On the other hand, the value for the 5b monolayers matched well with the theoretically calculated diameter (14.8 ) of TPP, 48 indicating that the molecules were tightly packed due to p-p stacking, which was also revealed in the AFM images (Fig. 2). Consistent with the AFM analysis, the surface area covered by 5b was several fold that covered by 5a. SIMS of the 5a monolayers showed mass peaks at m/z 795, 691, 675 and 596 amu, while for the 5b monolayers peaks appeared at m/z 777 and 386 amu (Fig. 4), revealing that the molecules remained intact during the grafting process. The observed higher mass fragments, in the case of the C-11 monolayers, was consistent with its longer alkyl chain length vis-à-vis that of the C-6 monolayers. IR peaks due to -CH 2 vibrational modes can provide better insight into the van der Waals interactions between the alkylated porphyrin rings anchored parallel on the Si surface. This, in turn, may help explain the better packing of the 5b monolayers vis-à-vis that of 5a. In pure solid alkane monolayers, the hydrocarbon chains exist in an all-trans confguration such that the carbon backbone of each molecule lies in a single plane. However, in liquid form, there is substantial out-of-plane twisting around the individual bonds, altering the frequency of the -CH 2 vibrational modes. 45,49 The polarized FTIR spectrum (Fig. 5) of the 5a monolayers exhibited a N-H stretching frequency at 3251 cm 1 along with symmetric (n s ) and asymmetric stretching (n a ) vibrational modes for the CH 2 groups at 2856 and 2927 cm 1 , respectively. In contrast, the respective IR absorption peaks of the 5b monolayers were at 3255, 2840 and 2921 cm 1 . Our results showed that the alkyl chains in the monolayers of 5b are more rigid like those in pure solid alkanes, while those in the monolayers of 5a are twisted. This clearly explained the observed improvement in packing for the 5b (C-11 linker) monolayers over the 5a (C-6 linker) layers. 50 From the Xray photoelectron spectroscopy (XPS) data, the peak for the monolayers at 99.5 eV could be attributed to the Si-C bonds, while the absence of a SiO 2 peak at 103 eV confrmed that the monolayers were free of SiO 2 (Fig. SL3 †). I-V measurements. In order to measure the I-V characteristics, a metal-molecule-Si (n++) structure was constructed (Fig. 6(a)), using a tiny drop of liquid Hg (40 mm diameter) as the counter electrode. The area in contact with the grafted monolayer, measured using a goniometer, was 0.002 mm 2 . The I-V curves (Fig. 6(b)) of the devices constructed using the monolayers of 5a and 5b showed current rectifcation in the reverse bias. The maximum RR was observed at AE1 V for both of the devices. The ratio for the monolayers prepared using 5b was very high (10 7 ), while that for 5a (C-6 linker) was $100. However, while both systems were stable during repeated voltage scanning up to 100 scans at a scan rate of 0.01V s 1 , the RRs reduced gradually from their original values to $10 000 for 5b and $10 for 5a after 50 scans. It is already well known that when electrodes are asymmetric (and especially when the work functions of the electrode materials are different) any molecule can show current rectifcation. 51,52 Therefore, to see the effect of using asymmetric electrodes (if any), we also recorded the I-V curves of two control devices made of (n++) Si/Hg (SL4 (a) †) and (n++) Si/C-11 alkyl monolayers/Hg (SL4 (b) †). These showed marginal rectifcations with RR I + /I values of 0.75 and 2, respectively. Recently, we constructed a n + -Si/pyrene C-11 monolayers/Hg device, which exhibited current rectifcation in the positive bias with RR of 100 at 1 V. 53 Furthermore, a n + -Si/5-(4-undecenyloxyphenyl)-10,15,20-triphenyl porphyrin (TPP C-11) monolayers/Hg system constructed by our group showed a marginally asymmetric I-V with signifcant hysteresis. In the positive bias scan (0 to +0.8 V), the current jumped by an order of magnitude at +0.6 V. However, on the reverse scan (+0.8 to 0 V) the current did not retrace the curve and remained at a higher value. 54 Taken together, the I-V curves of all these devices clearly indicated that the results for the present devices were not due to electrode asymmetry. Moreover, the I-V results for the n + -Si/TPP C-11 monolayers/Hg device indicated that the observed rectifcations in the negative bias using the 5a/5b molecules were not due to resonance tunnelling through the TPP moiety. The AFM and fast scan CV results showed better molecular stacking of the 5b monolayers, which may be due to the longer alkyl bridge in 5b than in 5a. 55 This may contribute to the better electrical characteristics of the 5b monolayers because the overlapping of electron clouds favours the generation and transport of charge carriers to induce intrinsic conductivity. Consequently, a signifcantly higher maximum RR was observed for the 5b monolayers. Control experiments, carried out with a blank Si sample as well as C-11 alkyl chain-grafted Siwafers showed nearly symmetrical sigmoidal I-V curves (Fig. SL4 †), eliminating any doubt about artifacts. The void sizes ($0.2-0.4 nm) of the present Si-alkyl porphyrin/Hg junctions were small compared to the size of the Hg drops ($40 mm). Therefore, Hg drops are unable to penetrate through the pinholes of the SAMs and the measured I-V is expected to be direct. Statistical analyses of the data and junction yields are extremely valuable to discriminate artifacts from the real data. Previously, Kim et al., 56 and Nijhuis et al., 57,58 employed extensive statistical analyses to assess the performance of SAM-based devices. In the present work, we constructed only 80 devices for each of compounds 5a and 5b. Nevertheless, we analyzed the statistics of our I-V results as shown in Fig. 7 and Table 1, and summarized below. For compound 5a, only 25% of the devices showed RR values of 80-100, while an additional 15% of the devices showed RR values of 50-80. However, the RR values of 44% of the devices were <50, while 16% of the devices didn't show any rectifcation. The device statistics for the monolayers of compound 5b were very impressive. The RR values for the majority (35%) of the devices were 10 5 -10 6 , while 21% of the devices showed RR values of 10 6 -10 7 . Another 20% showed RR values of 10 4 -10 5 , and the rest had RR values of 1-10 4 . The performance of the devices made of compounds 5a and 5b were satisfactory. In particular, the RR values of the Hg/5b/Si (n++) devices were far superior to that of molecular rectifers reported so far. 53,57,58 The current rectifcation properties of various D-s-A-based SAMs and LB flms in contact with noble metal electrodes have been described. 59 Results for some representative examples clearly establishes the signifcantly superior performance of the devices described in the present study. For example, a LB flm of a pyrenyl carbamate in a M-M-M junction exhibited a RR value of 130 at $2.5 V. 60 Likewise, quinolinium and tetrahydroisoquinolinium iodide-based SAMs deposited on Au substrates showed RR values of 50-150 and 30-80, respectively, at AE1 V. 61,62 Meanwhile, the RR value for SAMs of quinolinium salts joined by a truncated S-C 3 H 6 group on a Au surface was found to be 12 at AE1 V. 63 In another study, a LB monolayer of the D + -p-A molecule, hexadecylquinolinium tricyanoquinodimethanide on Au electrodes showed a maximum RR of 27.5 at 2.2 V. 64 However, literature reports on metal-porphyrin-semiconductor junctions are scarce. SAMs of 4-aminothiophenol/ ZnTPP/fulleropyrrolidine (PyC 2 C 60 ) on a Au (111) surface showed a modest RR of 24 at 1.8 V. 35 Interestingly, self-assembled layers of Fe(III)-5,15-di[4-(s-acetylthio)phenyl]-10,20diphenyl porphyrin on annealed Au crystal facets on glass substrates showed asymmetric I-V curves with the highest RR up to 9000, but the majority of the devices showed RR ¼ 20-200 at AE1 V. 65 To confrm our current rectifcation results, we computed the theoretical I-V curve of the device made of 5b. Initially, the ground state (GS) geometry of molecule 5b was optimized using an ab initio molecular orbital theory based LCAO-MO approach as implemented in the GAMESS software. The ionic optimization of molecule 5b was carried out without any symmetry constraint at the B3LYP/6-31G(d,p) level of theory. To calculate the transport characteristics, a suitable device was constructed using the optimized confguration of the molecule as the central device region between two electrodes. Besides the active parts of the device, the central region also included a sufficient part of the contacts, such that the properties of the electrode regions could be described as bulk materials. This could be ensured by extending the central region into a few layers of the metallic contacts. The calculation of the electron-transport properties of the system was divided into two parts: (i) a self-consistent calculation for the electrodes with periodic boundary conditions in the transport direction, and (ii) a self-consistent open boundary calculation of the properties of the central region, where the electrodes defne the boundary conditions. The complete details of the method are described in the literature. 66 In the present experimental set-up, we have used a highly doped Si substrate that is expected to undergo reconstruction. This restricted accurate modelling of the molecule-substrate interface using the ab initio formalisms of our computational resources. Therefore, a model for a two-probe system was constructed (Fig. 8(a)) by placing the molecules between two Au electrodes. We modelled the electrodes as part of truncated solid crystals. Unlike Hg, which is liquid under the experimental conditions and has a complicated structure, Au possesses a well-defned face-centred cubic crystal structure. Additionally, the pseudo-potential for Au is robust and it has been tested and used by many groups as a model electrode. 67,68 Understandably, the chosen system is not ideal for verifcation of the experimental results. However, our calculations were primarily aimed at a qualitative understanding of the electron transport through these molecules, and not a quantitative comparison, justifying our choice of Au electrodes. It is worth noting that Zheng et al. 69 recently reported the NDR properties of C 60 based electronic devices, wherein they claimed that the fndings were independent of the type of electrodes used. For construction of the theoretical device, a thiol end group was used for attachment of the molecule with the electrode. The interface geometry of the thiol-terminated molecule and the electrode was optimized to ensure good overlap between the device and the electrodes. Previously, we have reported the interaction of methyl thiol, a prototype device molecule, with an extended Au(111) surface using a plane wave based pseudopotential method. 70 The results showed that the terminal S atom binds at the hollow site of the Au(111) surface and the distance between the Au and S atoms is 2.52 . Using this information, we constructed the model for our present calculations. Two Au(111)-8X8 surfaces were used as the left and right electrodes. The Au/molecule/Au confguration was divided into three parts: left electrode, right electrode, and the central scattering region. In our models, there were three Au layers in each of the left and right electrode unit cells. The scattering region was composed of the isolated molecule together with the respective two Au layers on the left and right sides. The electron-transport properties of the Au/molecule/Au systems were investigated using the ATK 11.2.3 program, where semi-empirical extended Hückel theory, in combination with a frst-principle NEGF, was employed. 71 A k-point sampling of 100 was used in the electron-transport direction (Z direction). 70 Consistent with the experimental results, the theoretical I-V curve also showed rectifcation in the reverse bias (Fig. 8(b)). In general, the forward bias current-flow should be determined by the HOMO states of the molecules, while their respective LUMO states would dictate the reverse bias current. The observed rectifcation in the reverse bias is a result of an alignment of the LUMO levels of the molecules with the Fermilevels of the electrodes. To verify this, we determined the HOMO and LUMO energy levels (Table 2) of 5b by theoretical calculations using an ab initio method (GAMESS software). Ionic optimization without any symmetry constraints was carried out at the B3LYP/6-31G(d,p) level of theory where the exchange correlation functions are expressed using hybrid density functional theory. It was observed that the HOMO of the molecule was at 4.707 eV and the LUMO at 2.062 eV. The HOMO of the molecule (located at p-aminophenol group) is in close proximity to the Fermi level of electrode, but due to non contact with the electrode, resonance tunnelling will be difficult. For the LUMO, the energy difference with the electrode Fermi level is too large to undergo resonance tunnelling. Therefore, the observed current rectifcation is unlikely to be due to resonant tunnelling, but molecular asymmetry (D-s-A), which favours Aviram and Ratner's mechanism for rectifcation. From the spatial distribution of the HOMO and LUMO energy levels (Fig. 9), it could be seen that the HOMO was localized on the p-aminophenol segment, while the LUMO was on the porphyrin ring, and their separation by the spacer, CH 2 , allowed a unidirectional flow of electrons. This is consistent with the Aviram and Ratner theory that suggests that a single molecule with a donor-spacer-acceptor (D-s-A) structure should behave as a diode when placed between two electrodes if a non-conjugated s-bond bridge prevents direct overlap of the donor and acceptor energy levels. Moreover, a monolayer will rectify if its molecules are aligned in register between two electrodes such that they work together when electrons flow from the electrode MD (attached to the acceptor) to D, and then exit from A to the electrode MA (attached to the donor). To underscore the mechanism of electron flow between the donor and acceptor moieties, we also calculated the energy levels of the individual components (p-aminophenol and TPP) of 5b using the same computational approach. The energy level diagram of the constituent species is shown in Fig. 10. The LUMO energy level of p-aminophenol was at a higher energy than the TPP moiety. This will make it the donor. Therefore, under reverse bias, when electrons flow from MD to D, the electrons will move from the HOMO of p-aminophenol to its LUMO, tunnel through the bridge to the vacant LUMO of TPP, and fnally transfer to the Hg electrode to complete the reversedirection flow. In principle, reversing the orientation of the molecular dipole would change the I-V curves of the devices. This would also confrm that the observed rectifcation is not simply because of the non-symmetric nature of the device system. 72 However, construction of this type of device is synthetically more demanding for the following reasons. The synthesis has to start with an unsymmetrical porphyrin, wherein the porphyrin needs to have the alkenyl attachment at one of its pyrrole moieties for grafting to the Si wafers. Synthesis of the required porphyrin proceeds via mono-bromination of the porphyrin followed by a Stille coupling with 11-trimethylstannyl undecene. Next, the unsymmetrical porphyrin needs to be formylated for the subsequent attachment of the aniline moiety. However, formylation of unsymmetrical porphyrins is never clean, and we experienced the formation of a mixture of products, formylated at different sites, and their purifcation was extremely cumbersome. Moreover, the grafting of such a molecule may not selfassemble in a similar manner to that used in the present work. Thus, the I-V results of the newly constructed device may provide unreliable results, despite having the reverse geometry. To confrm this hypothesis, the GS geometry of a 5b-congener having the alkenyl attachment at one of the pyrrole moieties in the porphyrin core was theoretically optimized, as done for 5b. The GS geometry of the congener revealed that both the porphyrin and aniline moieties lie on same side (Fig. SL5 †). Therefore, the Si-grafted 5b-congener would be positioned in such a manner that the Hg electrode would preferably interact with the porphyrin core. However, 5b has many possible conformers, and the possibility of touching the mercury electrode by one of its straight conformers cannot be excluded. Nevertheless, based on the I-V results of the TPP C-11 monolayers 54 and that of 5a/5b, it is tempting to propose that the observed current rectifcation is because of the nature of the molecules. ## General Experimental details. Synthesis: all reagents and solvents (Sigma-Aldrich and Fluka) were of synthetic grade. Propionic acid, pyrrole, benzaldehyde, p-nitrophenol were used after recrystallisation. All solvents were dried and distilled before use. Tetrahydrofuran (THF) and hexane were distilled over Na under argon. DMF was dried with CaH 2 and distilled under vacuum. The 1 H NMR and 13 C NMR spectra were recorded with 200/300/500 (50/75/100) MHz spectrometers using deuterated solvents as internal standards. The mass spectrometry was carried out with an MS/MS (410 Prostar Binary LC with 500 MS IT PDA Detectors, Varian Inc, USA) and MALDI-TOF/TOF (Bruker Ultraflex II) data systems. The IR spectra were recorded as flms with a Jasco model A-202 FT-IR spectrometer and only the pertinent bands are expressed. Synthesis 5,10,15,20-Tetraphenylporphyrin (1). To a refluxing solution of benzaldehyde (5.30 g, 50 mmol) in propanoic acid was added pyrrole (3.35 g, 50 mmol) in propanoic acid drop wise. After refluxing for 2 h under stirring, the mixture was cooled to room temperature and left for 12 h. The mixture was fltered, then the precipitate washed with methanol, dried in vacuum, and subjected to column chromatography (silica gel, 50% CHCl 3 / hexane) to obtain 1 (1. 62 38 To a refluxing and stirred solution of 1 (0.500 g, 0.814 mmol) in CHCl 3 (75 mL) was added Cu(OAc) 2 $H 2 O (0.179 g, 0.895 mmol) in MeOH (12 mL). On completion of the reaction (cf. TLC, 30 min) the reaction mixture was brought to room temperature and triturated with MeOH to obtain the corresponding Cu(II)porphyrinato acetate (0.647 g, $quant.) as a purple powder. m.p.: >250 C (MeOH/CHCl 3 ); UV-Vis (CH 2 Cl 2 ) l max [nm]: 302, 416, 540, 576, 617; MALDI-TOF: m/z (%): 795 d. POCl 3 (7.9 mL, 52 mmol) was added drop wise to anhydrous DMF (5.5 mL, 75.6 mmol) at 0 C to obtain the Vilsmeier complex as a thick golden liquid. To this was added a cold suspension of the Cu(II)-porphyrinato salt (0.500 g, 0.73 mmol) in dichloroethane (50 mL). The reaction mixture was brought to room temperature, refluxed for 5 h, cooled to room temperature, and left overnight. Concentrated H 2 SO 4 (10 mL) was added to the ice-cold mixture and stirring continued for 10 min. The green mixture was poured into ice-cold aqueous NaOH (0.625 M, 1 L) with occasional shaking until disappearance of the green colour. The mixture was extracted with CHCl 3 (2 200 mL), then the organic layer was washed with saturated NaHCO 3 ( 4-(5-Hexenyloxy)-1-nitrobenzene 3a and 4-(10-undecenyloxy)-1-nitrobenzene 3b. A stirred mixture of p-nitrophenol (0.500 g, 3.59 mmol), 10-undecenyl bromide (0.920 g, 3.95 mmol) or 5hexenyl bromide (0.645 g 3.95 mmol) and K 2 CO 3 (0.644 g, 5.14 mmol) in dry acetone was refluxed for 12 h. The reaction mixture was cooled, fltered over celite, concentrated and residue dissolved in CHCl 3 (20 mL). The organic phase was washed with H 2 O (2 20 mL) and brine (1 5 mL), and then dried. Removal of the solvent followed by column chromatography (silica gel, 2% EtOAc/hexane) of the residue furnished 3a (0.791 g, 99%) and 3b (1.0 g, 99%) as gels. 3a: d H (600 MHz; CDCl 3 ; Me 4 Si) 8.18 (m, 2H), 6.93 (m, 2H), 5.83 (m, 1H), 5.01 (m, 2H), 4.06 (t, J ¼ 6.0 Hz, 2H), 2.14 (q, J ¼ 6.0 Hz, 2H), 1.84 (quint, J ¼ 6.8 Hz, 2H), 1. 58 4-Hexenyloxyaniline 4a and 4-undecenyloxyaniline 4b. To a stirred mixture of 3a or 3b (0.684 mmol) and HCO 2 NH 4 (0.068 g) in MeOH (5 mL), was added Zn dust (0.054 g, 0.82 mmol) under Ar. After 10 min, the mixture was fltered through celite and washed with Et 2 O (2 20 mL). The organic layer was washed with H 2 O (2 15 mL) and brine (1 5 mL), and then dried. Removal of the solvent in vacuo afforded 4a (0.116 g, 89%) and 4b (0.162 g, 91%) as white powders. The samples turned brown very fast, so were used for the next step without further purifcation. Porphyrins 5a and 5b. 39 A mixture of 2 (0.075 g, 0.12 mmol), 4a or 4b (0.17 mmol), 4 molecular sieves (0.030 g) and AcOH (2 drops) in THF (5 mL) was refluxed until consumption of the starting materials (cf. 3 h). NaBH 3 CN (0.008 g, 0.15 mmol) in MeOH (5 mL) was added into the respective mixtures, which were then refluxed for an additional 3 h. The mixtures were diluted with H 2 O (5 mL) and extracted with CHCl 3 (50 mL). The organic layers were washed with H 2 O (2 50 mL) and brine (1 5 mL), and then dried. Removal of solvent in vacuo afforded residues, which upon purifcation with column chromatography (silica gel, 40% EtOAc/hexane) gave 5a (0.075 g, 78%) as a gel and 5b (0.080 g, 78%) as a solid. Characterization of the monolayers. The monolayers were characterized in terms of thickness, using an ellipsometer (Sentech, model SE 400adv); surface morphology was measured by AFM imaging (Nanonics, Multiview 4000 system), de-ionized water contact angle (Data Physics System, model OCA20), FT-IR (Bruker, 3000 Hyperion Microscope with Vertex 80 FTIR System, LN-MCT 315-025 detector) in polarized ATR mode (20 objective) at an angle of 45 for 500 scans and the data were background corrected with freshly prepared Si-H monolayers. The molecular mass was measured by SIMS (BARC make, Kore's Technology software) keeping Si-H as a reference. The XPS analysis of the deposited flms was carried out using a Mg Ka (1253.6 eV) source and a MAC-2 electron analyzer. The XPS analysis chamber was maintained at a base vacuum of 10 9 mbar. The XPS binding energy scale was calibrated to the Au 4f 7/2 line at 83.95 eV. Preparation of H-terminated Si wafers. n-Type silicon wafers (orientation: 111; resistivity: 0.001-0.005 U cm) and 40% NH 4 F were purchased from Siltronix and Fluka, respectively. The Si (111) wafers were cut into small pieces ($0.5 cm 1.5 cm) and cleaned by heating in a 3 : 1 (v/v) mixture of conc. H 2 SO 4 : 30% H 2 O 2 (piranha solution) for 10 min at 80 C. The wafers were washed with excess H 2 O and immersed successively in deaerated (purged with Ar for 30 min) 40% aqueous NH 4 F for 10 min and 2% aqueous HF for 2 min. The wafers were washed with deionized H 2 O for 1 min, dried under a stream of N 2 and immediately taken into the electrochemical cell for electro-grafting. Monolayer formation. The electrochemical deposition of 5a and 5b was carried out by cyclic voltammetry (CV) with a potentiostat/galvanostat system (model: Autolab PGSTAT 30) using Si wafers as the working electrode (WE), Pt as the counter electrode (CE) and Ag/AgCl as the reference electrode (RE). The solution contained 0.1 M Bu 4 NP as the electrolyte and 5a or 5b (1 mM) in dry CH 2 Cl 2 . The CV was run from 0 to 1 V for 30 cycles at a 0.05 V s 1 scan rate under an inert atmosphere. After the CV scans, the WE was sonicated in CH 2 Cl 2 for 10 min to remove the electrolyte and any unreacted or physisorbed 5a or 5b. The WE was further washed with acetone, isopropanol and methanol to obtain the respective grafted monolayers. Junction and measurement setup. To measure the I-V characteristics, a metal/molecule/Si (n++) structure was completed by using a tiny drop of liquid mercury of diameter 40 AE 2 mm as the counter electrode. The contact area in the grafted monolayer was 0.002 mm 2 . The I-V curves were recorded at room temperature in a dark box using a pA meter-dc voltage source (HP 4140). Theoretical calculations. The ground state geometry optimization and molecular orbital calculations of molecule 5b, TPP, p-aminophenol and the 5b congener were carried out using an ab initio molecular orbital theory based LCAO-MO approach as implemented in the GAMESS software. The ionic optimization of the molecules was carried out without any symmetry constraints at the B3LYP/6-31G(d,p) level of theory. characteristics of the monolayers revealed pronounced, stable and reversible current rectifcation at room temperature in the negative bias. To the best of our knowledge, such high RR values are rare, except for previous devices constructed by C. A. Nijhuis and Whitesides's group, 51,57,58 L. Venkataraman 52 and a recent publication from our own group. 53 The monolayer with the C-11 linker was more compact and showed a 10 5 times high rectifcation ratio (RR) relative to the other similar system having the C-6 linker, possibly because of the compact packing. The rectifcation mechanism was explained on the basis of Aviram and Ratner's theory of rectifcation by using ab initio molecular orbital calculations.

chemsum

{"title": "A novel design for porphyrin based D\u2013s\u2013A systems as molecular rectifiers", "journal": "Royal Society of Chemistry (RSC)"}

synthesis,_structural_characterization_and_photodecarbonylation_study_of_a_dicarbonyl_ruthenium(ii)‐

## Abstract: A photoactivatable ruthenium(II) carbonyl complex mer,cis-[Ru (II)Cl(BisQ)(CO) 2 ]PF 6 2 was prepared using a tridentate bisquinoline ligand (BisQ = (2,6-diquinolin-2-yl)pyridin). Compound 2 was thoroughly characterized by standard analytical methods and single crystal X-ray diffraction. The crystal structure of the complex cation reveals a distorted octahedral geometry. The decarbonylation upon exposure to 350 and 420 nm light was monitored by UV/VIS absorbance and Fourier transform infrared spectroscopies in acetonitrile and 1 % (v/v) DMSO in water, respectively. The kinetic of the photodecarbonylation has been elucidated by multivariate curve resolution alternating leastsquares analysis. The stepwise decarbonylation follows a serial mechanism. The first decarbonylation occurs very quickly whereas the second decarbonylation step proceeds more slowly. Moreover, the second rate constant is lower in 1 % (v/v) DMSO in water than in acetonitrile. In comparison to 350 nm irradiation, exposure to 420 nm light in acetonitrile results in a lower second rate constant. ## Introduction Carbon monoxide (CO)-based therapy represents an exciting new frontier in biomedical research. A pre-condition for successfully achieving the therapeutic benefits from the otherwise toxic CO is the ability to deliver it in controllable dosage, at specific targets. [1b,c,h,2] CO releasing molecules (CORMs) offer the unique possibility of exerting dosage control by first allowing the storage of CO in solid form, and then releasing it only in response to an endogenous or exogenous trigger. [1a,c,e,2-3] In previous works, decarbonylation from CORMs has been triggered by photo-irradiation (photoCORMs), enzymes, pH and thermal changes as well as by ligand exchange reactions. [1c,d,g,2a,3-4] Of these methods, light-triggered decarbonylation is particularly attractive as it offers greater spatialtemporal control. [1f,3-,4c,5] In this regard, several organic and inorganic photoCORMs have been identified. [1e,f, 3-,4c,6] Transition metal carbonyls have featured the most in such studies. [1e,f,3-,4c,6a] This is not at all surprising considering the vastly rich photochemistry of metal carbonyl compounds. A plethora of Mn, Fe, Cr, Re, and W complexes, which contain one or more photolabile M CO bonds, have been evaluated for their aqueous solubility, stability, cellular toxicity and photodecarbonylation characteristics. [1f,g,4c,5-6,7] These studies have concluded that it is indeed possible to tune the CO releasing ability of such scaffolds by varying the ancillary ligand(s) around the metal center. More recently, increasing attention has been given to Ru(II)based photoCORMs. A number of research groups have shown that the Ru(II) dicarbonyl complexes can undergo photoinduced decarbonylation with relatively short half-lives. Using a combination of spectroscopic techniques, theoretical calculations and multi-curve resolution alternating least-squares (MCR-ALS) analysis, we have also demonstrated that the rate of decarbonylation is solvent dependent. [8f] Additionally, the feasibility to chemically link the Ru(II)-based photoCORMs to delivery vectors without compromising on their CO-releasing properties has also been demonstrated. [1d,8a,9] This offers interesting possibilities regarding the development of Ru(II) photo-CORMs as site-directed therapeutics. We have now extended our ongoing investigations on Ru(II) photoCORMs to include π-electron rich quinoline-based systems to overcome the poor photochemistry of terpyridine-based structures. Different strategies have been reported to shift the absorption band to the visible region by extending the π-system, altering the position of the nitrogen in the quinoline ring or adding electron-donating groups. In this article, we report the synthesis, characterization and photodecarbonylation study of a new Ru(II) dicarbonyl complex featuring a tridentate bisquinoline ligand (Scheme 1). In addition to determining the crystal structure of the prepared Ru(II) complex, we have studied its photdecarbonylation behavior using electronic absorption, Fourier transform infrared (FTIR) and NMR spectroscopy as well as MCR-ALS to gain some insight into the effect of the introduced bisquinoline-type π system on decarbonylation activity. ## Synthesis and Characterization The bisquinoline ligand BisQ 1 was synthesized by double Friedländer condensation of 2,6-acetylpyridine with orthosubstituted nitrobenzaldehyde in a one-pot reaction. The Ru (II) complex was prepared by refluxing the key precursor [RuCl 2 (CO) 2 ] n with 1 in methanol under inert atmosphere and exclusion of light to afford the pure mer,cis-[Ru(II)Cl(BisQ)(CO) 2 ] Cl 2 in moderate yield (51 %) (Scheme 1). The chloride counteranion was replaced with a weakly coordinating hexafluorophosphate (PF 6 -) anion by adding saturated KPF 6 solution in methanol to yield the complex as a PF 6 salt. Complex 2 has theoretically two possible stereoisomers (cis-(CO) and trans-(CO)). The latter one is thermodynamically unfavorable and may not exist. Therefore, only the mer,cis-(CO) isomer was isolated (confirmed by X-ray crystallography, as discussed later). The complex was fully characterized by 1 H/ 13 C, 1 H-1 H COSY NMR, ESI-MS, UV/VIS, FTIR, elemental and X-ray crystal structure analysis. The 1 H/ 13 C NMR in acetonitrile-d3 and DMSO-d6, 1 H-1 H COSY NMR, ESI-MS, IR and UV/VIS data of the complex are collected in Supporting Information (Figures S1-S7). Coordination of the ligand to the diamagnetic ruthenium(II) results in five signals, some of which are due to overlapping resonances, in the aromatic region with a slight up-field shift compared to the non-coordinated ligand in the 1 H NMR spectrum (Figure S1). The assignment of the aromatic protons was done by 1 H-1 H COSY NMR in acetonitrile-d3 (Figure S2). The two characteristic 13 CO signals in the 13 C NMR spectrum are observed at 201.0 and 187.8 ppm in acetonitrile-d3 (Figure S1) and at 195.5 and 187.0 ppm in DMSO-d6 (Figure S4). The FTIR spectrum of the ruthenium(II) carbonyl complex (Figure S6) shows two strong carbonyl CO vibrations in the expected region from 2100-1900 cm 1 . The symmetrical stretching mode is assigned to higher (2075 cm 1 ) and the antisymmetrical one to lower energy (2006 cm 1 ). The UV/VIS absorption spectrum of complex 2 was recorded in acetonitrile and 1 % (v/v) DMSO in water (Table 1, Figure S7). In acetonitrile, the ruthenium complex displays four main transitions and a shoulder at ~275 nm; two bands below and two above 350 nm. Interestingly, the electronic spectrum of complex 2 in 1 % (v/v) DMSO in water is similar to that in acetonitrile, except that the shoulder at ~275 nm has diminished and five rather than four distinct bands can be observed, three below 350 nm and two above. The extinction coefficients are quite intense (Figure S7, Table 1), but similar to the reported Ru(II)complexes bearing quinoline ligands. The transitions can be assigned to π-π* (ligand-based absorption) and MLCT/ LLCT. In comparison, only two distinct bands have been reported in literature for the Ru(II) dicarbonyl terpyridine complex 4 (Scheme 2) in DMSO, with the absorption maxima observed below 350 nm. [8a] Thus, the extension of the π-system from having terpyridyl to quinolyl ligand in the coordination sphere results in a bathochromic shift. A single pale, light-green crystal of the monocationic complex 2 suitable for X-ray diffraction analysis was obtained by slow diffusion of chloroform into the acetone solution, over two weeks. The molecular structure of the cationic complex 2 is shown in Figure 1 and selected bond length and angles are listed in Table S1. The crystallographic parameters are shown in Supporting Information (Tables S1-S6). The ruthenium(II) bisquinoline chloro dicarbonyl cation exists as the mer,cis-(CO)-Scheme 1. Synthesis of mer,cis-[Ru(II)Cl(BisQ)(CO) 2 ]PF 6 2. Table 1. Absorption maxima and molar extinction coefficients for complex 2 in acetonitrile and 1 % isomer and adopts a distorted octahedral geometry, which is a result of the bite angle of the tridentate ligand (N(4) Ru N(3)) at 155.85( 9)°. The Ru N(3) and Ru N(4) bond lengths are slightly elongated with 2.130(2) and 2.127(2) , while the Ru N(2) bond length is compressed at 2.017(2) . Moreover, the axial Ru-CO bond length (CO trans to Cl) is shorter (1.887(3) ) than that of the equatorial Ru-CO bond (CO cis to Cl) with 1.908(3) . For comparison, the X-ray crystal structure of a similar ruthenium(II) dicarbonyl compound with a tridentate terpyridyl ligand (see structure 3, Scheme 2) revealed a slightly different situation. The X-ray data of the terpyridyl complex show that the Ru N(3) and Ru N(4) bond lengths are shorter with 2.092(2) and 2.097(2) , while the Ru N(2) bond length is similar at 2.016(2). Interestingly, the reported axial Ru-CO distance (1.873(3) ) is shorter than in complex 2 whereas the equatorial Ru-CO bond length is similar (1.912(3) ). This may influence the decarbonylation properties and hence the kinetics. The longer the Ru-CO bond length, the faster might be the decarbonylation step and the less energy is needed for excitation. Furthermore, the Ru Cl bond distance in 2 is slightly longer than in 3 with 2.3997(7) vs. 2.3917(7) . ## Light-Induced Decarbonylation Monitored by UV/Vis, FTIR and 13 C NMR Spectroscopy UV/Vis absorption spectra of complex 2 was measured over time, in both acetonitrile and 1 % (v/v) DMSO in water (Figure 2), to monitor the photolysis at exposure to 350 nm irradiation. Additionally, photolysis of 2 was examined at 420 nm in acetonitrile. Due to similar absorption profiles of the species formed upon irradiation at 350 and 420 nm, just the changes in the absorption spectrum upon irradiation at 350 nm are depicted in Figure 2 (for 420 nm irradiation cf. Figure S8). Prior to photolysis experiments, the complex was stored in the dark and kept under air in the respective solvent. The sample revealed no spectral changes over a period of 21 h (Figure S7). Furthermore, the experimental setup does not allow recording of spectra with less than 10 s irradiation. Upon irradiation at 350 and 420 nm in acetonitrile, the absorption spectra of 2 showed five bands of increasing intensity (λ max ~255, 316, 348, 430-750) within 3.5 h (Table 1 for comparison with nonirradiated species). Only two new broad bands appeared in the region between 430 and 750 nm. The band at 348 nm (nonirradiated species at λ max = 355 nm) underwent a slight hypsochromic shift (~7 nm). Whereas the band at 375 nm decreased in intensity and the shoulder at ~275 nm diminished within time. No more significant changes in the spectrum occurred after 210 min. Photolysis experiments at 350 nm were also performed in 1 % (v/v) DMSO in water in particular to test the complex under aqueous conditions with regard to in vitro experiments (Figure 2 bottom). The complex showed fewer changes in the absorption profile compared to the photolysis experiments in acetonitrile. The band at 273 nm diminished upon brief exposure to UV light (> 10 s) and a new one formed at ~290 nm. The UV bands at 250 and ~377 nm decreased in intensity and the latter one underwent a slight hypochromic shift (~5 nm), whereas the band at 314 nm increased in intensity. No more significant changes in the spectrum occurred after 4 h. To elucidate the mechanism and kinetics of photolysis, the UV/VIS spectra were analyzed by MCR-ALS analysis, as described previously [8f] and in the experimental section. Kinetic models were determined for the observed changes in the spectra in both solvents. Two consecutive steps and three individual species (A, P1 and P2; Scheme S1) were calculated in acetonitrile at 350 and 420 nm (Table 2). The first reaction occurs very quickly, i. e., the absorption spectra change drastically upon the initial irradiation. Only two spectra were available for the fit and, consequently, the rate constants k 1 are considered lower limits only. The second rate constant (k 2 ) for the photoreaction at 350 nm is a bit higher than k 2 at 420 nm. In 1 % (v/v) DMSO in water three consecutive steps and four individual compounds (Table 2 and Scheme S1) were fitted by MCR-ALS. The fitted spectra and concentration profiles for complex 2 and their photoproducts are shown in Supporting Information (Figures S9 and S10). It should be noted that although the rate constant results from an exponential decay as function of concentration, it is not a unimolar rate constant for the associated photoreaction as such (see Eqs. 1 to 4). It is rather a result of the reaction conditions of this photochemical reaction [1g,18] , which is not defined as the classical rate constant. Thus, it should not be misunderstood as a common rate constant. The rate of a photochemical reaction [18a] is best described as: where c represents the concentration, t time, F the quantum yield and I abs is the absorbed light intensity. Inserting the definition of the absorbance AðlÞ gives: where I 0 is the incident light intensity. When AðlÞ ! 1, the right hand side can be approximated by using Taylor series ((1-10 -A(λ) )ffi2.3 A(λ)) and the absorbance can be replaced using the Lambert-Beer-Law: where ɛ is the molar extinction coefficient (M 1 cm 1 ) and d is the pathlength in a cell (cm). As a consequence, the rate constant resembles a first order kinetics: However, for identical measurement conditions and similar extinction coefficients (Table 1), the determined values of k n are a good representation of the quantum yield and therefore, can be meaningfully discussed instead. Moreover, the UV-light-induced decarbonylation characteristics of the complex in acetonitrile was assessed by FTIR spectroscopy. FTIR spectra of the irradiated complex were recorded after the deposition of an aliquot of an irradiated complex solution onto a silicon ATR accessory and drying it with N 2 gas. Spectral changes (Figure 3) in the characteristic CO region (2150-1850 cm 1 ) were observed upon irradiation of 2 at 350 nm. The intensities of the two CO bands associated with the starting complex diminished over a certain period upon exposure to UV light. One intense new stretching vibration at 1984 cm 1 appeared after 5 min. We assume that the monocarbonyl species [Ru(II)Cl(BisQ)(solvent)CO] + was generated [8f] (see FTIR spectrum after 5 min, Figure 3). The intensity of the CO band associated with the monocarbonyl species progressively diminished with time, indicating the complete decarbonylation within the time course of the experiments. The complete photolysis profile of complex 2 was also monitored by 13 C NMR spectroscopy. As shown in Figure S4, the typical 13 CO signals at δ = 195.5 and 187.0 ppm completely disappear after exposure to 350 nm irradiation for 4 h in a DMSO-d6 solution. 13 C NMR was not possible to measure in deuterated acetonitrile due to the precipitation of the species formed upon irradiation. Nonetheless, the complete decarbonylation of complex 2 is supported by the FTIR spectroscopy measurements in acetonitrile and the 13 C NMR data for the irradiated sample in DMSO-d6 (Figures 3 and S4). In summary, the photoreaction involves a stepwise decarbonylation. The first decarbonylation occurs very quickly, which can be observed by the dramatic change of the absorption profile even upon brief (> 10 s) UV-light exposure. The second decarbonylation step (k 2 @ k 1 ) proceeds much slower and k 2 is even lower when light irradiation is performed at 420 nm compared to 350 nm. In general, the absorption profile of the photodecarbonylation is less pronounced in 1 % (v/v) DMSO in water and process seems to be slower than in acetonitrile. That the dicarbonyl Ru(II) complex 2, where the bisquinoline ligand coordinates in a meridonal tridentate mode, is capable of undergoing complete decarbonylation can be confirmed by FTIR and 13 C NMR spectroscopy. In contrast, terpyridyl ligand in the reported -dicarbonyl Ru(II)CORM complex 5 (Scheme 2) initially exhibits a bidentate coordination mode, which changes into the meridonal tridentate only after the exposure of the complex to UV light and its subsequent monodecarbonylation. ## Conclusions The synthesized ruthenium(II) dicarbonyl complex based on tridentate bisquinoline ligand exhibits beneficial electronic properties, which manifest in a complete decarbonylation after photoactivation at 350 nm and 420 nm, as confirmed by FTIR and 13 C NMR spectroscopy. Furthermore, the conjugated πsystem can be altered and extended, respectively, to actually push the absorption range into the visible spectrum. Anchoring ancillary groups to the skeleton of the bisquinoline ligand gives the possibility to attach biomolecules for targeted therapy. ## Experimental Section All reactions were performed under nitrogen atmosphere using standard Schlenk techniques and assemblies were protected from light if necessary by wrapping them with aluminium foil. Analyticalgrade solvents were degassed by purging with nitrogen for at least 30 min before use if necessary. All solvents were purchased from Sigma-Aldrich or BioScientific and used as received. Ruthenium trichloride hydrate was purchased from Strem Chemicals (#44-5880). 2-Nitrobenzaldehyde (#A11501) was purchased from Alfa Aesar. Iron powder (#12310) and 2,6-Diacetylpyridine (#D8801) were bought from Sigma-Aldrich. Formic acid, 99 % was obtained from Fluka. All solvents were purchased from commercial sources (Sigma Aldrich, Fluka, VWR, Fisher Scientific) without further purification. A Direct-Q 3 UV water purification system from Millipore (Merck KGaA) was applied to produce ultrapure water. The resistivity of the ultrapure water was 18.2 MΩ/cm. Ligand BisQ 2,6-di(quinolin-2yl)pyridine 1 and the polymer [RuCl 2 (CO) 2 ] n , were synthesized according to literature. All characterization data were in agreement with literature reports. NMR Spectra were recorded at 298 K on an Agilent DD2-400 MHz NMR or an Agilent DD2-600 MHz NMR spectrometer with ProbeOne Chemical shifts δ are reported in parts per million (ppm) relative to tetramethylsilane. Coupling constants J are given in Hertz. Abbreviations for the peak multiplicities are as follows: s (singlet), d (doublet), dd (doublet of doublets), ddd (doublet of doublets of doublets), t (triplet), and m (multiplet). UV/VIS Spectra were recorded on a Cary 300 spectrophotometer (Agilent Technologies). Elemental Analysis (C, H, N) was carried out at the Campbell Microanalytical Laboratory, University of Otago, New Zealand. Electrospray ionization mass spectrometry (ESI-MS) was carried out on an Agilent 6120 series Quadrupole LC/MS system. ATR-FTIR Spectra were acquired on a Bruker model Equinox 55 FT-IR spectrometer fitted with a N 2 -cooled mercury-cadmium-telluride (MCT) detector. A Specac golden gate diamond or a Harrick silicon multiple reflection ATR accessory was used for spectral acquisition. The silicon ATR accessory was used to acquire spectra of samples in solvated state. For this purpose, 4 μL acetonitrile solutions of complex (1 mg/mL) were deposited on the silicon ATR followed by evaporation (10 min) under a gentle N 2 flow resulting in a thin film of compound for direct measurement. The spectra were acquired with OPUS software 6.0. ATR spectra were collected in the wavenumber range between 4000 and 400 cm 1 at a spectral resolution of 4 cm 1 , and 50 interferograms were co-added. Preprocessing of the spectral data was performed with the OPUS 7.2 software. ATR-FTIR spectra were baseline-corrected using concave rubberband correction. X-Ray Crystallography. Intensity data for green crystals of 2 (0.75 × 0.25 × 0.05 mm 3 ) were measured at 123 K on a Bruker Apex II CCD fitted with graphite monochromated Mo Kα radiation (0.71073 ). The data was collected to a maximum 2θ value of 50°and processed using the Bruker Apex II software package. Crystal parameters and details of the data collection are summarized in Table S1 in the Supporting Information. The structure was solved by direct method and expanded using standard Fourier routines in the SHELX-97. All hydrogen atoms were placed in idealized positions. A mixture of chloroform and acetone molecule of crystallization was located on the Fourier difference map and refined anisotropically, fixing their respective site occupancy factors as 50 %. All non-hydrogen atoms were refined anisotropically. Photolysis Experiments were conducted using a Rayonet Photoreactor (RPR-200 model) fitted with two Rayonet lamps with an emission wavelength centered at 350 nm (full width at half maximum = 45 nm; E ν � 2.5 mW cm 2 ) and 420 nm (full width at half maximum = 30 nm). A 1 cm quartz fluorescence cuvette (3 mL) was used as the reaction vessel. The power density was measured using a power meter with a built in photodiode sensor (PM200 series) from Thorlabs. The rate of decarbonylation upon exposure of UV light was measured by recording the UV/VIS absorption spectra (200-800 nm) of complex 2 in acetonitrile or 1 % (v/v) DMSO in H 2 O at certain time intervals. All experiments were performed in duplicate. Multivariate Curve-Fitting Analysis. UV/Vis absorbance spectra for all time points were imported into a Matlab matrix. The Matlab Toolbox MCR-ALS 2.0 was used to estimate the number of components by the inbuilt singular value decomposition algorithm and to extract the spectrum and transient concentration development of each component using a MCR-ALS algorithm. The number of components was determined from the number of eigenvalues � 1 and confirmed later by manual variation of the number of components, available time points, and an adapted kinetic model constraint. The initial estimates of the spectra were determined by means of the purest variable detection method. The following constraints were set for the ALS optimization: (i) nonnegativity for all species concentrations and spectra via nonnegative least squares (nnls) and (ii) a kinetic constraint to correlate the different species via a consecutive kinetic model, A!P1…!Pi with rate constant k i (Scheme S1), with the predefined number of species. At the beginning, only one species exists, and therefore only the initial concentration (of species A) was different from zero and set to 35 μM. Estimates for the reaction rate constants were k 1 = 1 min 1 , k 2 = 0.1 min 1 , and k 3 = 0.01 min 1 . No normalization of any spectrum was applied, and convergence was typically achieved in less than 100 iterations with a convergence limit of 10

chemsum

{"title": "Synthesis, Structural Characterization and Photodecarbonylation Study of a Dicarbonyl Ruthenium(II)\u2010Bisquinoline Complex", "journal": "Chemistry Open"}

high_performance_p-type_molecular_electron_donors_for_opv_applications_via_alkylthiophene_catenation

## Abstract: The synthesis of key 4-alkyl-substituted 5-(trimethylsilyl)thiophene-2-boronic acid pinacol esters 3 allowed a simplified alkylthiophene catenation process to access bis-, ter-, quater-, and quinquethiophene π-bridges for the synthesis of acceptor-π-bridge-donorπ-bridge-acceptor (A-π-D-π-A) electron donor molecules. Based on the known benzodithiophene-terthiophene-rhodanine (BTR) material, the BXR series of materials, BMR (X = M, monothiophene), BBR (X = B, bithiophene), known BTR (X = T, terthiophene), BQR (X = Q, quaterthiophene), and BPR (X = P(penta), quinquethiophene) were synthesised to examine the influence of chromophore extension on the device performance and stability for OPV applications. The BT x R (x = 4, butyl, and x = 8, octyl) series of materials were synthesised by varying the oligothiophene π-bridge alkyl substituent to examine structure-property relationships in OPV device performance. The devices assembled using electron donors with an extended chromophore (BQR and BPR) are shown to be more thermally stable than the BTR containing devices, with un-optimized efficiencies up to 9.0% PCE. BQR has been incorporated as a secondary donor in ternary blend devices with PTB7-Th resulting in high-performance OPV devices with up to 10.7% PCE. ## Introduction Bulk heterojunction (BHJ) organic solar cells (OSC), a blend of p-type and n-type conjugated polymers or molecular materials (MM), have attracted significant attention as alternative solar cell technologies as they are light-weight, low-cost and offer the opportunity of cheaper manufacturing employing roll-to-roll printing processes . Recent advances in materials synthesis and device architecture has pushed OSC power conversion efficiencies (PCEs) to 11.5% . Further materials design and device optimizations have been proposed to deliver OSCs with PCEs up to 15% . Although the field has been dominated by polymeric conjugated organic semiconductors, there has been a rapid advance in the development of MMs with PCEs over 10% now reported . The switch to MMs has in part been due to their discrete structure and relative ease of purification, which offers significant advantages, especially reduced batch-to-batch variation . We recently used side-chain engineering, through regioregular placement of hexyl side chains on a thiophene π-bridge , to generate a MM with a planar core structure and enhanced device performance, up to 9.3% power conversion efficiency (PCE) . This material, built from three key building blocks benzodithiophene-terthiophene-rhodanine (BTR), has been shown to have intriguing materials behaviour and excellent device performance when combined with -phenyl C 71 butyric acid methyl ester (PC 71 BM). Maximum PCEs of 9.3% for OSCs containing BTR are achieved after solvent vapor annealing, for devices with an active layer up to 310 nm thick. In this case fill-factors (FF) remain above 70%. However, OSC devices containing BTR are not stable to thermal annealing, a requirement for scale up using common printing processes, where temperatures >80 °C are required for drying or annealing of printed layers . BTR has extremely interesting properties worth further study and leads to three key questions; 1. Synthesis: Can we simplify the synthesis of BTR removing some chromatographic purification steps and use of toxic tin containing Stille condensation reactions? 2. Scale-up: Can we develop a multi-gram synthesis route to facilitate translation to printing programs? 3. Structure-property relationships: Can we modify the BTR chromophore length or alkyl side-chain length thereby improving device thermal stability and device performance? We report here a simplified synthetic route to a series of BTR analogues (Figure 1), where we have varied the chromophore length through the BXR series, where X = monothiophene (M), bithiophene (B), the known terthiophene (T), quaterthiophene (Q), and quinquethiophene (P), respectively and allowing isolation of products on the multigram scale. The simplified synthesis was translated to a second series of products where the oligothiophene sidechain length for the parent (BTR) was systematically varied, i.e. BT x R, where x = 4 (butyl), or 8 (octyl). Incorporation of the BXR series in devices with PC 71 BM has demonstrated that with increasing chromophore length, the thermal stability of the OSC devices increases giving a PCE of 8.9% for BQR after thermal annealing at 120 °C for 10 minutes. We also report an initial result of PCE of 10.7% for ternary blends of BQR with the commercially available PTB7-Th as the donor and PC 71 BM as the acceptor. ## Results and Discussion Synthesis: Our modified synthesis of BTR and its analogues starts with the lithiation of 3-alkylthiophene 1a-c by lithium diisopropylamide formed in situ from the reaction of n-butyllithium with diisopropylamine (DIA) in the presence of the alkylthiophene, followed by quenching with trimethylsilyl chloride to generate the previously unreported 4-alkyl-2-(trimethylsilyl)thiophenes 2a-c, which could be purified by distillation to ensure removal of unreacted 3-alkylthiophene, Scheme 1. Deprotonation of 2 with n-butyllithium and reaction with 2-isopropoxy-4,4,5,5-tetramethyl-1,3,2-dioxaborolane Scheme 1: Synthesis of the key intermediates TMS-T x -BPin (3), i) diisopropylamine (DIA), THF, n-BuLi, −78 °C TMS-Cl, ii) n-BuLi, iPrOBPin, THF, −78 °C. Scheme 3: Synthesis of the bithiophene through palladium catalyzed direct arylation, a) i) Pd(OAc) 2 , PCy 3 , PivOH, K 2 CO 3 , toluene 100 °C, 4 h, 1%, ii) Pd(OAc) 2 , PPh 3 , K 2 CO 3 , DMF 120 °C, 6 h, 10%, iii) Pd(OAc) 2 , dppp, KOAc, DMAc 120 °C, 5 h, 32%, iv) Pd(OAc) 2 , dppb, KOAc, DMAc 120 °C, 5 h, 32%. (iPrOBPin) resulted in formation of the key intermediates 3a-c, after distillation, in high yield of 60-70% (see refs for recent similar chemistry). Intermediate 3b has been scaled to the mole scale with no issues noted. With 3a-c in hand, synthesis of the required series of alkyl substituted oligothiophene π-bridges by simple Suzuki-Miyaura cross-coupling reactions could be completed. Starting with commercially available 5-bromothiophene-2-carboxaldehyde and then reaction with 3a-c to generate the required bithiophenes, then terthiophenes, while further catenation with 3b resulted in synthesis of the hexyl-substituted quater-and quinquethiophenes 8b and 10b, Scheme 2. Conversion of the intermediate TMS-protected oligothiophenes to the iodo-oligothiophenes (n =1 and 2) was achieved using iodine monochloride (ICl), however for n = 3 and 4 a number of side reactions leading to unidentified side products significantly reduced the yield. For the quater-and quinquethiophenes N-iodosuccinimide (NIS) was used to give a clean product. Ideally, the synthesis could be further simplified by direct palladium-catalyzed CH-activation, arylation of 2 followed by reaction with the commercially available 5-bromothiophene-2carboxaldehyde to generate the bithiophene 4, Scheme 3. It has previously been reported that direct coupling of 2-(trimethylsilyl)thiophene with aryl halides proceeds in good yield with protodesilylation being the major side reaction under the reaction conditions, even at short reaction times . An initial reaction screening, investigating ligand, base and solvent variation, showed positive results with up to 32% yield of the required bithiophene 4. We are currently examining catalyst opti- mization to improve the yields of this simplified route to the required oligothiophenes. To avoid large scale use of tin reagents we required the key bisborylated benzodithiophene (BDT) core 13, which was synthesised from the known BDT core 12 using iridium catalyzed borylation via CH-activation. The bis-borylated product was isolated by precipitation on addition of isopropanol (IPA), and an analytically pure material isolated by filtration in excellent yields >90%, Scheme 4. This simplified purification is in direct contrast with reported procedures for the bis-iodinated or bisstannylated analogues . . BQR can then be compared to BTR where, surprisingly, a single phase change is seen on heating, while three phase changes are observed on cooling, Figure 2a. Even on slowing the heating rate to 0.1 °C per minute no change in the single phase change on heating was observed. The three phase changes at 190 °C, 180 °C, and 164 °C on cooling appear to be analogous to that seen in BTR. Modification of the BTR oligothiophene alkyl chain lengths in the BT x R series results in an intuitive change in the temperatures of the relevant phase transitions, with an inverse correla-tion observed between alkyl chain length and the specific phase change temperatures. Interestingly, as with BQR the DSC traces of the BT x R analogues reveal markedly different phase behaviour relative to that of BTR. BT 4 R has a single endothermic (206 °C) and a single exothermic peak (199 °C) that are higher than the phase transitions in BTR. Two exothermic peaks at 148 °C and 182 °C are observed in BT 8 R, and two endothermic peaks are recorded at 100 °C and 166 °C. As can be observed, even these small changes in alkyl chain length result in a significant impact on the phase change behaviour. However, unfortunately no correlation can be made at this stage between subsequent thermal stability of OPV devices and the phase transition of the BXR and BT x R materials. ## Polarized optical microscopy (POM). POM was utilized in conjunction with a heating platform to directly observe these phase transitions and elucidate thin film structure. On heating at 10 °C•min −1 BQR shows a single phase transition to the isotropic melt at 202 °C, while on cooling we have identified an initial transition to a high-temperature NLC phase at 190 °C, and then a crystallisation at 180 °C. On further cooling a thermochromic phase change is observed at 164 °C, see Figure 2 (and UV-vis discussion below). Even with the much slower cooling rates used for POM studies we did not observe more than the single phase change on heating the BQR sample. We have repeated POM studies on the new batches of BTR and they are identical to those reported; see Supporting Information File 1, Figure S8.2 . Upon examination of BT 4 R with POM a single phase change on heating was observed, with a highly crystalline state below 205 °C giving way to an isotropic melt at 206 °C. Surprisingly, when the sample was cooled a characteristic NLC was observed at 199 °C, with a change to its crystalline state at 193 °C, Figure 3. This NLC transition was not observed in the DSC, even with a slowed cooling rate. The thermal transitions in BT 8 R are not as defined as in the other cases with two broad endothermic (148 °C and 182 °C) and exothermic transitions (166 °C and 100 °C) (Supporting Information File 1, Figure S8.5). No high temperature NLC phase was observed when examined under POM. The dramatic changes in the thermal behaviour and phase change properties for the BX x R series materials are induced by either changes to the chromophore length, or by altering the side chain length on the oligothiophene. These changes have a dramatic impact on the presence or absence of a NLC phase in these materials. All, however, show the appearance of long needle like crystal forms in the POM images obtained. UV-vis and fluorescence spectroscopy. Solution and thin film UV-vis absorption profiles of the BX x R are shown in Figure 4 and Figure 5, respectively, with selected data collected in Table 1 (all spectra can be found in the Supporting Information File 1). The members of the series all exhibit absorption maxima between 450-600 nm in chloroform. In solution it is evident that, although all the materials have a similar onset of absorption at around 600 nm, the peak maxima progress in the reverse order to that expected with BPR having a maximum absorption at 490 nm, while BMR has a maximum absorbance at 541 nm, Table 1. While the BPR and BQR spectra show broad featureless peaks, the spectra for BMR (and BBR) show more complicated structure indicating possible association in solution with the development of strong aggregates. The UV-vis spectra for thin films of the BX x R series, cast from chloroform and subject to both solvent vapour annealing (SVA, THF 20 s) or thermal annealing (TA, 120 °C 10 min, N 2 ), are shown in Figure 5. On increasing the conjugation length from BMR to BPR the expected red-shift in the absorption peaks is now evident (Figure 5a). However, λ max is dependent on the degree of formation of π-π stacking and development of the lowest energy transition with two clear sharp peaks at 552 nm and 590 nm (λ max ) for BMR, while for BPR λ max is at 594 nm with a shoulder at around 630 nm indicating poor formation of the aggregates (Figure 5d). Side-chain modification from butyl to octyl in the BT 4 R, BTR and BT 8 R series also impacts on the thin film formation as seen in as-cast films with poor development of the crystalline order in BT 4 R and BT 8 R, as well as a λ max blue-shift of both BT X R analogues, when compared to BTR, Figure 5a and d. Crystalline order develops for all films after SVA or TA, (Figure 5b and c). While after annealing λ max corresponds to the lowest energy band for most of the thin films, or the two peaks are close in intensity, BPR is the exception where the long wavelength absorption is a shoulder that is not well resolved and λ max corresponds to the higher energy band. In both cases BQR shows well-ordered films with the largest red shift and a λ max at 625-630 nm. After annealing the absorption profiles of BT 4 R, BTR and BT 8 R are almost identical with only a small change in the intensity of the peak at around 570 nm, therefore indicating that the underlying packing structures are not significantly altered through side-chain substitution. To better understand the annealing process UV-vis spectra have been replotted for each material, see Figure 6 (and Supporting Information File 1, Figures S9.1 and 2). As each material is annealed a small blue shift, 10-20 nm, is seen in most spectra with a concomitant increase in the prominence of the low energy peak. The shift is smallest in BPR (Figure 6c), and no significant change is seen for BT 4 R (Supporting Information File 1, Figure S9.2f). It is evident that subtle changes in molecular orientation and packing, with a tendency to H-aggregate formation, are present, however, further work is being undertaken to better understand the underlying processes leading to these changes. Fluorescence emission spectra were collected using the same films as those used to collect the UV-vis spectra above, with selected graphs shown in Figure 7, and extracted data in Table 2. The full spectra are presented in Supporting Information File 1, Figure S9.3. The as-cast films do not contain simple symmetric emission bands, indicating a significant level of structural complexity in the as-cast thin films. BMR shows two emission peaks at 670 nm and 704 nm with a long tail at around 800 nm. A number of absorption peaks are evident for BMR, Figure 5a and d, and these may represent the multiple environments for emission. BBR, BTR and BQR show surprisingly asymmetric peaks with long linear tails from a peak maximum located at around 715 nm. Following SVA a broad, more sym- a Taken from ref. . The absorption spectrum of BQR collected on the POM heated stage shows and extra shoulder located at around 730 nm that was not present in the as-cast films of BQR or the thermally annealed films (measured at room temperature). The BQR thin films were annealed up to 220 °C to probe the effect on the UV-vis spectrum (spectrum collected at room temperature after cooling) of cycling the BQR thin film up to the NLC phase change temperatures (Figure 10) but there is no appearance of the new shoulder, however as the films are rapidly cooled there may be a kinetic effect (see below). Variable temperature fluorescence emission spectra were recorded on BTR and BQR (Figure 9) using a similar setup as for the UV-vis measurements (Figure 8), but employing the Hg fluorescence excitation lamp of the microscope. However, the lamp used introduced a significant thermal load on the sample (approx. 18 °C) and therefore the apparent phase change temperatures are offset relative to the absorption data for this experiment. The data for BQR are shown below, while the data for BTR are included in Supporting Information File 1. At room temperature (after cycling once) BQR has two emission peaks at around 750 nm and 690 nm. On heating the low energy peak reduces in intensity with a concomitant increase in the peak at 690 nm and a blue shift to 655 nm at the sample melting point to the isotropic phase (Figure 9a). A similar shift is observed on cooling the sample, Figure 9b. The emission peaks appear to reflect the two absorption peaks observed in the variable temperature UV-vis absorption spectra, however the underlying structural changes remain unclear and are the subject of further structural studies on BTR and BQR thin films. The appearance of the second fluorescence emission peak in variable temperature spectra on the POM stage, again not seen on the thin films (Figure 7c), has been examined in more detail. The thin films used to obtain the UV-vis spectra, were heated to above the NLC phase change temperature, and collected fluorescence emission spectra are shown in Figure 10. It is clear that when the thin films are heated to 220 °C, i.e., above the NLC phase change temperature, a second blue shifted peak appears at 730 nm, however this is not at the same position recorded on films heated on the POM stage (690 nm). The thin films for these tests were heated to the annealing temperature and after the set time the microscope slide was removed, causing rapid cooling (220 °C fast in Figure 10). It is possible that the rate of heating/cooling impacts on the crystallization of the thin films, with the slower rates used for the fully enclosed, temperature ramped POM stage, leading to equilibrium phases, while rapid thermal quenching of isolated thin films on glass slides gives different results. To further probe this effect, the cooling rate for the thin film heated to 220 °C was modified by turning off the hotplate and allowing the thin film to cool slowly (220 °C slow in Figure 10). When the film is left to cool at a slow rate (220 °C to room temperature over 45 min) the emission spectrum is an almost perfect replica as for the as-cast film emission. Detailed variable temperature X-ray analysis of BQR thin films is currently underway to better understand these changes. ## CV and PESA: The electrochemical properties of the BXR series of materials have been examined by cyclic voltammetry (CV), photoelectron spectroscopy in air (PESA), and UV-vis to determine approximate energy HOMO-LUMO energy levels, and the data are summarized in Table 3. From the UV-vis absorption onsets, determined from the as-cast thin films, we determined the E g(opt) levels, which demonstrate a clear trend in the reduction of E g(opt) on increasing the conjugation length in the BXR series dropping from 1.92 eV to 1.74 eV. Pseudo reversible oxidation potentials in the CVs have been recorded on thin films for each of the materials and the data are listed in Table 3, (see Figure S7.1 Supporting Information File 1 for the cyclic voltammograms). The CV data show the expected gradual increase in HOMO levels as we increase the conjugation length and the expected general downward trend in energy gap from 2.40 eV to 1.77 eV for the BXR series. The first reduction potential could also be measured allowing an estimation of the LUMO levels for the BXR series and therefore an electrochemical energy gap (E g(CV) ), and these match well to the E g(opt) values reported above and listed in Table 3. Ionisation potentials have also been measured by photoelectron spectroscopy in air (PESA), and give a direct measure of the DFT calculations. To further understand the impact on varying the conjugation length of the oligothiophene bridging arm on the distribution of the HOMO/LUMO energy levels and overlap, density functional theory (DFT) calculations were performed. Geometry optimization and molecular orbital surfaces were determined and are shown in Figure 11. Geometries of the BXR series were obtained at the D2 dispersion corrected B3LYP/6-311G(d,p) level of theory. Subsequent time-dependent DFT (TD-DFT) calculations were carried out on the optimized structures with PBE0/def2-TZVP level of theory based on our benchmark calculations (Supporting Information File 1, chapter S11). It is apparent in Figure 11 that as the BXR molecular materials increase in size the overlap of the HOMO and LUMO decreases. The HOMO of the BXR series extends as the number of the thiophene rings increases. In contrast, the LUMO becomes more localized towards the N-hexylrhodamine acceptor moiety as the conjugation length increases. The calculated HOMO values and HOMO-LUMO energy difference follows the same trend as the observed values determined by CV and PESA. Photovoltaic performances. The BX x R series of materials were incorporated into bulk heterojunction devices with a conventional architecture, i.e. ITO/PEDOT:PSS/active layer/ Ca/Al (Figure 12a). The active layer composition was held at 1:1 BX x R:PC 71 BM (by weight) and deposited from CHCl 3 . The active layer was ≈250 nm thick and Ca/Al was used as the back cathode. We report here preliminary BHJ device data to indicate the impact of small structural variations on the device performance. Further device optimization is currently being completed and will be reported at a later date. Devices assembled with as-cast films (Table 4, entries 1-5 and J-V curves Figure 12b) show acceptable device performance without annealing, with BQR delivering the best device performance at 5.3%. All devices show high open circuit voltages (V oc ) above 0.90 V, but low fill factors (FF) <45%. Using previously optimized SVA conditions for BTR, THF for 10 seconds, devices based on the BXR series were fabricated and device data collected, see Table 4 (entries 6-10) and J-V curves in Figure 12c. The V oc decreases from 1.04 V for BMR to 0.82 V for BPR as the conjugation length increases. The V oc drop reflects the measured increase in the HOMO level across the series from −5.51 eV for BMR to -5.02 for BPR. The measured device data for BTR with J sc = 13.9 mA cm −2 , V oc = 0.92 V, FF of 72% and PCE of 9.3% are almost identical to those previously reported at J sc = 13.9 mA cm −2 , V oc = 0.90 V, FF 74.1% and PCE of 9.3% , showing the batch to batch reproducibility in device data for molecular materials. Except for BMR, the FFs for SVA devices lie above 70%, indicating excellent morphology development. The best device contains BQR with a PCE of 9.4% and a J sc = 15.3 mA cm −2 . BPR shows promise with a high FF (74%), however a lower V oc (0.82 V) and a reduced J sc (14.3 mA cm −2 ) reduce the PCE to 8.7%. UV-vis data indicate that under these SVA conditions the π-π stacking is not fully developed indicating that optimizing SVA conditions may lead to improved light harvesting. It is important for commercialization of printed BHJ devices that any active layer can withstand the requirements of a printing process, which normally requires a drying or curing step for printed electrodes of >80 °C. To evaluate our new materials for possible translation to a printing process devices incorporating the BXR series of donors were assembled and the active layer thermally annealed at 120 °C for 10 minutes before electrode deposition, data collected from the devices are listed in Table 4 (entries 11-15) and J-V curves are shown in Figure 12d. The thermally annealed devices do not show as clear a trend as seen for SVA annealed devices with, e.g., no clear systematic decrease in the V oc s on going from BMR through to BPR. Also, apart from BPR, device FFs remain below 70%. This suggests that further device optimization is required. The device performances of BMR (PCE 1.1%), BBR (PCE 3.3%), and BTR (PCE 5.7%) are significantly lower than the SVA devices, primarily due to lower FFs and J sc values. Both BQR (PCE 8.9%) and BPR (PCE 8.1%) do not show significant performance loss after thermal annealing, maintaining good FF's, J sc values and V oc 's. The drop in performance compared to the SVA devices indicates that further optimization may be required. It is evident that modification of the chromophore length has a large impact on the device stability and performance. BQR as a molecular electron donor is the stand-out performer with the best initial results under all device assembly conditions, and shows thermal stability compatible with printing processes. The influence on the oligo-thiophene alkyl chain length on molecular packing, and thereby device performance, was examined in the BT x R series. BHJ devices using BT 4 R and BT 8 R were assembled using the same device architecture described above. The collected device data are summarized in Table 5, and the J-V curves are shown in Figure 13. Examination of the BT x R UV-vis data for as-cast films (Figure 5d) indicates that BT 8 R does not have a well-developed π-π stacking peak in as-cast films, unlike BTR. Also, both BT 4 R and BT 8 R are blue-shifted in comparison to BTR, by 18 nm and 26 nm respectively for BT 4 R and BT 8 R. As it is not expected that modifications of the oligothiophene bridge side-chain length should significantly impact the chromophore energy levels, variations in measured properties will be due to impacts of side-chain variation on intra-/intermolecular interactions. The differences are reflected in the performance of BT 4 R and BT 8 R containing devices, Table 5, entries 1-3 and the J-V curves reproduced in Figure 13b, where the device efficiency for BTR at 4.6% PCE remains above that for BT 4 R (3.8% PCE) and BT 8 R (2.4% PCE). The major change is a significant drop in short circuit current for BT 8 R down to 5.7 mA cm −2 , from over 10.3 mA cm −2 for BTR and BT 4 R. The open circuit voltage is also lower for both BT 4 R and BT 8 R in comparison to BTR, however there is no obvious trend. After SVA the UV-vis spectra for BT 4 R and BT 8 R match more closely that for BTR, however the π-π stacking peak remains poorly resolved for BT 8 R. Again this is reflected in the lower device performance for BT 8 R (5.2% PCE) in comparison to BTR (9.3% PCE) and BT 4 R (9.0% PCE). In fact, the device parameters for BT 4 R are almost identical to those for BTR, except for a significant drop in V oc to 0.88 V from 0.92 V. One can only speculate on the cause of the V oc drop until further structural characterisation of the thin films is completed. The poor J sc and FF for the BT 8 R devices indicates a poor development of morphology and indicates that devices optimization is still required. The performance of BQR as a molecular electron donor and the stability of BQR containing BHJ devices encouraged the examination of BQR in ternary BHJ devices. It has been reported that addition of a small percentage of a molecular electron donor to polymer:fullerene BHJ devices leads to an improve- ment in overall device performance . The underlying reason for the improved performance in these ternary devices is not yet clear with a combination of favourable morphology, energy level cascading and recombination in the ternary blend being suggested , however the performance enhancement is real and reproducible. Ternary blend devices containing BQR have been assembled using poly [4,8- BQR and PTB7-Th, if this is important. Inverted devices with a structure ITO/ZnO/PTB7-Th:BQR:PC 71 BM/MoO 3 /Ag were assembled, and device data are collected in Table 6 and J-V curves are shown in Figure 14c. Devices were spun cast from chlorobenzene containing 3% diiodooctane as a processing additive. When PTB7-Th was as the polymeric donor with our standard inverted device architecture and processing conditions, we were able to assemble BHJ devices with a PCE of 9.6%. For these PTB7-Th only devices we achieved a good J sc =17.2 mA cm −2 and a FF of 69% with the expected V oc of 0.80 V for devices containing PTB7-Th (Table 6, entry 1). These results compare very well with previously reported devices containing PTB7-Th:PC 71 BM as the active layer, with a similar simple device architecture (see for example ref , J sc =17.23 mA cm −2, FF 63.42%, V oc of 0.793 V, and a PCE 8.81%, 1:1.5 PTB7-Th:PC 71 BM). Inclusion of 15 wt % of BQR in the donor phase resulted in a significantly enhanced efficiency from 9.6% to 10.7% PCE, with increase in the device J sc to 19.8 mA cm -2 , while the V oc and FF remain effectively unchanged. To investigate the enhanced performance due to addition of BQR, we have studied the absorption spectra (Figure 14d) of the ternary and binary blend films. The normalized absorption intensity of ternary blend (PTB7-Th:BQR:PC 71 BM) active layer shows the enhanced absorption intensities in the wavelength range between 500-600 nm comparing to the binary blend (PTB7-Th:PC 71 BM). This enhanced absorption in the ternary blend is due to the inclusion of BQR as confirmed from the absorption spectrum of BQR: PC 71 BM, which shows the absorption maximum in the wavelength range 500-600 nm. Further device optimization and active layer morphology investigation for enhanced performance of ternary blend OPV devices are being explored currently and the results will be communicated elsewhere. ## Conclusion We have reported a simplified synthesis of alkylsubstituted oligothiophenes used as π-bridges in A-π-D-π-A molecular electron donors via chain extension catenation of alkylthiophenes. We have used commercially available 3-butyl-, 3-hexyl-and 3-octylthiophene to form the key intermediate TMS-alkylthiophene boronic acid pinacol esters (3) in high yield on a large scale and in high purity as they can be purified by distillation. Access to the mono-, bis-, ter-, quater-, and quinquethiophene π-bridge oligothiophenes by alkylthiophene catenation has allowed the synthesis of chromophore extended versions of the previously reported BTR, the BXR series of materials, that is BMR (X = M, monothiophene), BBR (X = B, bithiophene), the known BTR (X = T, terthiophene), BQR (X = Q, quaterthiphene), and the BPR (X = P, quinquethiophene). The impact of the oligothiophene alkyl side-chain on OPV device performance was studied using the 3-butyl and 3-octylthiophene starting materials to generate the BT X R analogues with butyl-and octyl-substituted oligothiophene π-bridges, the BT x R series of materials, where x = 4 (butyl) and x = 8 (octyl). Thin films of the pure materials have been analysed by UV-vis absorption spectroscopy which indicated that extension of the BXR chromophore through oligothiophene extension and side-chain variation impacts significantly on the development of highly π-π stacked materials. Shorter chromophore length leads to good stacking in thin films with dominant absorption transitions even in as-cast films for BMR, BTR and BQR. Molecular organization is improved in all films with SVA, except for the longest chromophore (BPR), where good π-π stacking is not observed, even on extended thermal annealing. Small changes in the molecular structures lead to larger impacts on the thermal behaviour of the materials. DSC thermograms for materials indicate that short and long chromophores (BMR, BBR, and BPR) show single phase changes, while BTR and BQR show a number of phase changes and high temperature NLC phases. Surprisingly when BT 4 R was examined using DSC a single exothermic and endothermic peak were recorded, however when the materials were examined by POM a high temperature NLC phase was observed indicating a rich and more complicated phase space than indicated by simple thermal analysis. For BQR, variable temperature UV-vis spectroscopy mapped the transmitted light variations that accompany the phase change noted in the POM studies of BQR. We are currently studying the structural changes occurring in thin films of our BX x R materials to better understand the changes observed and these will be reported in due course. All the new materials have been tested as electron donors in OPV devices with PC 71 BM as the acceptor with thick active layers (approx. 250 nm). The preliminary results show interesting patterns with good OPV device performance for both solvent vapour and thermally annealed devices, up to 9.4% PCE. Device performance improved with chromophore extension in SVA devices increasing from 3.5% PCE for BMR containing devices through to 9.4% PCE for BQR. The results indicated an improved performance for BQR over that for BTR previously reported, 9.3% PCE, also reproduced for materials made in this study with our modified procedure. Further extension of the chromophore length in these preliminary studies, for BPR, results in a lower PCE of 8.7%, mainly due to a lower J sc and V oc . However, again with a FF of 74% there is scope for device improvements through more optimization. Device performance improved with chromophore extension in TA devices, increasing from 1.1% PCE for BMR through to 8.9% PCE for BQR. The results indicated an improved thermal stability for OPV devices containing BQR over that for BTR previously reported. Incorporating BPR in OPV devices, with the longest chromophore length in this study, also resulted in thermally stable devices, but with a lower PCE of 8.1%, mainly due to a lower J sc and V oc . However, with a FF of 71%, the highest in this thermally annealed series, there is again scope for device improvements through more optimization. In an extension of these studies, we have used the best material (BQR) as a secondary donor in ternary blend devices with commercially available PTB7-Th as the main polymeric donor. In initial studies using these ternary blends we have recorded OPV device efficiencies of up to 10.7% PCE. The improved efficiency in these devices is a result of a significantly higher J sc , rising from 17.2 to 19.8 mA cm −2 , with no significant change in V oc or FF. Therefore, we have shown using a simplified synthesis that chain extended chromophores can be accessed, and thereby the thermal stability of OPV devices containing these new materials can be improved. We are currently examining BQR in printed solar cells. In all cases in our structure-property relationship studies, devices incorporating materials that exhibited a high temperature NLC phase gave the best results. The role of the high temperature NLC behaviour in device performance remains unclear, and as we do not anneal to temperatures where the NLC phase change temperature is reached, its presence is unlikely. However, it may be that structural properties leading to a high temperature NLC phase may help to pre-organise the donor material into a morphology best suited for OPV devices. We are currently probing the structure of these materials in thin films, and these results will be published in the near future, along with device optimization studies and translation to large area devices. In summary, we have developed a simplified synthetic route to afford a range of MMs analogues of BTR. This simplified route has allowed large-scale synthesis of intermediate building blocks and of a multi-gram synthesis of the required MMs. Detailed structure-property studies have identified BQR and BPR as excellent materials for further optimization with an improved performance over BTR. OPV devices containing BQR or BPR show a good thermal stability at 120 °C for 10 min, maintaining a high PCE (BQR, 8.9% and BPR, 8.1%) and FF (BQR, 65% and BPR, 71%). These are promising results for high performance OPV devices and the translation to large area and printed OPV devices.

chemsum

{"title": "High performance p-type molecular electron donors for OPV applications via alkylthiophene catenation chromophore extension", "journal": "Beilstein"}

the_impact_of_the_polymer_chain_length_on_the_catalytic_activity_of_poly(n-vinyl-2-pyrrolidone)-supp

## Abstract: Poly(N-vinyl-2-pyrrolidone) (PVP) of varying molecular weight (M w = 40-360 kDa) were employed to stabilize gold nanoclusters of varying size. The resulting Au:PVP clusters were subsequently used as catalysts for a kinetic study on the sized-dependent aerobic oxidation of 1-indanol, which was monitored by time-resolved in situ infrared spectroscopy. The obtained results suggest that the catalytic behaviour is intimately correlated to the size of the clusters, which in turn depends on the molecular weight of the PVPs. The highest catalytic activity was observed for clusters with a core size of ~7 nm, and the size of the cluster should increase with the molecular weight of the polymer in order to maintain optimal catalytic activity. Studies on the electronic and colloid structure of these clusters revealed that the negative charge density on the cluster surface also strongly depends on the molecular weight of the stabilizing polymers.The catalytic activity of gold nanoclusters (AuNCs) strongly depends on their size 1 , whereby smaller clusters usually exhibit higher catalytic activity due to quantum size effects. On account of the charge distribution on the gold surface, the electronic effects are also more advantageous in AuNCs with smaller cores 2 . The catalytic activity of AuNCs is also strongly affected by their interfacial environment. Previous studies have demonstrated that poly(N-vinyl-2-pyrrolidone) (PVP)s are well suited to stabilize the surface of AuNCs (Au:PVP) 3 , and that the interactions via the carbonyl group do not affect the catalytic activity of these nanoclusters 2, 4 . In general, ligands that strongly coordinate to the AuNC surface, such as thiols, inhibit the catalytic activity of the clusters 5, 6 . However, weakly adsorbing stabilizers such as PVP represent an ideal stabilizing polymer system for AuNCs. In addition to its stabilizing function, PVP may also act as an electron donor, and the coordination of PVP to the gold surface has already been investigated [7][8][9][10] .Previous experimental and theoretical studies have shown that the interaction between a stabilizing polymer and AuNCs leads to a donation of negative charges onto the AuNC surface, which is important for catalytic aerobic oxidations involving molecular oxygen (Fig. 1) 11,12 . The increased electron density on the AuNC surface leads to an enhanced catalytic activity via the formation of catalytically active superoxo species, which are generated by an electron transfer from the high-lying LUMO of the anionic AuNC to the π*-orbital of the adsorbed dioxygen molecule 2 .The fact that the stability of nanoclusters can be controlled by the molecular weight of the stabilizing polymer is well established, whereby polymers with lower molecular weight usually afford less protection against agglomeration [13][14][15] . Although several reports discuss the stabilization of AuNCs by polymers as a function of the molecular weight, morphology, and rheology , studies on the effect of the molecular weight of the polymers on the catalytic activity of the AuNC surface still remain elusive. This is most likely due to the assumption that AuNCs that are stabilized by the same polymer should exhibit comparable activity on account of the comparable environment of the metal surface. We have previously reported a preliminary investigation on the effect of the PVP chain length on the homocoupling of phenylboronic acid. AuNCs stabilized by PVP (K-30; M W = 40 kDa), i.e., Au:PVP (K-30) exhibited higher catalytic activity and selectivity relative to AuNCs stabilized by PVP (K-90; M W = 360 kDa), i.e., Au:PVP (K-90). Nevertheless, it should be noted that on the basis of this report, it is impossible to compare the reactivity between polymers of different chain length, as AuNCs of the same size are required for the suppression of the size effect. In the case of high-viscosity PVP (K-90), AuNCs (1.3 nm) could not be obtained from the conventional batch method 18 . However, the size-controlled synthesis of AuNCs has meanwhile developed further, and a wide range of AuNCs (1-9 nm) stabilized by PVPs of different molecular weight is now available 20,21 . Therefore, we conducted intensive investigations on the correlation between the catalytic activity of Au:PVP and the PVP chain length. Moreover, we examined the structures of AuNCs in order to determine how these are affected by their interaction with the polymer matrix. ## Results Size matters, but 'smaller' is not necessarily better than 'larger'. AuNCs of varying size, stabilized by PVPs of varying molecular weight, were prepared according to previously reported procedures (Fig. S1-S3 in Supplementary Information) 18,20,21 . The size-dependent AuNC-catalyzed oxidation of 1-indanol (Fig. 2a) was selected due to its very short reaction times, the absence of potential alternative pathways, and the possibility to monitor the reaction in situ 22,23 . This allows avoiding the agglomeration of AuNCs during the reaction, and hence only the catalytic activity of AuNCs is determined. To quantify the matrix effect as well as the size-dependence on the catalytic activity, we examined the aerobic oxidation of 1-indanol (1a) catalyzed by AuNCs containing PVP of varying molecular weight, i.e., PVP (K-30) (M W = 40 kDa), PVP (K-60) (M W = 160 kDa), and PVP (K-90) (M W = 360 kDa). The progress of this reaction was monitored in situ using a fourier transform infrared (FTIR) spectrophotometer with an attenuated total reflection (ATR) accessory unit, which allowed recording the C = O stretching frequency of the reaction product 1-indanone (2a). The IR spectra were collected in intervals of 27 seconds, and the obtained time-resolved IR spectra are shown in Fig. 3. The AuNC size and the properties depending on the length of the PVP chain are summarized in Table S1 and Fig. S4 (in Supplementary Information), wherein the normalized rate constant k norm refers to the reaction rate per unit of surface area which can be obtained by assuming a spherical shape (Supplementary Information) 1,24 . The size-dependent properties of Au:PVP (K-30) exhibited a trend similar to that observed in the reaction of p-hydroxy benzyl alcohol: 1 smaller catalysts promote the reaction faster than larger catalyst (Fig. S5 in Supplementary Information). After normalizing the rate constants, the catalytic activity of larger clusters was slightly elevated. In case of Au:PVP (K-60), a similar result was obtained, although the trend was found to shift in favor of larger clusters. However, the size-dependence of Au:PVP (K-90) was substantially different to those of Au:PVP (K-30) and Au:PVP (K-60). For smaller clusters (core size < 2 nm), Au:PVP (K-90) showed inferior activity compared to Au:PVP (K-30), while the activity increased drastically upon increasing the particle size in the range of 0.8-7.0 nm (Fig. 4). The normalized rate constant of Au:PVP (K-90) (core size: 7 nm) was five times higher than that of Au:PVP (K-30) (core size: 1.3 nm), which was previously used as a benchmark catalyst. Considering the superior catalytic activity of Au:PVP (K-90) (core size: 7 nm) in the catalytic aerobic oxidation of 1a, we wanted to find out if the observed reactivity is substrate-specific. For that purpose, we examined other typical AuNC-catalyzed reactions, e.g. the aerobic homocoupling of organoboron compounds such as potassium trifluorophenyl boronate ([PhBF 3 ]K, 2a) 25 , and the intramolecular hydroalkoxylation of unactivated alkenes such as 1,1-diphenyl-4-penten-1-ol (3a) 26,27 (Fig. 2b,c). Subsequently, we investigated the size and matrix effects of these reactions under previously reported conditions . The aerobic homocoupling of potassium trifluorophenyl boronate (2a) affords biphenyl (2b), while the intramolecular hydroalkoxylation of 1,1-diphenyl-4-penten-1-ol (3a) furnishes cyclization product 3b. The approximate activity of each catalyst was determined based on the yield of the corresponding products. The reaction was stopped after 4 h, and the yield of 2b and 3b was determined by GC (Table S2 and S3, Fig. S6 in Supplementary Information). In these reactions, similar trends as in the aerobic oxidation of 1a were observed, i.e., the highest catalytic activity was observed for Au:PVP (K-90) (core size: 7 nm). It should also be noted that the reaction time for the aerobic homocoupling, as well as for the intramolecular hydroalkoxylation was substantially shortened from 16-24 h (conventional conditions) to 4 h when Au:PVP (K-90) (core size: 7 nm) was used. As these results clearly demonstrate that the polymer matrix effect is not limited to specific substrates or reactions, the polymer matrix should play an important role for the catalyst activity. Moreover, these results suggest that it should be possible to tune the catalytic activity of the PVP-stabilized AuNCs via the polymer matrix. The electronic structure of Au:PVP. Among the series of Au:PVP derivatives tested, Au:PVP (K-90) (core size: 7 nm) exhibited extremely high catalytic activity in all three reactions. In order to better understand the origins of this outstanding catalytic performance, we wanted to determine the electronic structure of these AuNCs, and find out if an effect of the polymer matrix could be established. The analysis of the electronic structure was achieved by X-ray adsorption (EXAFS and XANES) and X-ray photoelectron spectroscopy (XPS) (Fig. S7-S8 in Supplementary Information). Table 1 and Table S4 (in Supplementary Information) show the EXAFS and XANES data, whereby the former are in good agreement with the TEM results. The Au-Au coordination number CN, as well as the interatomic bond lengths R increase with increasing core size. All Au:PVP clusters show a noticeable decrease of bond lengths compared to Au foil (2.88 ). The contraction of the metallic bond distance in these AuNCs should also lead to changes of their electronic properties 28 . However, the relatively small alteration of the bond lengths in the larger AuNCs relative to Au foil suggests that these should be less effective in catalytic reactions. Moreover, in most of the cases and virtually for all core sizes, Au:PVP (K-90) exhibited significantly smaller coordination numbers than PVP (K-30) and PVP (K-60) (cf. entries 1-4, 7-9, 13-15). It can therefore be concluded that the increasing molecular weight of the stabilizing polymer resulted in a morphological change of the metal surface. Although an analysis of the XANES-derived edge energy values for the Au:PVP clusters suggested that their Au cores exhibit a more anionic character than Au foil, it should be noted that the comparison of the observed small edge energy differences was difficult, mostly due to the low resolution and concentration of gold in these clusters system. Therefore, we decided to further examine the size-dependence of the electronic structures of the Au:PVP clusters by XPS. The electron density of the Au cores can be determined experimentally by XPS via the binding energy (BE) of Au4f 7/2 . The XPS data showed a primary Au4f 7/2 band for the Au atoms accompanied by the corresponding satellite peaks. The width of the observed peaks is thereby indicative of a contribution of Au atoms with different electron density. Particularly, the significantly smaller BE of Au4f 7/2 relative to bulk gold (84.0 eV) suggests that the negative charge on the gold surface in Au:PVP arises from the interaction between the gold surface and the polymer matrix (Table 1, Fig. S8 in Supplementary Information) 2 . The BE of Au:PVP (K-90) (entries 13-17) decreases with increasing core size (≤ 9 nm) before increasing with further increasing core size. Au:PVP (K-90) (core size: 9 nm) (entry 18) exhibits a significantly increased BE relative to the smaller clusters, due to the decreased electron density on the gold surface. The Au 4f 7/2 BE for Au:PVP (K-90) (core size: 7 nm) (entry 17) is clearly smaller than that for the other clusters, indicating the highest negative charge on its surface, which is in good agreement with the results of the catalytic activity study. Although the BE of Au4f 7/2 and its catalytic activity are not perfectly proportional (Fig. 5), it should nevertheless be useful to compare these correlations with similar catalyst systems, e.g., with Au:PVP (core size: 1.3 nm). The smallest BE value was observed for Au:PVP (K-30) (entry 1), followed by Au:PVP (K-60) (entry 7), and Au:PVP (K-90) (entry 14). These results are consistent with the experimental data that Au:PVP (K-30) with small cores exhibited the highest catalytic activity (Fig. 4a). Another possible explanation is the geometrical effect by changing the crystallinity. For example, the TiO 2 -supported poly-crystalline Au showed twice the activity of that single crystalline Au/TiO 2 system 29 . Tsukuda and co-workers also reported that icosahedral Au 144 clusters supported on hierarchically porous carbon exhibited superior catalytic activity to the smaller fcc Au clusters 30 . To confirm the structural differences in each clusters, we carried out the high-resolution TEM (HR-TEM) and powder X-ray diffraction (PXRD) measurements. HR-TEM measurement revealed that, in case of relatively larger cluster (more than 5 nm), most of the cores were polycrystals regardless to the polymer length (Fig. S2, Supplementary Information). In addition, PXRD measurement suggested that the larger clusters mainly keep fcc structures (Fig. S3, Supplementary Information) 31 . Although this morphological effect on the catalytic activity should not be ignored, it is difficult to explain the significant difference of the catalytic activity of 7 nm-sized Au:PVP (K-30) and Au:PVP (K-90). Therefore, AuNCs should thus be intimately correlated with mainly the electronic properties on the surface of the gold cluster, and it can hence be concluded that the excellent catalytic activity of Au:PVP (K-90) (core size: 7 nm) may be attributed to the highly negative charge density on its surface. The electronic structure studies also revealed that Au:PVP (K-90) (core size: 7 nm), which exhibited high reactivity, possessed more anionic nature. This behavior may be rationalized in terms of electron affinity on the gold surface, as the partial electron transfer from PVP to the core gold leads to bond activation of the substrate prior to participation in the catalytic cycle 25 . Structural analysis of the colloidal Au:PVP. Considering the aforementioned results in their entirety, it can be concluded that the excellent catalytic activity of Au:PVP (K-90) (core size: 7 nm) should be ascribed to the high negative charge on the cluster surface. This however poses the question: why does Au:PVP (K-90) (core size: 7 nm) possess such a highly electronegative surface? In order to answer this question, we examined the physical structure of the AuNCs and the effect of the stabilizing PVP on the AuNCs surface. We measured the size of the Au:PVP colloids by the induced grating method (IG method), which includes an activation procedure induced by dielectrophoresis to form a particle grating and measures the decay of the diffracted light. It thus provides a highly sensitive and reproducible means to measure single colloids down to the nanometer scale. The advantage of the IG method over dynamic light scattering (DLS) is that the former is immune to the inaccuracies arising from contamination with large particles, with which the results of the latter are inevitably associated 32,33 . Fig. 6 and Table S5 (in Supplementary Information) show the relationship between Au:PVP colloid size and core size. The colloid size of most Au:PVP (K-30) was 40 ± 10 nm, irrespective of the core size and the polymer chain length, while the colloid size of free PVP was ~100-200 nm. The colloid size of both Au:PVP (K-60) and Au:PVP (K-90) decrease with increasing core size (Fig. S9-S10 in Supplementary Information). Judging from the volume in the presence of short-chained polymers, Au:PVP colloids consist predominantly of water (>90%), indicating that the gold cluster should be drifting in diluted aqueous PVP solutions, reminiscent to an egg of frog model (Fig. 7a), which reminds us a frog egg surrounded by jelly-like blob in frogspawn. Recent results obtained from MD-DFT calculations indicated that PVP moieties bind much closer to metal surfaces than H 2 O. Consequently, the PVP concentration should be increased in close spatial proximity of the cluster, which might explain the stabilizing effect of PVP 31 . However, the physical structure of the interface between the gold surface and the PVP matrix must be loose and soft. In contrast, when the bigger clusters are wrapped with high-molecular-weight polymers, the polymer should wrap around the Au core and thus afford higher surface coverage 34 and more entangled structures (Fig. 7b). In such entangled structures, the hydrogen-bonding network should be eliminated, which would result in contracted colloids. Indeed, the aforementioned calculations suggested that virtually no water should be included in Au:PVP (K-90) (core size: 7 nm) colloids. Therefore, it can be concluded that a strong interfacial interaction is induced when the chain length of the stabilizing polymer and the core size of the metal cluster are matched, emerging the highly electronegative surface on the metal core. ## Conclusion In summary, we were able to demonstrate that Au:PVP catalysts promote a variety of reactions. In all cases, the properties of Au:PVP depend on the cluster size. However, morphological effects can surpass the size effect and control the catalytic activity. From an applications-driven perspective, it is highly promising that the catalytic activity of AuNCs can be modulated by the cluster size and by the polymer matrix. Our experimental results revealed that Au:PVP (K-90) (core size: 7 nm) exhibited a catalytic activity that was up to five times higher than that of Au:PVP (K-30) (core size: 1.3 nm). The extremely high activity of the former should be ascribed to the high surface coverage of the gold core with PVP (K-90), which should lead to a high density of negative charges on the core surface. The analysis of the physical structure suggested a difference in surface morphology. Even though the interface between the Au surface and the PVP matrix is usually loose, this is not the case for long PVP chains. AuNCs stabilized by long-chain polymers result in highly entangled structures via entanglement of the Au surface with PVP, which increases the electronegativity on the surface of the cluster. However, it is still extremely difficult to experimentally determine the structure of this interface precisely, and new experimental methods as well as the computational studies are required to advance research in this area.

chemsum

{"title": "The Impact of the Polymer Chain Length on the Catalytic Activity of Poly(N-vinyl-2-pyrrolidone)-supported Gold Nanoclusters", "journal": "Scientific Reports - Nature"}

investigating_the_influence_of_aromatic_moieties_on_the_formulation_of_hydrophobic_natural_products_

## Abstract: Many natural compounds with interesting biomedical properties share one physicochemical property, namely a low water solubility. Polymer micelles are, among others, a popular means to solubilize hydrophobic compounds. The specific molecular interactions between the polymers and the hydrophobic drugs are diverse and recently it has been discussed that macromolecular engineering can be used to optimize drug loaded micelles. Specifically, π-π stacking between small molecules and polymers has been discussed as an important interaction that can be employed to increase drug loading and formulation stability. Here, we test this hypothesis using four different polymer amphiphiles with varying aromatic content and various natural products that also contain different relative amounts of aromatic moieties. While in the case of paclitaxel, having the lowest relative content of aromatic moieties, the drug loading decreases with increasing relative aromatic amount in the polymer, the drug loading of curcumin, having a much higher relative aromatic content, is increased.Interestingly, the loading using schizandrin A, a dibenzo[a,c]cyclooctadiene lignan with intermediate relative aromatic content is not influenced significantly by the aromatic content of the polymers employed. The very high drug loading, long term stability, the ability to form stable highly loaded binary coformulations in different drug combinations, small sized formulations and amorphous structures in all cases, corroborate earlier reports that poly(2oxazoline) based micelles exhibit an extraordinarily high drug loading and are promising candidates for further biomedical applications. The presented results underline that the interaction between the polymers and the incorporated small molecules are complex and must be investigated in every specific case. ## Introduction: For many years, natural products have played an important part in drug discovery. In the late 20 th century, a majority of drugs were either natural compounds or their derivatives. At the end of their review concerning the importance of natural products for drug discovery Newman et al. argued that well-defined drug delivery systems could overcome unfavorable physicochemical properties, like aqueous solubility, in the future. Through high-throughput screening, new chemical entities or lead structures are being identified and evaluated every day, but only a minute fraction ever ripen into an approved drug. Obviously, a large proportion of drug candidates are poorly water-soluble which calls for effective formulation strategies. Traditionally used surfactants like Cremophor EL and Tween 80 have drawbacks as they can elicit potentially life-threatening side effects and are limited with respect to their in solubilizing ability. Polymeric micelles have been discussed and evaluated as carriers for hydrophobic molecules for many years and thousands of papers praising the potential of polymer based drug delivery systems are published every year. However, until now, only one micelle-based formulation (Genexol-PM®, South Korea) has been used in the clinic with several other being under clinical development. Zhang et al. argued that the low drug loading capacity and poor in vivo stability typically displayed by polymeric micelles is responsible for this major discrepancy. These major problems concerning nanoformulations, drug delivery and the advancement of polymeric micelles for clinical cancer therapy were also critically reviewed by other researchers. Polymer micelles comprising a poly(2-oxazoline) (POx) based amphiphilic triblock copolymer (poly(2-methyl-2-oxazoline)-block-poly(2-butyl-2-oxazoline)-block-poly(2-methyl-2oxazoline) (PMeOx-b-PBuOx-b-PMeOx ≡ A-pBuOx-A)) constitute an unusual exception. Loading capacities (LC) of almost 50 wt.% for one of the most commonly used chemotherapeutic agents, paclitaxel (PTX), were reported by Kabanov, Jordan, Luxenhofer and co-workers. . Despite the very low water solubility, PTX is a high affinity substrate for P-glycoprotein (P-gp) which leads to PTX resistance in cancer cells. The drug loaded polymer micelles formed stable and injectable formulations and showed significantly increased therapeutic efficacy. The combination of high drug loading and stability was only seen in block copolymers with poly(2-n-butyl-2-oxazoline) as hydrophobic core. Testing a variety of structurally different taxanes led to similar high drug loadings and stability of the formulations. However, with etoposide and bortezomib, two well-known topoisomerase and proteasome inhibitors, respectively, no stable formulations could be obtained. Lübtow et al. investigated a small library of structurally similar ABAtriblock copolymers based on poly(2-oxazoline)s and poly(2-oxazine)s and explored their solubilization capacity for PTX and curcumin (CUR ), another well-known natural compound featuring extremely low aqueous solubility, bioavailability and stability. The authors observed significant and orthogonal specificities dependent on one methylene group. That research outlines the complexity of drug/carrier interactions. More recently, this CUR nanoformulation was characterized in detail and compared in 2D and 3D cell culture with CUR dissolved in DMSO. Another natural product with low water solubility can be found in fruits of Schisandra chinensis, which are widely used in traditional Chinese and Japanese herbal medicine and are said to have hepatoprotective, anti-asthmatic, anti-diabetic and sedative properties. Dibenzo[a,c]cyclooctadiene lignan metabolites are thought to be responsible for the majority of these biological effects. Many of such lignans have been extracted and chromatographically isolated from Schisandra chinensis fruits. The compounds have been predominantly identified by UV-Vis, IR-and NMR spectroscopy, mass spectrometry and circular dichroism. [33, In addition, the absolute configurations of some such lignans were determined via crystal structure analysis. Schobert and co-workers established a simplified extraction method followed by one saponification step to obtain the pure dibenzo[a,c]cyclooctadiene lignan schisandrol A by column chromatography. This was converted to a cinnamate and a titanocene derivative which both showed promising P-gp inhibition and increased activity against cervix and breast cancer cells. The formation of nanoparticles and nanocrystals to formulate schisantherin A, a related dibenzo[a,c]cyclooctadiene lignan, was described by Cheng et al. The drug/carrier aggregates could pass the hemato-encephalic barrier and showed effects potentially useful for the treatment of Parkinson´s disease. Several strategies for the development of drug specific drug delivery platforms have been followed, lately. In particular, Luo, Nangia and co-workers backed the synthetic work with extensive modeling and achieved very high drug loadings paired with excellent therapeutic efficacy. Börner and co-workers employed a high-throughput screening to find an improved drug loading for different cargo. The driving forces considered relevant for drug incorporation are hydrophobic and electrostatic interactions, hydrogen bonding, π-π stacking and van der Waals forces. In the literature, the relevance of these interactions for drug formulation is widely discussed. In order to stabilize polymeric micelles and so increase their loading capacity for PTX (28 wt.%) and docetaxel (34 wt.%) Shi et al. synthesized amphiphilic block copolymers comprising the aromatic monomer N-2-benzoyloxypropyl methacrylamide (HPMAm-Bz) as a hydrophobic building block. The π-π stacking effect significantly increased the stability, loading capacity and therapeutic index of drug loaded polymeric micelles. Moreover, this led to a retardation of PTX release compared to polymers that did not contain aromatic moieties. Amphiphilic diblock copolymers containing poly(2-phenyl-2-oxazolin) (PPheOx) and PMeOx were tested on their self-assembly in aqueous milieu. Dependent on different block compositions, the researchers found polymeric micelles, vesicles and larger polymersomes. Furthermore, the hydrophobic drug indomethacin could be successfully formulated. Tiller and co-workers described ABAtriblock copolymers based on PPheOx and PMeOx and discussed their usage in drug delivery. The size and morphology of the aggregates depended strongly on the overall block length and the balance between hydrophilic and hydrophobic moieties. Here, we present a small library of POx based amphiphiles, in which the aromatic character was increased systematically and the solubilization capacity for drugs with different aromatic content was investigated. ## Materials and Methods: All substances and reagents for the polymerizations were obtained from Sigma-Aldrich (Steinheim, Germany) or Acros (Geel, Belgium) and were used as received unless stated otherwise. Curcumin powder from Curcuma longa (Turmeric) was obtained from Sigma-Aldrich (curcumin = 79%; demethoxycurcumin = 17%, bisdemethoxycurcumin = 4%; determined by HPLC analysis). Paclitaxel was purchased from LC Laboratories (Woburn, MA, USA). Deuterated solvents for NMR analysis were obtained from Deutero GmbH (Kastellaun, Germany). All substances used for polymerization, specifically methyl trifluoromethylsulfonate (MeOTf), MeOx, BuOx, PheOx, and BzOx were refluxed over CaH 2 and distilled and stored under argon. Benzonitrile (PhCN) was dried over phosphorus pentoxide. The monomers 2-n-butyl-2-oxazoline (BuOx) and 2-benzyl-2-oxazoline (BzOx) were synthesized following the well-known procedure by Seeliger et al. The Pt-NHC-complex (Pt-NHC) and the fluorinated curcuminoid derivative (CUR-F 6 ) were synthesized according to literature (Figure S6). Fruits of Schisandra chinensis were obtained from Naturwaren-Blum (Revensdorf, Germany) and were dried and powdered prior to the extraction procedure. Schizandrin (SchA) was obtained from powdered Schisandra chinensis using the simplified extraction procedure investigated by Schobert et al. Thin layer chromatography (TLC), NMR-, IR-, UV-Vis-spectroscopy and electrospray ionization mass spectrometry (ESI-MS) were used for analytical issues. TLC were perfomed on Sigma-Aldrich ® TLC Plates containing silica gel matrix (stationary phase) using n-hexane/ethylacetat (1:1) as mobile (phase). For NMR measurements a small fraction of the purified compound SchA was dissolved in deuterated dichloromethane and 1 H-, 13 C-, COSY-(correlation spectroscopy), HSQC-(heteronuclear single quantum coherence) and HMBC-(heteronuclear multiple bond correlation) experiments were recorded. For IR-analysis small amount of SchA were recorded on an FT-IR-spectrometer 4100 from 500 to 4000 cm -1 from Jasco (Gross-Umstadt, Germany). For the UV-Vis measurement a 1 g/L ethanolic solution of SchA was filtered through 0.2 µm PTFE filters (Rotilabo, Karlsruhe) and recorded at 25 °C from 700 to 180 nm. The purity of SchA was determined to be 97.9% by analytical high pressure liquid chromatography (HPLC). The polymers A-pBuOx-A, A-p(BuOx-co-BzOx)-A, A-pBzOx-A and A-pPheOx-A were synthesized by living cationic ring opening polymerization (LCROP) as described previously. The reactions were controlled by 1 H-NMR-spectroscopy. The lyophilized polymers were characterized by 1 H-NMR and gel permeation chromatography (GPC). The critical micellar concentrations of the ABA-triblock copolymers were determined by pyrene fluorescence measurements. The I 1 /I 3 -ratio in dependence of varying polymer concentrations and the total redshift of I 1 in dependence of varying polymer concentrations were detected and used for the determination of the CMC-values. Drug loaded polymeric micelles were prepared by thin film method (Figure 1 b). The loading capacities (LC) and efficiencies (LE) were determined by HPLC measurements, according to equation ( 1) and (2). where m drug and m excipient are the weight amounts of the solubilized drug and polymer excipient in solution and m drug,added is the weight amount of the drug initially added to the dispersion. Therefore, it was assumed that no loss of polymer during micelles preparation occurred. The aggregation behavior of the polymers (10 g/L in PBS) and polymer-drug solutions (1:0.5 g/L in PBS) were investigated by dynamic light scattering (DLS) measurements at 27 different angles (temperature was fixed at 25 °C). The glass transition temperatures (T g ) and melting points (mp) were determined by differential scanning calorimetry (DSC) measurements. All methods are described in more detail in the supporting information. ## Results and Discussion: Inspired by reports on benefits for drug delivery via π-π stacking between drug carrier and loaded API , we wanted to investigate this issue in poly(2-oxazoline) based polymer amphiphiles. In particular, the Hansen-solubility parameters calculated by Dargaville and coworkers suggested a benefit regarding drug loading using polymer amphiphiles comprising a hydrophobic poly(2-phenyl-2-oxazoline) block. In the case of paclitaxel and docetaxel, Hennink and co-workers reported that incorporation of aromatic side chains into thermosensitive block copolymers of modified hydroxypropyl methacrylamides improves drug loading. However, it is important to note that in this study PTX precipitation rather than release was quantified. The authors argue that this was done as it is difficult to upload proper sink conditions for the extremely poorly soluble PTX. In contrast, in a preliminary study, we did not observe any benefit with respect to PTX formulation when we included aromatic moieties into the hydrophobic block. The inclusion of an aromatic moiety (A-p(BuOx-co-BzOx)-A; LC PTX : 36 wt.%) led to significant loss of loading capacity in comparison to A-pBuOx-A (LC PTX: 49 wt.%). Therefore, the present study investigates the influence of different proportions of aromatic moieties within poly(2-oxazoline) based ABA triblock copolymers on the formulation of different hydrophobic drugs with varying aromatic content in more detail. To this end, we used a small library of four different polymers. As in previous work, the hydrophilic block A was poly(2-methyl-2-oxazoline) (pMeOx). The hydrophobic blocks were in order of poly(2-phenyl-2-oxazoline) A-pPheOx-A, respectively. The polymers were prepared by living cationic ring opening polymerization (LCROP) and characterized by 1 H-NMR and GPC (Table 1 and supporting information). All polymers exhibited CMC values in the low μM range (determined by pyrene fluorescence, Figure S3), often deemed favorable for intravenous administration (Figure S3, Table S1). Interestingly, in the present library, it appears that the introduction of aromatic side chains has no marked influence on the CMC values. For formulation, we focused on three natural compounds. On the one hand, we employed the well-known and extremely water-insoluble compounds paclitaxel PTX (0.4 mg/L) and curcumin CUR (0.6 mg/L) . On the other hand, we tested the poorly soluble dibenzo[a,c]cyclooctadiene lignan schizandrin (SchA, solubility: 0.19 g/L determined via HPLC). These three natural compounds differ in their relative aromatic content. While PTX contains 3 phenyl rings at a molar mass of 854 g/mol, SchA contains two rings at 432 g/mol and CUR contains also 2 phenyl rings, however connected with a bridging π-system at a molar mass of 368 g/mol. The ratio of C arom. , the carbon atoms, which are included in the aromatic system, and C total , the total number of carbon atoms in the hydrophobic compound, represents the aromaticity of the three different cargos (Table 2). Additionally, the insoluble compounds Pt-NHC and CUR-F 6 (for structures, please see Figure S6) were investigated, the latter of which can be viewed as derivative of natural compound CUR. All four polymers were tested for the solubilization of PTX and CUR, while A-pPheOx-A was not used to formulate SchA. As previously reported, A-pBuOx-A is an excellent solubilizer for PTX but much less so for CUR. This was corroborated also in the present study (Figure 2). Interestingly, SchA, having an intermediate relative aromatic content was solubilized very well, but less than PTX. As previously reported and independently reproduced here using a newly synthesized polymer, the introduction of benzylic moieties (A-p(BuOx-co-BzOx)-A) does not help in the formulation of PTX but rather reduces the maximum drug loading. Interestingly, while in the case of SchA the maximum drug loading increased slightly, the LC CUR that could be achieved increased significantly (Figure 2). In the case of A-pBzOx-A, the LC PTX In this report, the authors found an increased PTX loading with increasing content of aromatic co-monomer, which was attributed to π-π stacking between polymer and drug molecules. While apparently valid in some cases, it appears that this rationale is not generally helpful to increase drug loading in polymer micelles. Due to insufficient amounts of SchA, this compound was not tested using A-pPheOx-A. It should be noted that we also attempted formulation of CUR-F 6 , but no stable formulation could be obtained by film hydration method with either polymer. The NHC-Pt-complex was solubilized with A-pPheOx-A by film hydration method using dichloromethane. However, it appears the complex does not exhibit sufficient stability as HPLC analysis revealed multiple signals after formulation whereas the compound was pure initially. Therefore, quantification was not possible. from Ref . In all cases, the polymer concentration was fixed at 10 g/L. Data is given as means ± SD (n = 3). The drug content was quantified immediately after preparation. In order to gain a basic understanding of the stability of the drug formulations, we stored the formulation under ambient conditions and took samples after 10 and 30 days. For A-pBuOx-A, we have previously observed excellent stability of the PTX formulation without any precipitation after several months. In general, all tested formulations of PTX, CUR and SchA showed very good stability over 30 days (Figure 3). No significant loss of SchA and very little variability of the drug concentration was observed. In the case of PTX, the variability was somewhat higher, nevertheless overall formulation stability was excellent. The stability compares favorably with π-π interaction stabilized PTX formulations reported by Hennink et al. In that report, 50% to 100% of the solubilized PTX precipitated within 10 days, depending on the aromatic content of the micelles. In the case of CUR, we made unexpected observations. First of all, the stability of CUR formulated in the POx based micelles was remarkably high, especially considering the well-established low chemical stability of CUR in aqueous media. In case of A-p(BuOx-co-BzOx)-A and A-pPheOx-A, we did observe some loss of CUR concentration and increased variability after 30 days, but the average drug loading remained high. Notably, according to HPLC analysis, CUR did not show any signs of degradation, even though it is often reported that CUR is not stable in aqueous environment. Interestingly, in the case of A-pBzOx-A, the CUR concentration in solution (after centrifugation and filtration) was much higher on day 10 (LC CUR : 45.6±3.3 wt.%) and day 30 (LC CUR : 44.6±4.6 wt.%) than on day 0 (LC CUR : 20.8±2.1 wt.%). It should be noted, for our stability studies, the drug formulations were stored under ambient conditions over the pellet of unformulated drug (if any). Thus, it appears that CUR that initially precipitated during thin film hydration became incorporated over time into the micelles. Unfortunately, we cannot explain this phenomenon at this point, but further studies are certainly warranted. In addition to the described single-drug formulations, we also investigated binary coformulations. Previously, we have reported several POx-based binary and ternary drugformulations with very high loading of ≥58 wt.%. Here, we investigated A-pBzOx-A for co-formulation of PTX with SchA and CUR, respectively. Both combinations could show interesting pharmaceutical synergies. For these preliminary studies, we fixed the relative drug weight ratio at 1/1 for both combinations (10 g/L A-pBzOx-A and 6 g/L in the case of PTX/SchA and 8 g/L in the case of PTX/CUR). In both combinations, the loading efficiency was excellent and total LC exceeded 50 wt.%. The individual LC for PTX (40.8 wt.%) and CUR (40.1 wt.%) yielded an overall LC of 58.8 wt.%, while with the combination of PTX (35.6 wt.%) and SchA (36.9 wt.%) an overall LCs of 53.2 wt.% was obtained (Figure 4). Previously, we have investigated PTX formulation with A-pBuOx-A in great detail using electron microscopy, dynamic light scattering (DLS) and small angle neutron scattering. For preliminary elucidation of the aggregation behavior, we investigated aqueous polymer solutions and formulations by DLS (Figure 5). All polymers form aggregates in the size range expected for polymer micelles. The polymers containing aromatic moieties form very small and rather defined polymer micelles. At 10 g/L in PBS hydrodynamic radii between 10-20 nm were found (Figure 5 a, Figure S5 c). In the case of A-pBuOx-A, we found a rather broad distribution centered around a hydrodynamic radius of R h = 25 nm. Previously, using a different batch of the same polymer we observed a bimodal size distribution, originating from spherical and worm-like micelles. Possibly, the broad distribution observed in the present case could be caused by an unresolved bimodal distribution. The drug formulations studied here are not only potentially interesting as drug loaded micelles but also in form of solid dispersions. In this context, it is particularly interesting whether the drug is present in amorphous or crystalline form. Neither polymer exhibited a melting point, therefore being fully amorphous structures with glass transitions temperatures predictable by the Fox equation using the T g -values of the homopolymers (Figure 6 a, Table S9, Equation 6, Table S10). SchA did not undergo thermal degradation at temperatures up to 200 °C, but exhibited a melting point of 129 °C (Figure 6 b), which corroborates values found in the literature. Upon cooling, we did not observe recrystallization at the chosen experimental parameters, but also a T g -value was detectable. The second heating cycle revealed a T g of 30.7 °C for SchA. For comparison, we also analyzed a simple physical mixture of SchA and A-pBzOx-A with the nanoformulation (both 1/2, w/w) obtained via thin-film hydration and subsequent lyophilization. For the nanoformulation of A-pBzOx-A and SchA no melting point could be detected in the first heating cycle (Figure 6 c). Only a T g of 67.5 °C could be discerned, showing that the nanoformulation with a LC of about 30 % was fully amorphous. ## Moreover, we investigated drug formulations of A-pBzOx The fact that only one T g is observed, elucidates that no independent domains of amorphous drug and amorphous polymer are present, but that the two entities are intimately and molecularly intertwined. In case of the physical mixture, we could clearly observe the melting point of SchA in the first heating cycle (Figure 6 d). Also in this case, no recrystallization of SchA was observed upon cooling. The second heating cycle only revealed a T g at 70.4 °C. Interestingly, the Fox-equation predicts a much lower T g for a mixture of 33 wt.% SchA and Fox-equation yields an expected T g of only 52 °C. The origin of this discrepancy is unclear at this point. Melting points (endothermal maximum) were determined in the first heating cycle, respectively. The particular T g values were determined in the second heating cycle. The scans were performed under N 2 -atmosphere using constant cooling and heating rates of 10 K/min. ## Conclusion: We investigated the influence of incorporation of aromatic moieties into poly(2-oxazoline) based ABA triblock copolymers. In addition to the varying degree of aromaticity within the polymers, three cargo compounds with different relative aromatic content were tested. In contrast to previous reports on a different polymeric system, incorporation of 2-benzyl-2oxazoline or 2-phenyl-2-oxazoline did not increase drug loading or formulation stability in the case of paclitaxel. Interestingly, the formulation of the natural compound schizandrin A was barely affected while the loading with curcumin benefitted significantly from incorporation of 2-benzyl-2-oxazoline but less so for 2-phenyl-2-oxazoline. Therefore, it appears that π-π stacking can be beneficial for drug loading and formulation stability in some cases, but this should not be considered a general phenomenon and must be assessed on a case-by-case basis. ## Syntheses and extraction procedure Monomer synthesis: The building blocks for the hydrophobic core of ABA-triblock copolymers General synthetic procedure 1, GSP 1 For the reaction 1 eq of respective nitrile, 1.2 eq of alkanolamine and catalytic amounts of zinc acetate dehydrate (0.025 eq) were added to a nitrogen flushed flask and heated to 130 °C under reflux for 72 h until the reaction mixture turned dark brown. Reaction progress was controlled by FTIR-and 1 H-NMR-spectroscopy. The mixture was dissolved in dichloromethane and washed with H 2 O (3x). The organic phase was dried with MgSO 4 , filtered and concentrated under reduced pressure. The raw product was mixed with CaH 2 and purified via vacuum distillation. If necessary, distillation was repeated and the product was kept under nitrogen atmosphere. ## 2-n-butyl-2-oxazoline (according to GSP1)Ref[27] 2-benzyl-2-oxazoline (according to GSP1) Benzyl cyanide: 204 g (1. ## ABA-triblock copolymer synthesis according living cationic ring opening polymerization General synthetic procedure 2, GSP 2 The used technique were carried out as described previously. Briefly, 1 eq of initiator ## Extraction of schisandra chinensis fruits The extraction procedure of fruits of schisandra chinensis was carried out as described by Schobert et al. The dried fruit powder (200 mg) was first extracted at RT with n-hexane (3.5 L) for 10 h. This extract was filtered off and the remainder was dried and re-extracted with ethyl acetate (2.5 L) for 10 h. After concentration under reduced pressure the raw product was purified three times by column chromatography on Silica Gel (60 230-400 mesh particles) from Sigma-Aldrich (Steinheim, Germany) with nhexane/ ethyl acetate (1:1) as the eluent to maintain Schizandrin (SchA) as a colorless powder. ] of the Polymers A-pBuOx-A (purple), A-p(BuOx-co-BzOx)-A (dark red), A-pBzOx-A (black) and A-pPheOx-A (dark blue) (10 g/L in PBS) determined using monoexponential fit function. d) Hydrodynamic radii R h [nm] in dependence of the particular scattering vector q 2 [cm -2 ] of the nanoformulation A-pBzOx-A/SchA (1/0.5 g/L in PBS) determined using triexponential fit function (circles: Species 1; triangles: Species 2; stars: Species 3; squares: weighted average). e) Hydrodynamic radii R h [nm] in dependence of the particular scattering vector q 2 [cm -2 ] of the nanoformulations A-pBzOx-A/PTX (blue) and A-pBzOx-A/CUR (red) (1/0.5 g/L in PBS) determined using monoexponential fit function. ## Differential Scanning Calorimetry The DSC measurements were performed in aluminum crucibles on a calibrated DSC 204 F1 Phoenix system from NETZSCH (Selb, Germany) equipped with a CC200 F1 controller unit. The heating and cooling rate was constantly 10 K/min using a constant N 2 -atmosphere. For each sample three heating cycles and two cooling cycles were performed. Glass transitions temperatures of ABA-triblock copolymers (binary system) (𝑇 ) ) are predictable using the well-known Fox equation and the T g -values of the homopolymers (𝑇 ),I , 𝑇 ),S ) and the mass fractions (𝑚 I , 𝑚 S ), respectively (Equation 7; Table S10). Table S10. T g -values of the homopolymers used for the prediction of the T g -region of ABA-triblock copolymers. References are noted. ## Homopolymer Building blocks T g [°C] Reference pMeOx 60 75

chemsum

{"title": "Investigating the influence of aromatic moieties on the formulation of hydrophobic natural products and drugs in poly(2-oxazoline) based amphiphiles", "journal": "ChemRxiv"}

chemically_fueled_self-sorted_hydrogels

## Abstract: Narcissistic self-sorting in supramolecular assemblies can help to construct materials with more complex hierarchies. Whereas controlled changes in pH or temperature have been used to this extent for two-component self-sorted gels, here we show that a chemically fueled approach can provide three-component materials with high precision. The latter materials have interesting mechanical properties, such as enhanced or suppressed stiffness, and intricate multistep gelation kinetics. In addition, we show that we can achieve supramolecular templating, where pre-existing supramolecular fibers first act as a templates for growth of a second gelator, after which they can selectively be removed. ## Introduction Narcissistically self-sorted hydrogels have been made by methods like rapid mixing of components, 1 solvent switching, 2 chiral recognition, 3 electrostatic interactions, 4 pH change, 5,6 and supramolecular catalysis. 5,6 Thermal annealing is most commonly used, where different gelation temperatures of the molecules allow them to self-assemble sequentially. A second approach, uses a gradual change in pH to create two-component self-sorted hydrogels. Gluconolactone (GdL) hydrolysis leads to sequential gelation of co-dissolved monomers at their respective pKa values, leading to materials with improved mechanical, optoelectronic, and photoconductive properties. van Esch and co-workers have shown fabrication of self-sorted hydrogels using 'kinetic selfsorting' of both a charged and neutral hydrogelator which form in situ via hydrazone formation. The reaction kinetics of the hydrogelators were found to be comparable, but they could still self-sort in certain cases due to their differing minimum gelation concentrations. 20 Here we show that a chemically fueled functional group transformation-that is, aldehyde-tohydroxysulfonate (and back)-can lead to exquisite control over self-sorting, providing access to well-structured three-component hydrogels. This approach is useful, since chemically very similar gelators with different innate reactivity can be used, which would otherwise coassemble when using controlled cooling. Moreover, since the functional group transformation is reversible, we can achieve supramolecular templating, where first a self-assembled fiber guides the growth of a second, after which the first can be selectively removed. Overall, we believe chemically fueled approaches are promising to get more exquisite control over supramolecular structures and the mechanical properties of multi-component gels. ## Results and discussion We recently reported a chemically fueled reaction cycle capable of gel-sol-gel transitions using aldehyde-containing saccharide derivative (compound 3 in Figure 1a). 21 In the latter, a thermally annealed hydrogel of 3 was first disassembled using sodium dithionite DT by converting the aldehyde moiety into its water-soluble hydroxysulfonate analog 3'. Formaldehyde (HCHO), produced in situ with a time-delay, then converted sulfonate 3' back to aldehyde 3, again leading to gelation. In the current work, we synthesized compounds 1 and 2 with close structural similarity to 3 and studied their assembly behavior in heat-cool cycles and their response to chemical fuels, both for the pure compounds and that of their mixtures (Figure 1). Surprisingly, this led to up to three-component well-structured self-sorted hydrogels. ) addition from 1 H-NMR. (l) Self-assembly kinetics from turbidity measurements at 500 nm using UV-Vis spectroscopy (see Figure S9 for detailed concentration dependent measurements). Assemblies of 1, 2, and 3 (21.6 mM) from 1', 2', and 3' upon addition of HCHO (47 eq.). Lines to guide the eye. Lag phases are indicated with asterisks. ## Thermally annealed 'thermogels' We started from the traditional 'controlled cooling' approach to try and obtain self-sorted hydrogels. Both pure 1 as well as 2 formed free standing hydrogels by thermal annealing with critical gelation concentration (CGC) of 23.5 mM and 21.6 mM, respectively. As a reminder, we had previously shown that thermogel 3 had a CGC of 25.8 mM. Such 'thermogels' have gelation temperatures (Tgel at 35 mM) of 74°C for 1, 78°C for 2, and 60°C for 3, as described in SI section S3.1. Confocal laser scanning microscopy (CLSM, including transmitted light imaging) showed that thermogel 1 (35 mM) formed a bimodal distribution of green fluorescent fibers: small fibers of width < 1 µm and length ~ 200 µm, and long fibers of width ~10 µm and length >500 µm. The latter were also much more emissive as compared to the short fibers. Thermogels of 2 showed large ~ 400 µm blue fluorescent spherulites while 3 formed thin (1-2 µm) and long (>1000 µm) non-fluorescent fibers (SI Figure S1). Next, we tested thermally annealed two-and three-components combinations of these molecules (total concentration is always constant at 35 mM). Thermogel 1+2 (ratio 0.5 : 0.5) formed a free-standing hydrogel composed of co-assembled spherules (Figure 1c) with an intermediate emission wavelength λem= 523 nm, as compared to the pure assemblies that were 548 and 485 nm, respectively. The latter suggests that 1 and 2 co-assemble, 22 when thermally cycled. In addition, the fibrous structure of 1 was completely suppressed in the 1+2 gel, further supporting co-assembly. Thermogel 1+3 only formed spherical co-assembled aggregates (Figure 1d), whereas pure 1 and 3 both form long fibers. The fluorescence emission wavelength is identical to that of 1, since 3 is non-emissive. In addition, gelation was suppressed, whereas pure 1 and 3 both form gels at 35 mM at room temperature. This shows that co-assembly can be detrimental for heatcool thermogels. The combination of 2+3 formed self-sorted hydrogels (Figure 1e), where the characteristic features of pure 2 and 3-being blue spherulites and non-emissive fibers, respectively (cf. Figure S1)-can be recognized. The latter makes sense, as 3 has a Tgel that is 18 degrees lower than that of 2. Therefore, during cooling from 85°C to room temperature, first 2 has time to form, followed by 3 later on. The same argumentation, however, does not hold for 1+3, which have gelation temperatures that are 14 degrees apart, but still co-assemble. Clearly, predicting narcissistic self-sorting in structurally similar molecules at equilibrium is not straightforward. Three-component mixtures of 1+2+3 (ratio: 0.2/0.2/0.6 and 0.33/0.33/0.33) showed features of co-assembled spherical 1+2, and some fibrous 3 albeit much shorter than in pure 3 (Figure 1e). Overall, multicomponent thermogels were mostly unable to self-sort except for the combination 2+3. Instead, co-assembly was preferred, leading to loss of their fibrillar morphology and in select cases their gel-forming ability. ## Chemically fueled 'chemigels' We now move to chemically fueled gels or 'chemigels' as we will refer to them. Typically, ~6 equivalents of DT were added to a previously formed thermogel, leading to chemical conversion of the aldehyde moiety to a hydroxysulfonate (i.e., 1', 2', or 3'), which resulted in complete dissolution. After 21 hours to ensure full disassembly and dissolution, HCHO was added to revert the hydroxysulfonate back to the aldehyde inducing re-assembly. Looking first at pure chemigels, we see that 1 still forms green emissive fibers as compared to the thermogel. However, they are not bi-modal in size distribution, but instead more uniform and straight. Compound 2 is also still forming spherulites, but they are 20 times smaller (at 20-30 µm) as compared to the thermally annealed case (cf. Figure S4). The latter indicates that there are more frequent nucleation events when chemically fueled. And lastly, compound 3 forms non-emissive long fibers both in the thermogels and chemigels (see Figure S13 for fluorescence emission data). Thermally, the fibers are randomly distributed in space (homogeneous nucleation), whereas chemically they grow more from defined nucleation centers into fractal-like structures, due to secondary nucleation as we showed previously. 21 For multicomponent systems, 1', 2', and 3' were mixed when fully disassembled, followed by addition of HCHO to form the multicomponent chemigels. That is, no heating or cooling procedures were involved to make chemigels. Strikingly, all multicomponent chemigels give rise to self-sorted assemblies (see Figure 1g-j), whereas this was only the case for 2+3 thermally. Even the 1+2+3 chemigel is self-sorted into the three characteristic green/blue/black (nonemissive) colors. Upon closer inspection, there is another interesting change in the assembly process of 2. Instead of self-nucleating and forming blue spherulites (cf. Figure S4), it grows on top of green fibers of compound 1, if present. That is, heterogenous nucleation of 2, using assemblies of 1 as nucleation sites, is more favorable than homogeneous nucleation. The result is that green fibers are formed, which have blue protrusions from its sides (Figure 1g, SI Figure S5, SI Movie 1 and 2). Compound 2, however, does not perform a heterogeneous nucleation on top of 3 (see Figure S5), likely because their chemical structures are too different from each other, favoring full narcissistic self-sorting. Overall, excellent self-sorting behavior is achieved using our chemically fueled (HCHO) approach. To understand why this is the case, we have examined the chemical and self-assembly kinetics of each building block, which is described next. ## Chemical reactivity The rate at which the individual hydroxysulfonates revert back to their respective aldehydes was determined by time-dependent 1 H NMR kinetics (Figure 1k). The rates of hydroxysulfonate consumption were found to be 1' > 2' > 3', which may be explained by looking at their chemical structures. Namely, compounds 1 and 2 have electron donating groups: 1 has ortho-hydroxyl and meta-methyl substituents, and 2 has two methoxy groups at the ortho and meta positions. These substituents cause 1' and 2' to react faster with HCHO to form their corresponding aldehydes as compared to 3'à3. 23 Moreover, the -OH group next to the -CHO can further stabilize 1 as a product through intra-molecular hydrogen bonding and thus further accelerate its hydroxysulfonate to aldehyde conversion 31 (as confirmed by NMR, see SI Figure S12). Thus, the overall rates of reaction 1 > 2 > 3 are reasonable considering their aromatic substitution patterns. ## Cooperative self-assembly for all derivatives Once 1'-3' has been chemically converted to its aldehyde form 1-3, it is charge neutral and can start assembling. The assembly kinetics were followed by UV-Vis turbidity measurements, where the optical density (O.D.) at 500 nm was tracked after addition of a large excess (~47 equivalents) of HCHO to hydroxysulfonate solutions (Figure 1h). UV-Vis turbidity measurements show that a cooperative mechanism of self-assembly for all three systems is present. Specifically, a lag time is observed for 1-3 (Figure 1l), and adding pre-formed self-assembled seeds leads to immediate growth without a lag phase (Figure S9). The kinetic time traces, however, show evidence of biphasic behavior and are not accurate enough to be analyzed by available nucleation / elongation / fragmentation models. 27 Corresponding to the rate of hydroxysulfonate consumption by NMR studies, the rate of selfassembly obtained by UV-Vis measurements and kinetic fitting gave the order of assembly as 1 > 2 > 3 (Figure 1l, Figure S9, and section 3.6 of SI). Overall, 1 aggregates faster, followed by 2, and 3 has the slowest assembly kinetics. This is evidenced by a rate constant for primary nucleation (kn) 10 orders of magnitude smaller for 3 when compared to 1 and 2 (which have values in the same order of magnitude). After nucleation, aggregates of 1 grow faster than 2 due to a ~2.3 times higher elongation rate constant (kp) and the presence of secondary nucleation that is not observed in the aggregation of 2 and is only important after nucleation. Confocal images and videos further confirmed this order of assembly in multicomponent mixtures forming self-sorted structures (SI Figure S7, SI Movie 1, 2, 3). ## Selective supramolecular template removal As shown above in Figure 1g, compound 2 can grow on top of assemblies of compound 1 due to heterogeneous nucleation, leading to 1+2 structures. Interestingly, we found that upon addition of DT to 1+2 structures, we could selectively remove 1 (see disappearance of green 1 fibers in Figure 2a; see also SI Figure S8, SI Movie 4). Considering their reactivity-where the rate of 1'à1 was faster than 2'à2 (Figure 1g)-we had expected that 2à2' would be faster than 1à1'. However, the reverse is observed, and 1 that forms first upon addition of HCHO (see Figure S8), also disappears first when adding DT. It is not entirely fair to make such simple assumptions based on chemical reactivity considering electron donating groups. In fact, when measuring the conversion rates of 1'à1 and 2'à2 we are starting from completely homogenous and monomeric hydroxysulfonates that react with HCHO. When viewing the conversion of 1à1' and 2à2', we start in the assembled state with micrometer-sized structures. It takes time for the DT to penetrate and react with structures of these sizes. However, DT can react more quickly with species that are in their monomeric state. NMR studies showed a higher proportion of soluble molecules for 1 than 2 (Figure 2b), due to their solubilizing hydroxyl groups. We therefore think that DT reacts preferentially with soluble 1 molecules, and therefore induces the selective disassembly of 1 fibers, as we have observed experimentally. That is, a depletion of 1 monomers below the critical aggregation concentration, causes 1 molecules to be extracted from 1 fibers. In effect, 1 fibers act a removable supramolecular template for the growth of 2 structures. We could confirm the latter hypothesis using NMR by treating chemically fueled assemblies of 1 with different DT concentrations. DT when below the net concentration of HCHO+1 (soluble monomers) did not lead to 1', and was preferentially consumed by excess HCHO. At concentrations comparable to HCHO+1 (soluble monomers) we could observe quick conversion of soluble 1 monomers to 1'. Once the DT was consumed, we observed monomers of 1 reappearing in the solution along with 1' due to dissolution of the aggregates. At much higher concentration of DT all the molecules of 1 (soluble+aggregates) were quickly converted to 1' (SI section 3.7, SI Figure S12). The mechanical properties of single and multicomponent self-sorted hydrogels were evaluated by rheology (Figure 3, see triplicate runs in Figure S10). To this end, solutions of 1', 2', or 3' or mixtures of the latter three always at a total concentration of 35 mM, were quickly mixed with an excess of HCHO and placed between the parallel plates of the rheometer (see Section 3.8 of the SI). Compound 1 formed unstable hydrogels that expelled solvent under slight perturbation (ca. 500 Pa) probably because its rigid crystalline fibers could not percolate solvent properly. Hydrogels of 2 evolved quickly to reach a high G' (ca. 2000 Pa) but eventually stabilized to lower values (ca. 700 Pa). The latter can be seen in Figure 3b, where a maximum in G' and G" was reached around 17 minutes, after which both decrease and reach a plateau. This behavior can be due to quick formation of numerous small assemblies (see SI Figure S4), but in the longer run, absence of long fibers would results in partial sedimentation of the assemblies to give the final G' values. Hydrogels of 3 had the best mechanical response (see black bar in Figure 3a and black lines in panel b), forming the stiffest of the three materials due to their long wavy fibers that are typically seen in supramolecular hydrogels. Interestingly, hydrogels consisting of 1+2 structures-formed by secondary nucleation of 2 on 1-had significantly higher mechanical strength (ca. 2300 Pa) than either 1 or 2 alone (Figure 3a,c). In contrast, solvent expulsion (for 1 alone) or settling of aggregates (for 2 alone) was not observed. ## Mechanical properties of multi-component gels Instead, the blue branches of 2 on fibers of 1, seem to give rise to better entanglement and thus the formation of a more stable hydrogel. Another interesting feature, was the step-wise evolution of G' for 1+3 self-sorted gels (red line in Figure 3c). From microscopy we know that 1 forms first, followed by 3 that is the slowest to nucleate (see Figure S7d and SI Movie 3). Interestingly, the total G' of 1+3 is ~50 % higher than that of 3 alone, at just half the concentration of 3. The high mechanical strength can be attributed to the presence of long extended fiber networks from both the individual components where the second network fills in the empty spaces to create a more densely packed hydrogel. In contrast, the 2+3 combination formed hydrogels with a lower mechanical strength (1200 Pa) than pure 3, but slightly above that of pure 2 (Figure 3a,c). The three component system (1/2/3 in ratio 0.33/0.33/0.33) was comparable to 1+3 gels. The ability of these gels to self-heal after applying a high shear rate (1000 s -1 ) for 30 seconds was also evaluated. Hydrogels of 1, once sheared, could not recover (G'≈ G" ≈ 10 Pa) and separation of solvent from fibers was observed. Further, gels of 2 could only partially recover to about 10% of their initial G'. The long fibers of 3 somewhat resisted total disruption of the gel properties, but the self-healing only recovered ~ 4% of the initial G'. Similarly poor recovery after shear damage was observed for 1+2 and 1+2+3, whereas 1+3 did not show any recovery. In sharp contrast, 2+3 could recover and form gels that were stronger even than the initial self-sorted 2+3 gels. Apparently, the 2+3 gel shows a synergistic interaction between fibers of 3 and spherulites of 2. The latter synergy presents intriguing prospects for other multicomponent self-sorted gels and materials, which can have materials properties-such as self-healing-that are not observed in the respective single component materials. ## Conclusions and outlook We showed how chemical fuels can be used to construct multicomponent self-sorted hydrogels. Subtle differences in the chemical structure of the hydrogelators affected both their reactivity toward the chemical fuels, as well as their propensity to self-assemble. The result is that intricate self-sorted materials could be made of molecules that using traditional approaches (e.g., heat/cool) would form poorly ordered co-assemblies. Our approach even allows for supramolecular templates to be used. That is, a first assembly guides the second, after which the first can selectively remove. Man-made chemically fueled systems have already shown fascinating properties such as oscillations, 28 dynamic vesicles, 29 and transient assemblies 21,30,31 , but have not been applied to control the hierarchy of multicomponent systems. Although ATP-powered transiently selfsorted colloids have been shown using DNA building blocks, 32 a similar approach in chemically fueled synthetic materials was lacking. We believe chemically fueled self-sorting provides a new method to achieve complex functional materials consisting of programmed orthogonal networks.

chemsum

{"title": "Chemically Fueled Self-sorted Hydrogels", "journal": "ChemRxiv"}

drug_repurposing_for_candidate_sars-cov-2_main_protease_inhibitors_by_a_novel_in_silico_method

## Abstract: The SARS-CoV-2 outbreak caused an unprecedented global public health threat, having a high transmission rate with currently no drugs or vaccines approved. An alternative powerful additional approach to counteract COVID-19 is in silico drug repurposing. The SARS-CoV-2 main protease is essential for viral replication and an attractive drug target. In this study, we used the virtual screening protocol with both long-range and short-range interactions to select candidate SARS-CoV-2 main protease inhibitors. First, the Informational spectrum method applied for Small Molecules was used for searching the Drugbank database and further followed by molecular docking. After in silico screening of drug space, we identified 57 drugs as potential SARS-CoV-2 main protease inhibitors that we propose for further experimental testing. ## Introduction An outbreak of the novel coronavirus disease in December of 2019 in Wuhan, China, has spread promptly to more than 213 countries with over 1 918 138 confirmed cases and over 123 126 confirmed deaths worldwide as of 15 April 2020 . The outbreak has been declared a global pandemic by The World Health Organization (WHO) on March 11, 2020 . It is uncertain whether a COVID-19 pandemic will cause multiple concurrent epidemics over 1-3 years, and SARS-CoV-2 may become an endemic virus globally . Moreover, millions of people have been disturbed as a result of mandatory isolations/quarantines and every part of society is severely affected with health care systems and economy adversely affected . The outbreak caused by SARS-CoV-2 is an unprecedented global public health threat owing to the high transmission rate of the virus, coupled with currently no drugs or vaccines approved. In the pandemic setting with rapid virus transmission, new vaccine production is of exceptional importance besides much needed therapeutic that are both expected to need months to years to develop. The rapid response action to the emergent pandemic is repurposing of approved antiviral, antimalarial, antiparasitic agents and those based on immunotherapy approaches to treat COVID-19, with some clinical trials already started . An alternative efficient additional strategy to tackle COVID-19 is in silico drug repurposing approaches. The main protease M pro , also called 3CLpro, represents an attractive drug target due to its essential role in the viral life cycle, crucial for viral replication. The pp1a and pp1ab, two overlapping polyproteins, important for viral replication and transcription, are encoded by the SARS-CoV-2 replicase gene . The M pro cleaves large polyprotein 1ab in at least 11 sites. The M pro is highly conserved across the Coronaviridae family and any mutation here can be disastrous for the virus . As one of the best-characterised drug targets among coronaviruses, in the absence of closely related human homologues, the M pro represents one of the most attractive SARS-CoV-2 drug target. Since there is no human protease with similar cleavage specificity, the inhibitors are expected to be nontoxic . SARS-CoV-2 M pro is active in a dimer form, consisting of two monomers arranged nearly perpendicular to one another . The dimerization is necessary for the M pro enzymatic activity as the N-finger of each of the two monomers interacts with Glu166 of the other monomer support the correct orientation of the S1 pocket of the substrate binding site. M pro active site comprises a catalytic dyad that consists of the conserved residues H41 and C145 . The available high-resolution experimental structure of the main protease of SARS-CoV-2 was used in the current study as the target for molecular docking based virtual screening (VS) . In this study, we used VS protocol with sequential filters, based on the both long-range and short-range interactions, to select candidate SARS-CoV-2 M pro inhibitors. First, the Informational spectrum method applied for Small Molecules (ISM-SM) was used for searching Drugbank database , and further was followed by molecular docking. By applying a new combo filter, we select 57 compounds for further experimental testing. ## Informational spectrum method analysis In the present study, we have used Informational spectrum method (ISM) for the structure/function analysis of the highly conserved SARS-CoV-2 protein M pro . According to the previous studies, the informational characteristic of the protein, identified in the analysis, corresponds to the protein key biological function. The informational spectrum (IS) of M pro contains three characteristic peaks at the frequencies F(0.1923), F(0.3183), and F(0.4414) shown in Figure 1. To find the domains of a protein crucial for the information related to the three frequencies, M pro was computationally scanned. As a result of scanning with the ISM algorithm, with overlapping windows of different lengths, we identified regions with the highest amplitudes at these frequencies. It was shown that the regions, including residues 131-195, 151-183 and 72-136, are essential for the information represented by the frequency F(0.1923), F(0.3183) and F(0.4414) respectively. Two dominant frequencies of M pro , F(0.1923) and F(0.3183) correspond to the catalytic domain of the enzyme, while F(0.4414) to the allosteric domain (Figure 2). In the recent study, Ebselen has shown in vitro M pro inhibition activity . We calculated Cross-spectrum (CS) for M pro and Ebselen and found a dominant peak at the F (0.1054) (Figure 3). This frequency was computationally mapped to domain 182-214, corresponding to the catalytic domain. We further searched CS of Drugbank candidates with M pro at the F(0.1923), F(0.3183), F(0.4414) and F (0.1054) to find potential M pro inhibitor candidates, with additional condition that candidates' IS contained main peaks on those values. With this search, we selected 57 candidate drugs (Tables 1 and 2). ## Molecular docking To further filtering of the selected compounds, we carried molecular docking into the catalytic and allosteric domains of SARS-CoV-2 (Tables 1 and 2). The cut-off binding energy value for the best candidates was set to -7.0 kcal/mol. From the initial docking, as the best candidates were found Mezlocillin, Camazepam and Spirapril, targeting the catalytic site. Raltegravir, Rolitetracycline, Tolvaptan, Ciclesonide and Rescinnamin were found targeting the allosteric domain. All compounds have a better docking score than Ebselen, which suggests that they could be potentially promising inhibitors of SARS-CoV-2 M pro . Because Ebselen is an inhibitor of HIV-1 capsid C-terminal domain dimerization , from this study is assumed that it analogously hinders M pro dimerization. Docking in the catalytic site showed that all candidates (Table 3, Figure 4, Figures S1-S3) interact with Cys 145 and His 41, which are essential for the catalytic activity of M pro . Types of intermolecular interactions that candidates form with amino acid residues are hydrogen bonds, aromatic π-π, alkyl-π, S-π and cation-π interactions. Table 3. Protein-ligand interactions in the catalytic site. From the docking results in the allosteric site (Table 4, Figure 5, Figures S4-S7), noticeable protein-ligand interactions are with Lys 5, which is next to Arg 4, an essential residue for the dimerization process . The other interacting residues (Lys 137, Gly 138, Glu 290, Tyr 126 and Leu 286) are in accordance with those found from the biological assembly in PDB 6LU7 (Table S1). Table 4. Protein-ligand interactions in the allosteric domain. ## Discussion Current prevention and treatment options for SARS-CoV-2 infections are insufficient due to lack of approved drug therapy or vaccines . In a search for preventive and therapeutic options to counter threats of pandemics, the fundamental problem is that drug development is a costly, time-consuming, and a risky enterprise. Therefore, drug repurposing is a promising therapeutic strategy for many viral diseases and the most realistic in the present pandemic. Various predictive in silico approaches have been applied to identify drug repositioning opportunities against SARS-CoV-2 . In this work, we have used the ISM for the structure/function analysis of the highly conserved SARS-CoV-2 protein M pro and identified the key informational characteristic of the protein, which corresponds to the protein key biological function. The ISM was recently used for prediction of potential receptor, natural reservoir, tropism, and therapeutic/vaccine target of COVID-19 . In another recent study, ISM was used for the analysis of the COVID-19 Orf3b, suggesting that this protein acts as a modulator of the interferon signalling network . To select drug candidates for SARS-CoV-2 M pro inhibitor, further in this work, we used the VS protocol based on the application of successive filters. First, the ISM-SM was used for the fast screening of large compound libraries through candidate selection at a specific frequency value. Molecules were treated as quasi-linear entities, analogous to peptides, and ISM was applied to predict potential candidates for the SARS-CoV-2 M pro . Previously this novel approach in bioinformatics treatment of small molecules was successfully used for analyzing GPCR drugs of Golden dataset with amino acid sequences of corresponding receptors. The essential information that can be extracted from protein-ligand CS spectra is the domain of the binding site in the corresponding receptor . As the second step of VS to meet short-range compatibility, we used molecular docking. One of the best ranked allosteric inhibitors from our computational study is ciclesonide. Another computational study also found ciclesonide as a potential inhibitor of M pro . In in vitro studies, ciclesonide showed good antiviral activity against SARS-CoV-2, however against a different target . The potential multitarget activity of ciclesonide may help to overcome drug resistance in COVID-19. Additional favorable results were reported from studies identifying that ciclesonide inhalant may improve the respiratory status in severe COVID-19-induced pneumonia and in cases of mild to mid-stage COVID-19 . Ciclesonide is a safe drug commonly used for inhalation in premature babies and newborns, as well as the elderly. It is effective in controlling chronic inflammation of the respiratory tract and the only steroid that showed anti-SARS-CoV-2 activity . These studies gave rise to a recently initiated an open-labeled, randomized Phase 2 clinical trial to evaluate the antiviral effect of ciclesonide on the reduction of viral load in patients with mild COVID-19 . Raltegravir, the first approved human immunodeficiency virus type 1 (HIV-1) integrase inhibitor and the best M pro allosteric inhibitor according to our study, was among the 30 compounds, with potential SARS-CoV-2 activity shown by a joint research team of the Shanghai Institute of Materia Medica and Shanghai Tech University against SARS-CoV-2 from in silico and in vitro analysis . Another of the best ranked allosteric inhibitors from our computational study was rolitetracycline, the first of the semi-synthetic tetracyclines. In recent molecular docking study it also showed the best binding with the catalytic center of the SARS-CoV-2 M pro through binding with CYS 145 and HIS 41 . Tolvaptan is also in the group of potentially the best M pro allosteric inhibitors The efficacy and safety of tolvaptan therapy was reported in patients with the COVID-19-associated syndrome of inappropriate antidiuretic hormone secretion . In our work, mezlocillin was the best candidate for catalytic site inhibitor of M pro .The same results was also reported from another in silico study . The second best best candidate for catalytic site inhibitor, camazepam, is a benzodiazepine. Although an anxiolytic, there have been earlier reported antiviral activities of benzodiazepines . According to our study, spirapril and ACE inhibitors, could be a promising catalytic site inhibitor candidate of SARS-CoV-2 M pro . It was proposed that ACE inhibitors could have both potentially harmful and beneficial effects on COVID-19. Membrane-bound angiotensinconverting enzyme 2 (ACE2) participates in the entry of SARS-CoV-2 into human cells, and animal studies show that ACE inhibitors could upregulate ACE2 expression, and on the other side, the beneficial effect can be expected with upregulated ACE2 converting angiotensin II to angiotensin-(1-7) with potentially advantageous vasodilatory and anti-inflammatory properties . We can assume that ACE inhibitors through M pro inhibition, in addition to the already assumed theory of the advantageous effect of upregulated ACE2, could explain that treatment with ACE-inhibitors is associated with less severe disease in SARS-CoV-2 infection . A number of drugs selected in our study as repositioning candidates that potentially bind to the catalytic site were also identified in other in silico studies that analyzed the same target. In the molecular docking study, bacampicillin was among the best repurposed drugs against the main protease of SARS-CoV-2 . Carbinoxamine showed in another docking study potential activity against SARS-CoV-2 M pro . In the docking study of FDA approved drugs against protease and spike protein of COVID-19, paromomycin was found to have a strong binding affinity against both Spike and M pro of SARS-CoV-2 according to its glide score . Phensuximide was found in molecular docking simulations of FDA-approved small compounds associated with protection against COVID-19, M pro . In VS-based study magnesium Ascorbate, a buffered (non-acidic) form of vitamin C, was found to be the top lead compound among 106 nutraceuticals against SARS-CoV-2 s M pro , and tizanidine was amongst the 11 approved drugs predicted to show a high binding affinity in VS study with M pro . The potential allosteric inhibitors found in our study were also selected as M pro inhibitors in several following in silico studies. In the VS study, cefotiam was found among eight compounds potential SARS-CoV-2 M pro inhibitors . In another docking study, voriconazole, tobramycin and kanamycin showed potential activity against SARS-CoV-2 M pro . Ospemifene was among 51 hits selected in silico against M pro , and propylthiouracil was among the top 20 drugs showing the highest docking score . Besides, we found oseltamivir, influenza neuraminidase inhibitor to have higher docking score against SARS-CoV-2 M pro than ebselen (Table 2). The use of oseltamivir was already reported during the COVID-19 epidemic in China, either with or without antibiotics, and clinical trials are ongoing with oseltamivir with multiple combinations with chloroquine and favipiravir . Selected drugs from our computational study may represent an initial step for further experimental investigations in a quest for safe, new treatments for COVID-19. ## Informational spectrum method In this work, we analyze SARS-CoV-2 protein M pro protein using the informational spectrum method (ISM). A comprehensive explanation of the sequence analysis based on ISM is available elsewhere . According to this approach, sequence (protein or DNA) is transformed into a signal by assignment of numerical values of each element (amino acid or nucleotide). These values correspond to electron-ion interaction potential (EIIP) , determining the electronic properties of amino acid/nucleotides, which are essential for their intermolecular interactions. The EIIP descriptors are easily calculated using following formulas: Where i is type of the chemical element, Z is valence of the i-th chemical element, n is number of the i-th chemical element atoms in the compound, m is number of types of chemical elements in the compound and N is total number of atoms. The EIIP signal is then transformed using Fast Fourier Transform (FFT) into information spectrum (IS) as a representation of a sequence in the form of a series of frequencies and amplitudes: Where m is the summation index, x(m) is the m-th member of a given numerical "signal" series (from a transformed, encoded primary protein sequence in our case), N is the total number of points in this series), n is the number of a discrete frequency (ranging from 1 on up to N/2) in the DFT, X(n) are the discrete Fourier transformation amplitude coefficients corresponding to each discrete frequency n, and 2π*(n/N) is the phase angle at each given m in the amino-acid series of the protein in question. However, in the case of protein analysis, the relevant information is primarily presented in energy density spectrum, which is defined as follows: By this, the virtual spectroscopy method is feasible to analyze protein sequences without any previous experimental data functionally. Its extension for small molecules, ISM-SM was developed and published recently . A small molecule is imported in smiles notation and decoded by atomic groups into an array of corresponding EIIP values. Using FFT, the corresponding IS of a small molecule is computed. This spectrum is further multiplied by IS of the protein receptor to obtain a Cross-spectrum (CS). Cross-spectral function is the function which determines common frequency characteristics of two signals. For discrete series it is defined as follows: Where X(n) and DFT coefficients of the series x(m), and Y(n)* are complex conjugated DFT coefficients of the series Y(m). From common frequencies in CS, one can determine whether protein interacts with small molecule and determine the corresponding binding region in the protein. ## Data preparation Protein sequences were downloaded from downloaded from the UniProt database (https://www.uniprot.org/) with the following accession numbers: Gaba alpha subunit , P14867 D-alanyl-D-alanine carboxypeptidase Q75Y35, Angiotensin-converting enzyme P12821, HIV 1 Integrase Q7ZJM1, 30S ribosomal protein S9 P0A7X3, Vasopressin V2 receptor P30518. FASTA COVID-19 M pro sequence was downloaded from RCSB, PDB ID 6LU7 and corresponding IS was calculated. A set of 1490 approved Drugbank drugs with corresponding smiles was subjected to IS and CS calculation with M pro . All calculations were carried using our in-house software. ## Molecular docking Molecular docking of selected candidates into the crystal structure of M pro was carried. Receptor threedimensional structure was downloaded from RCSB, PDB ID 6LU7 . All ligands, waters and ions were removed from PDB file. Two grid boxes with dimensions 24 x 24 x 24 were set to span all amino acid residues interacting with co-crystallised inhibitor N3 in case of catalytic site. For the allosteric domain, it was set to span the residues interacting in the dimer interface. The (x,y,z) centers of the grid boxes were (-11.775, 13.910, 66.706) for the catalytic site and (-22.521, 1.749, 50.782) for the allosteric domain. Selected drugs from previous step were converted from smiles to 3D SDF and further to PDB files and protonated at physiological pH. Geometry optimization was carried in MOPAC 2016 at PM7 level of theory . Exhaustiveness was set to 50. Molecular docking was carried in Autodock Vina , implemented in VS software PyRx . Figures were made in BIOVIA Discovery Studio 2017 , Schrodinger Maestro 11.1 and Origin 9.0 software .

chemsum

{"title": "Drug repurposing for Candidate SARS-CoV-2 main protease inhibitors by a novel in silico method", "journal": "ChemRxiv"}

fast_and_stable_vapochromic_response_induced_through_nanocrystal_formation_of_a_luminescent_platinum

## Abstract: A hybrid vapoluminescent system exhibiting fast and repeatable response was constructed using periodic mesoporous organosilica with bipyridine moieties (Bpy-pMo) and a pt(ii) complex bearing a potentially luminescent 2-phenylpyridinato (ppy) ligand. An intense red luminescence appeared when the pt(ii)-complex immobilised Bpy-pMo was exposed to methanol vapour and disappeared on exposure to pyridine vapour. The ON-OFF vapochromic behaviour occurred repeatedly in a methanol/ pyridine/heating cycle. Interestingly, a rapid response was achieved in the second cycle and cycles thereafter. Scanning and transmission electron microscopies (SEM/TEM), absorption and emission, and nuclear magnetic resonance spectroscopies, mass spectrometry, and powder X-ray diffraction indicated that methanol vapour induced Si-c cleavage and thus liberated [pt(ppy)(bpy)]cl (bpy = 2,2′-bipyridine) from the BPy-PMO framework. Furthermore, the self-assembling properties of the Pt(II) complex resulted in the formation of highly luminescent micro/nanocrystals that were homogeneously dispersed on the porous support. The unique vapoluminescence triggered by the unprecedented protodesilylation on exposure to protic solvent vapour at room temperature is attributable to Bpy-pMo being a giant ligand and an effective vapour condenser. Consequently, this hybrid system presents a new strategy for developing sensors using bulk powdery materials. Porous materials are extremely attractive for the creation and control of nano-and meso-spaces where specific effects and novel phenomena are expected. Various nano-and mesoporous structures have been developed using inorganic, organic, and hybrid materials, which have been applied in catalytic systems, sensors, gas storage, electronic devices, and biological systems . To build a mesoporous silica structure for luminescence, De Cola and co-workers developed mesoporous silica particles containing photofunctional metal complexes in their pores by using amphiphilic metal complexes as surfactants 4,5 . It was remarkable discovery that the confinement effect of mesoporous silica induced an enhancement of the luminescence intensity of the included metal complexes. In contrast, periodic mesoporous organosilicas (PMOs) are relatively new materials with uniformly distributed organic and inorganic moieties within their frameworks . Compared with other hybrid porous materials such as mesoporous silica and metal-organic frameworks (MOFs), PMOs have an advantage in that they possess large regulated pores of several nano-meters in diameter (i.e. mesopores), which allow construction of particular chemical reaction fields. Inagaki and co-workers also developed a sophisticated mesoporous organosilica that incorporated various aromatic organic moieties 10 . In addition to the high stability and light-harvesting effect of the PMOs, assembly control of the functional metal complex systems was expected 11,12 . In particular, BPy-PMO, composed of silicate and 2,2′-bipyridine (bpy) connected by covalent bonds, allowed the metal ions to be captured through direct coordination to the periodic mesoporous framework (Fig. 1a) 13 . This PMO was then utilised as a photocatalytic system with polypyridine-Ru(II) and Ir(III), Re(I), Cu(II), and Pt(II) complexes as well as other selective catalytic systems 10, . However, the full potential of the periodic structure of this huge ligand (BPy-PMO) has not yet been studied. Pt(II) complexes are known to exhibit a characteristic colour and luminescence when the square-planar complex units are stacked with Pt•••Pt short contacts . Thus, the colour and luminescence of such assembly-induced luminescent Pt(II) complexes should be extremely sensitive to changes in the stacking structures. External stimuli such as heat, pressure, vapour, and mechanical forces readily induce such changes. In particular, a vapour-induced reversible colour change, vapochromism, is a characteristic property of Pt(II)-complex assemblies and has attracted significant attention based on an easy sensing of the environment . Various vapochromic systems have thus been developed using different metal complexes and organic crystals . For vapochromic materials, however, the response rate to dilute gaseous molecules is an issue to be overcome in bulk systems, although it can be used to detect long-term changes like such as blue silica-gel as a moisture indicator. Aiming at a rapid response, various systems have recently been developed using MOFs 36,37 and supramolecular porous crystals 33,38 , which have a large surface area directly accessible to the inner chromophores. The fabrication of thin films is an effective way to achieve a rapid response . Meanwhile, soft materials have an advantage in terms of stimulus responsiveness and chemical sensing . However, higher-order systems with less fluctuation are desired to improve the accuracy. In this context, a new category of promising stimulus-responsive materials that have both properties of soft and high-order materials, termed soft crystals, is proposed 47 . In addition, if the vapochromic properties are accompanied by changes in luminescence, a higher vapour-sensitivity will be possible. Therefore, the exploration of more sophisticated vapoluminescent systems having high sensitivity and selectivity remains a challenging subject. In this study, we developed a unique vapochromic luminescent system that exhibited a rapid and stable vapour response with clear colour changes using integrated coordination sites and the vapour absorptivity of BPy-PMO. By immobilisation of a Pt(II) complex on BPy-PMO with a high ratio, the [{Pt(ppy)} n (BPy-PMO)]Cl n (ppy = 2-phenylpyridinate, Fig. 1b) (Pt-PMO) system successfully achieved vapour-induced nanocrystal formation and repeatable vapour response. ## Results Synthesis and characterisation. Pt(II)-immobilised BPy-PMO (Pt-PMO) with a favourable immobilisation amount was successfully obtained by reacting BPy-PMO with [Pt(ppy)Cl(DMSO)] (DMSO = dimethyl sulfoxide) in CH 2 Cl 2 instead of DMSO, which was used in our previous report (Fig. 2a) 16 . The immobilised amount was estimated to be 12% based on the Pt/Si ratio using X-ray fluorescence (XRF) spectroscopy and UV-Vis absorption spectra (Fig. S1). The powder X-ray diffraction (PXRD) pattern of Pt-PMO clearly showed characteristic peaks at 2θ = 1.84° (d = 4.8 nm), 7.7°, 15.4°, and 23.0° after the immobilisation, indicating that the ordered mesoporous structure and pore wall structure of BPy-PMO was maintained even after the immobilisation of the platinum(II) complex (Fig. 2b). To discuss the local structure around the Pt sites in Pt-PMO, the extended X-ray absorption fine structure (EXAFS) spectra were measured. The k 3 -weighted EXAFS spectrum of the Pt-L III edge of Pt-PMO and the corresponding Fourier transform are shown in Figs S2a and 2c, respectively. The two broad peaks at ~1.4-1.9 and 2.2-2.7 were well-explained by the local structure of the first and second shells around the Pt site in the model complex [Pt(ppy)(bpy)](PF 6 ), whose crystal structure was elucidated by single crystal X-ray analysis of a linear chain structure of the Pt(II) complex units with a moderate Pt•••Pt distance (3.6048(1) ) at 93 K (Fig. S3). The X-ray absorption near-edge spectrum (XANES) of Pt-PMO was also in good agreement with that of [Pt(ppy)(bpy)](PF 6 ) (Fig. S2b). The X-ray photoelectron spectroscopy (XPS) revealed that the binding energies of Pt 4f 7/2 and Pt 4f 5/2 signals of Pt-PMO (72.3 and 75.1 eV) were almost identical to those of the model complex [Pt(ppy)(bpy)]Cl (72.1 and 75.3 eV) (Fig. 2d, Table S1), which further supported the desired complexation in the mesopores of BPy-PMO. The nitrogen adsorption isotherm of Pt-PMO revealed that the amount of adsorbed nitrogen was lower in comparison with BPy-PMO (Fig. S4). The Brunauer-Emmett-Teller (BET) surface area (S BET ) decreased to 393 m 2 g −1 for Pt-PMO from 680 m 2 g −1 for BPy-PMO, and the non-linear density functional theory (NLDFT) analysis also revealed that the pore diameter of Pt-PMO (3.9 nm) was smaller than that of BPy-PMO (4.7 nm), indicating that the platinum(II) complex was immobilised in the mesopores (Table S2). Using the average particle diameter (695 nm) determined through dynamic light scattering (DLS) measurements and the pore diameter of BPy-PMO (4.7 nm), the ratio of the inner and outer surfaces area of the mesopore of BPy-PMO was roughly estimated to be approximately 105 (Fig. S5), which meant that 99% of the surface was inside the BPy-PMO. Assuming a three-layered structure of the wall and a full occupation of the outer surface bpy moieties by the Pt complexes 13 , 37% of the bpy moieties on the inner surface of the BPy-PMO were estimated to be occupied by the Pt-ppy units for the present Pt-PMO sample with 12% Pt/Si (Fig. S5). The emission properties of Pt-PMO provide useful information regarding the assembly of the Pt(II) complex units. Pt-PMO with 12% Pt/Si exhibited a broad emission spectrum at 77 K (Fig. 3, Table S3). This is in contrast to that for Pt-PMO with its small loading amount of the platinum complex (1% Pt/Si) at 77 K, which showed a typical 3 ππ* emission with a vibronic structure similar to that for the model complex [Pt(ppy)(bpy)] + in a dilute methanol solution (Fig. 3). Considering that there was no change in the diffuse reflectance spectra of 1% and 12% Pt-PMO (Fig. S6), the broad luminescence for 12% Pt-PMO was attributable to the emissions from the excimeric or dimeric form of platinum complexes units. In fact, the excimer emission of another PMO incorporating biphenyl (Bp-PMO) has been reported previously 48,49 . Additionally, an emission at ~450 nm was observed in the system of BPy-PMO, which is completely different from that observed for 12% Pt-PMO (Fig. 3). This is an important point because the broad emission indicates an assembled immobilisation of the Pt(II) complex units in the meso-pore that allows the electronic interactions between the Pt(II) complex units in the excited states. Vapochromic response. Interestingly, Pt-PMO exhibited a distinct vapochromic behaviour. The yellow powder sample of Pt-PMO (12% Pt/Si) turned into a reddish colour upon exposure to MeOH vapour, accompanied by the appearance of a new band at 530 nm, as shown in the diffuse reflectance spectra (Fig. 4). Simultaneously, a broad emission band appeared with a maximum at 630 nm, and the emission intensity reached the maximum level after 8 h under nearly saturated vapour pressure at room temperature (Fig. S7). This emission band at 630 nm and the quantum yield (Φ) of 0.11 were consistent with that observed for the model complex [Pt(ppy)(bpy)]Cl in the solid state (Φ = 0.12 and τ = 123 ns at 298 K, as shown in Table S3), suggesting that the S1. emission is originated from the triplet metal-metal-to-ligand charge transfer ( 3 MMLCT) in an excited state arising from the Pt•••Pt electronic interaction 19,22 . In fact, the red emission of both the orange powdery sample (called Pt-PMO-R) and the crystalline sample of [Pt(ppy)(bpy)]Cl were red-shifted by 35-40 nm with sharpening at 77 K indicating typical spectral features of 3 MMLCT emissions (Figs S8 and S9). These results suggest that a regularly assembled form of the Pt(II) complex units was formed on the BPy-PMO upon exposure to the MeOH vapour. Next, Pt-PMO-R was heated at 403 K for 12 h to remove the MeOH molecules (Fig. S10a). The emission band did not change even after the removal of MeOH (Fig. S10b), which suggested an excellent stability of Pt-PMO-R. However, when it was exposed to the pyridine vapour, its colour readily changed from red to light yellow (hereafter, called Pt-PMO-LY) within ~10 min, demonstrating a considerable blue-shift (~140 nm) of the absorption edge (Fig. 4b). Along with the absorption spectral change, the emission intensity decreased and disappeared eventually (Fig. S11a). At 77 K, Pt-PMO-LY exhibited a green emission at λ max = 481 nm with a clear vibrational progression (Fig. S12) attributable to the ligand-based 3 ππ* emission, similar to that for the model complex S3), indicating the negligible Pt•••Pt interactions in this form in contrast to those of Pt-PMO-R. On heating the Pt-PMO-LY form at 353 K for 30 min for the removal of the pyridine vapour, it returned to a yellow powder (called Pt-PMO-Y), showing almost the same absorption spectrum as the Pt-PMO form (Fig. 4b). Surprisingly, the vapochromic response of Pt-PMO-Y was drastically accelerated as compared with that of the as-synthesised Pt-PMO. Upon exposure to methanol vapour, the colour of Pt-PMO-Y rapidly turned from yellow to red within tens of seconds and the Pt-PMO-R form was obtained again. Along with this change in colour, the broad emission band at 630 nm also increased rapidly upon exposure to methanol vapour (Fig. 5a). Subsequently, exposing Pt-PMO-R (second cycle) to pyridine vapour resulted in the regeneration of the Pt-PMO-LY form again. This three-state vapochromic cycle of Pt-PMO-R/Pt-PMO-LY/Pt-PMO-Y showed a high reversibility, as indicated in Fig. 5b, which was pursued through changes in the emission quantum yield. Furthermore, the present system also enabled the vapochromic response in a low relative pressure (P/P 0 ) region because of the mesoporous structure. Pt-PMO-Y showed a vapochromic luminescence even at P/P 0 = 0.1 (Fig. S14). As previously reported, the pore walls of BPy-PMO contains a considerable amount of silanol (Si-OH) groups 10 , which is expected to capture methanol through hydrogen bonding even at low relative pressure, in addition to the capillary condensation effect in the mesoporous channel. Indeed, thermogravimetric (TG) analysis revealed that BPy-PMO can adsorb a considerable amount of methanol molecules, whereas almost no adsorption was observed after the protection of silanol by trimethylsilyl groups (Fig. S10a). The methanol vapour adsorption isotherm (Fig. S15) showed a large amount of methanol adsorption at a very low relative pressure (~20 mol•mol −1 per bpy unit at P/P 0 = 0.1), and a significant hysteresis within the region of P/P 0 > 0.6. Therefore, the high affinity of BPy-PMO for methanol was responsible for the low detection limit of Pt-PMO-Y, and the fast vapochromic response. Vapochromic mechanism. To investigate the origin of this vapochromism in detail, we used microscopic techniques and PXRD to observe the morphological changes occurring in the vapochromic cycle. Interestingly, several new diffraction peaks appeared only for the Pt-PMO-R stage (red line in Fig. 6) in addition to those for the periodic structure of BPy-PMO, which indicated the formation of a new crystalline species on BPy-PMO through the methanol vapour exposure. After exposure to pyridine vapour, these crystalline peaks disappeared, with the BPy-PMO peaks remaining (blue line in Fig. 6); essentially the same patterns were also observed after heating (orange line in Fig. 6). The diffraction peaks for Pt-PMO-R were regenerated using the second methanol vapour exposure (lower red line in Fig. 6). The scanning and transmission electron microscopic (SEM and TEM, respectively) observations with energy dispersive X-ray spectrometry (EDS) clearly indicated the morphological changes (Figs 7 and S16). Initially, no crystalline species were observed on the Pt-PMO particles, and the EDS elemental map showed that Pt atoms were homogeneously immobilised on BPy-PMO (Figs 7a and S16a). After methanol vapour exposure (i.e. the formation of Pt-PMO-R), the SEM image showed that crystals with a length of 1-10 μm appeared on the BPy-PMO substrate (Fig. 7b), as suggested by the PXRD data. The EDS elemental map clearly showed that Pt atoms were mainly localised in the crystals (Fig. 7b), and the EDS single-point analysis suggested that crystals of a Pt(II) complex were generated (Fig. S16b). Such crystalline materials were observed neither for Pt-PMO-LY (Fig. 7c) nor Pt-PMO-Y (Fig. 7d). After the second methanol vapour exposure, crystalline materials were observed again. However, in contrast to the first observation, these crystals were much smaller with a length of 80-400 nm (Fig. 7e). These nano-sized crystals were confirmed to be those of a Pt(II) complex through EDS single-point analysis and TEM (Fig. S16e). The formation of such nanocrystals is likely the reason for the fast vapochromic response during the second cycle, as discussed in the next section. To identify this crystalline species on Pt-PMO-R, mass spectrometry (MS) and NMR analyses were conducted. Importantly, both ESI-MS for the soluble species of Pt-PMO-R and MALDI-MS for the powdery sample of Pt-PMO-R showed the main signal at m/z = 505.1 (Fig. S17), which is consistent with the Pt(II) complex ion, {Pt(ppy)(bpy)} + , including an isotropic peak pattern. The 1 H NMR spectrum of the extracted species from Pt-PMO-R with methanol was also identical to that of [Pt(ppy)(bpy)]Cl (Fig. S18). These results suggest that [Pt(ppy)(bpy)] + was formed through the Si-C(bpy) bond dissociation from the BPy-PMO framework. It is notable that the double Si-C bond dissociations occurred under such mild conditions through methanol vapour exposure considering that BPy-PMO is a known stable framework and Si-C bond dissociations typically occur only under the strongly acidic or basic conditions 13,50 . An immobilised Pt complex may allow a nucleophilic attack of methanol molecules on BPy-PMO (i.e. protodesilylation) 51 while maintaining the entire framework of BPy-PMO. This vapour-triggered protodesilylation process was further evidenced through the deuterium-labelling experiments, in which the deuterated complexes, [Pt(ppy)(bpy-d 2 )] + and [Pt(ppy)(bpy-d 1 )] + , were obtained when Pt-PMO was exposed to methanol-d 4 vapour, and confirmed through ESI-MS and 1 H NMR (Figs S19 and S20). Based on these results, we concluded that the methanol-vapour induced crystal formation of [Pt(ppy)(bpy)]Cl occurred through Si-C bond dissociation in Pt-PMO-R. Indeed, the diffraction peaks of the crystalline species of Pt-PMO-R qualitatively agreed with those of the model complex [Pt(ppy)(bpy)]Cl in a methanol atmosphere (Fig. S21). The XRF spectra indicated that ~2/3 of the immobilised Pt(II) complex was detached to form crystals through vapour-induced protodesilylation (Fig. S22). This conclusion was further supported by the fact that the BET surface area and average pore diameter of Pt-PMO-R (420 m 2 g −1 and 4.2 nm, respectively, as shown in Fig. S4 and Table S2) definitively increased from those of Pt-PMO (393 m 2 g −1 and 3.9 nm, respectively) owing to the detachment of [Pt(ppy)(bpy)] + ions through a vapour-triggered Si-C bond dissociation. Pt-PMO also exhibited a vapochromic response to other protic vapours such as H 2 O, EtOH, and i-PrOH, as indicated from the emission spectra and PXRD patterns, although it did not respond to less-polar vapours such as CH 2 Cl 2 , chloroform, and toluene (Figs S23 and S24). In addition, it is also noteworthy that the vapour-triggered protodesilylation is characteristic of metal-loaded BPy-PMOs and never occurred for a discrete Pt(II) complex bearing a precursor unit of BPy-PMO in the same conditions (Figs S25 and S26). ## Discussion The structural, spectroscopic, and microscopic investigations of the interesting behaviours of the mesoporous materials revealed not only the vapochromic mechanism but also the reason for the fast response. A schematic illustration of this vapour response cycle is shown in Fig. 8. Before the vapour exposure, the Pt(II) complex was homogeneously and densely immobilised on BPy-PMO through the coordination bonds. Based on the first methanol vapour exposure, however, the Si-C bonds on the immobilised Pt(II) complexes were cleaved using methanol vapour to form micrometre-sized crystals of [Pt(ppy) (bpy)]Cl, leading to a change in colour to red (i.e. Pt-PMO-R), as confirmed through isotropic experiments. The micrometre-sized crystals of [Pt(ppy)(bpy)]Cl on the BPy-PMO were dissolved in pyridine upon exposure to pyridine vapour, and uniformly adsorbed in the mesopores of BPy-PMO through the capillary condensation, as evidenced by SEM-EDS, resulting in a light yellow colour (i.e. Pt-PMO-LY) of the discrete molecules of the Pt(II) complex without any intermolecular interactions. This is in contrast to the lack of change in the luminescence spectrum when exposed to acetonitrile vapour instead of pyridine vapour (Fig. S11b), probably because of a weaker coordination ability of acetonitrile as compared to that of pyridine. After the removal of pyridine through heating, the Pt(II) complex was uniformly loaded onto BPy-PMO using physisorption without any Pt•••Pt interaction (i.e. Pt-PMO-Y), as suggested by the PXRD pattern (Fig. 6). It is interesting to note that the nanocrystals of [Pt(ppy)(bpy)]Cl were produced from Pt-PMO-Y on exposure to methanol vapour. This is in contrast to the recrystallisation of [Pt(ppy)(bpy)]Cl from a normal pyridine solution, which fails because of the decomposition of the complex (Fig. S27). The meso-space of BPy-PMO must also be effective also for the crystallisation. It is reasonable for the vapochromic response in the second and subsequent exposures to be much faster than during www.nature.com/scientificreports www.nature.com/scientificreports/ the first exposure because the Si-C bond cleavage is no longer necessary. In addition, it should be noted that a uniform loading of [Pt(ppy)(bpy)]Cl is key to obtaining such a fast and clear vapochromic response. We were unable to obtain similar results by using a mixture of [Pt(ppy)(bpy)]Cl and BPy-PMO instead of Pt-PMO (Fig. S28). For many vapochromic materials, the vapour response typically requires several hours to days in a bulk state. In addition, vapochromic materials generally suffer from poor stability in the vapour-included forms, and quickly restore the original colour through the desorption of vapour molecules . Stable and as-desired changeable systems are necessary for perfectly controlled vapochromic systems. It is noteworthy that the present three-step vapochromic system Pt-PMO enables a rapid response within several tens of seconds through exposure to methanol vapour in a bulk state accompanied by an ON-OFF switching of the emission. In addition, once Pt-PMO-R is formed in response to methanol vapour, it demonstrates high stability after its removal, maintaining the "vapour-detection history, " which has been difficult to observe thus far except for a few examples 52,53 . In addition, it is also extremely easy to erase the history (switch-off) through pyridine vapour exposure. In conclusion, we successfully constructed and elucidated a superior vapochromic system utilising a cooperative phenomenon arising from a huge mesoporous ligating support, BPy-PMO, with an assembly-induced Pt(II) complex. Our findings provide a new guiding principle for the development of photofunctional materials. ## Methods Materials. Reagents and solvents were purchased from commercial sources and used without further purification. BPy-PMO 13 , trimethylsilyl-protected BPy-PMO 13 , Pt-PMO (1% Pt/Si) 16 , and [Pt(ppy)Cl(DMSO)] 54 were prepared as previously described. ## Synthesis of model complexes, [pt(ppy)(bpy)]X (X = cl − ). Although the synthetic procedure of [Pt(ppy)(bpy)]Cl has been previously reported 16,55 , in this study, it was prepared using another method that is similar to the immobilisation of the Pt(II) complex. A solution of [Pt(ppy)Cl(DMSO)] (117 mg, 0.25 mmol) and bpy (43 ## Measurements. Elemental analyses and electrospray-ionisation mass spectrometry (ESI-MS) were conducted at the analysis centre of Hokkaido University. The 1 H NMR and 1 H-1 H COSY NMR spectra were acquired on a JEOL ECZ-400S or EX-270 spectrometer. Energy dispersive XRF spectra were acquired on a JEOL JSX-3100RII spectrometer using a Rh target. DLS analyses were conducted using an OTSUKA ELSZ-1000SCl analyser. Nitrogen and methanol vapour adsorption isotherms were measured using an automatic BELSORP-max (MicrotracBEL Co.) volumetric adsorption apparatus. Pore-size distributions were calculated using the density functional theory (DFT) method (DFT kernel: N 2 at 77 K on silica, cylindrical pores, and NLDFT equilibrium model). Thermogravimetry-differential thermal analysis (TG-DTA) measurements were recorded using a Rigaku Thermoplus EVO TG-DTA 8120 with Al sample pans under an Ar flow. Matrix-assisted laser desorption/ionisation mass spectrometry (MALDI-MS) was conducted using a Bruker Microflex LRF spectrometer in a linear mode. Electron microscopy. TEM was performed on a JEOL JEM-2010 FASTEM microscope at an accelerating voltage of 200 kV. Field-emission scanning electron microscopy (FE-SEM) using EDS was conducted on a JEOL JSM-7100F microscope at an accelerating voltage of 15 kV equipped with an electron backscatter diffraction detector (Oxford Aztec Energy-HKL). Powder X-ray diffraction. PXRD measurements were conducted using Cu K α radiation (λ = 1.5418 ) on a Bruker D8 Advance diffractometer equipped with a graphite monochromator and a one-dimensional LinxEye detector, or a Rigaku SPD diffractometer at the BL-8B beamline of the Photon Factory (PF), Japan. The synchrotron X-ray wavelength was 1.5455 . rescence mode at BL-12C beamline of the PF, Japan. The incident X-ray was made monochromatic using a Si(111) double-crystal monochromator. k 3 -weighted EXAFS function, k 3 χ(k), was extracted from the raw X-ray absorption data, which were obtained using the ATHENA software 56 . A Fourier transform of the EXAFS function was conducted within the k range of 2-14 −1 using a Hanning window, and it was fit to the Pt-PMO structure using ARTEMIS software 56 in the range (R) of 1-3 corresponding to the first and second shells. In this fitting, the amplitude and phase shift for all scattering paths were also calculated using FEFF6L in the ARTEMIS software 56 . X-ray photoelectron spectroscopy. XPS were recorded on a JEOL JPC-9010MC spectrometer at a vacuum pressure of less than 1 × 10 −5 Pa in an analysis chamber. A standard Al K α excitation source (1486.6 eV) was used during all experiments. Each sample was placed on carbon tape. The binding energies were calibrated against the carbon 1s (284.6 eV) peak position. A spectral fitting was conducted using a Gaussian-Lorentzian product function. Single Crystal X-ray crystallography. Diffraction data for [Pt(ppy)(bpy)](PF 6 ) were collected using a Rigaku XtaLAB-PRO diffractometer with a Hypix-6000HE area detector and a multilayer mirror-monochromated Cu K α radiation (λ = 1.54184 ) at 93 K. The crystal was mounted in a microloop with Paratone-N oil. Diffraction data were collected and processed using CrysAlisPro 57 at 93 K. The structures were solved using a SHELXT 58 structure solution program applying intrinsic phasing, and refined using the SHELXL 58 refinement package with least squares minimisation. Because the solvent molecules in the unit cell were significantly disordered and could not be properly modelled, the solvent mask routine implemented in Olex2 59 was used to calculate the solvent disorder area and remove its contribution to the overall intensity data. All non-hydrogen atoms were refined anisotropically and H atoms were refined using the riding model. The crystallographic data are summarised in Table S4 and have been deposited at the Cambridge Crystallographic Data Centre (CCDC 1910647). The molecular and stacking structures are shown in Fig. S3, and selected bond distances and angles are summarised in Table S5. Absorption and emission measurements. UV-Vis absorption spectra were recorded on a Shimadzu UV-2500PC spectrophotometer. UV-Vis diffuse reflectance spectra were obtained using the same spectrophotometer equipped with an integrating sphere apparatus. The reflectivity of the solid samples was converted using the Kubelka-Munk function. Emission spectra were recorded on a JASCO FP-8600 spectrophotometer. Luminescence quantum yields were recorded on a Hamamatsu Photonics C9920-02 absolute photoluminescence quantum yield measurement system equipped with an integrating sphere apparatus and a 150 W CW xenon light source. The accuracy of the instrument was confirmed based on a measurement of the quantum yield of anthracene in ethanol (Φ = 0.27) 60 . The emission lifetime measurements were conducted using a Hamamatsu Photonics Quantaurus-Tau C11367 system.

chemsum

{"title": "Fast and stable vapochromic response induced through nanocrystal formation of a luminescent platinum(II) complex on periodic mesoporous organosilica", "journal": "Scientific Reports - Nature"}

intense_near-infrared-ii_luminescence_from_nacef₄:er/yb_nanoprobes_for_<i>in_vitro</i>_bi

## Abstract: Near-infrared (NIR) II luminescence between 1000 and 1700 nm has attracted reviving interest for biosensing due to its unique advantages such as deep-tissue penetration and high spatial resolution.Traditional NIR-II probes such as organic fluorophores usually suffer from poor photostability and potential long-term toxicity. Herein, we report the controlled synthesis of monodisperse NaCeF 4 :Er/Yb nanocrystals (NCs) that exhibit intense NIR-II emission upon excitation at 980 nm. Ce 3+ in the host lattice was found to enhance the luminescence of Er 3+ at 1530 nm with a maximum NIR-II quantum yield of 32.8%, which is the highest among Er 3+ -activated nanoprobes. Particularly, by utilizing the intense NIR-II emission of NaCeF 4 :Er/Yb NCs, we demonstrated their application as sensitive homogeneous bioprobes to detect uric acid with the limit of detection down to 25.6 nM. Furthermore, the probe was detectable in tissues at depths of up to 10 mm, which enabled in vivo imaging of mouse organs and hindlimbs with high resolution, thus revealing the great potential of these NaCeF 4 :Er/Yb nanoprobes in deep-tissue diagnosis. ## Introduction Luminescent biolabeling is a powerful technique that employs optical probes for detecting biomolecular concentration or visualizing biological events. In order to avoid auto-fluorescence and improve the signal-to-noise (S/N) ratio, several luminescent probes have emerged based on unique optical properties such as long-lived downshifting (DS) luminescence or near-infrared (NIR)-triggered upconversion (UC) luminescence. Generally, the emission lights for these probes are located below 1000 nm, which is not optimal in bioapplications since the photon scattering may limit the tissue penetration depth. To solve this problem, luminescent materials exhibiting NIR-II emission (1000-1700 nm) in the second biological window have recently been proposed as an excellent class of probes that can signifcantly reduce light scattering and increase the probing depth in bioapplications. In the past few years, continuous efforts have been dedicated to developing NIR-II probes including organic fluorophores, carbon nanotubes, and semiconductor quantum dots (QDs). However, the use of these bioprobes has several limitations. For example, organic fluorophores commonly possess poor photostability and are susceptible to photobleaching. The applicability of QDs is compromised by photoblinking and high toxicity of heavy metal elements (e.g., cadmium and selenium). Moreover, both organic fluorophores and QDs may induce high background noise owing to a small Stokes shift, which decreases the detection sensitivity for bioassays. These concerns fuel high demand for a new generation of luminescent probes to circumvent the limitations of traditional ones. 18 Lanthanide (Ln 3+ )-doped nanocrystals (NCs), as another kind of promising luminescent probe, have received growing attention due to their tunable emissions from different Ln 3+ activators. Compared with organic fluorophores and QDs, Ln 3+ -doped NCs feature long luminescence lifetime, high photostability, low toxicity and sharp f-f emission peaks. Thus, they are widely applied for in vitro bioassays and in vivo bioimaging. 23,24 Nevertheless, most previous research studies focused on the exploration of UC nanoprobes with emission light in the UV or visible range, 25,26 which may restrict the tissue penetration depth. Several Ln 3+ ion (e.g., Pr 3+ , Nd 3+ , Sm 3+ , Dy 3+ , Ho 3+ , Er 3+ , Tm 3+ and Yb 3+ ) doped NCs have been reported to emit NIR-II light. 27 However, the NIR-II quantum yields for most of these Ln 3+ -based NCs are still low for practical application. To meet the requirement of sensitive bioassays, it is urgent to develop Ln 3+ -doped NCs with highly efficient emission in the NIR-II region. In this regard, we herein report the synthesis of monodisperse and size controllable Er 3+ /Yb 3+ -doped hexagonal NaCeF 4 core-only and core/shell NCs that exhibit intense NIR-II emission upon 980 nm excitation, by virtue of the efficient Yb 3+ -Er 3+ -Ce 3+ energy transfer. The maximum NIR-II quantum yield for the NaCeF 4 :Er/Yb NCs is determined to be 32.8%, which is $17.5 times higher than that of the widely reported NaYF 4 :Er/ Yb NCs. After surface modifcation, these NaCeF 4 :Er/Yb nanoprobes can be applied for sensitive and selective detection of uric acid (UA) in human serum through a simple mixand-measure type assay, with the limit of detection (LOD) down to 25.6 nM. Moreover, the tissue penetration depth of NIR-II emission from the proposed probe is found to be higher than that of the green UC emission of NaYF 4 :Er/Yb NCs of similar particle sizes under otherwise identical conditions. After tail vein injection of hydrophilic NaCeF 4 :Er/Yb@NaCeF 4 NCs into nude mice, the biodistribution of the nanoprobes is clearly monitored for 24 h using an in vivo bioimaging system (Scheme 1). ## Results and discussion Hydrophobic and monodisperse NaCeF 4 :Er/Yb NCs were synthesized via a facile high-temperature co-precipitation method. 28 The X-ray diffraction (XRD) patterns of the asprepared NCs can be indexed to pure hexagonal NaCeF 4 (JCPDS no. 75-1924), and no traces of other phases or impurities were detected (ESI Fig. S1 †). Energy-dispersive X-ray (EDX) spectroscopy confrms the successful doping of Er 3+ /Yb 3+ ions into the NaCeF 4 host (ESI Fig. S1 †). By changing the reaction time at 320 C, the sizes and morphologies of these NCs can be fnely tailored. Specifcally, longer reaction time resulted in larger particles. As shown in Fig. 1, when the reaction time increased from 20 to 30 min, the size of the obtained NaCeF 4 :Er/ Yb NCs increased markedly from 7.1 AE 0.5 to 25.2 AE 2.7 nm (Fig. 1a-c). With further increasing the reaction time to 45 min, the coexistence of small nanospheres and large nanorods was observed (Fig. 1d), which may be attributed to an Ostwald-ripening process where small particles dissolved and big nanorods grew simultaneously. After heating for 60 or 90 min, the small nanospheres completely transformed into larger nanorods with lengths of 103.3 AE 10.9 and 200.6 AE 16.5 nm (Fig. 1e-f), respectively. Besides the core-only NCs, core/shell NCs were also synthesized through epitaxial growth of inert NaCeF 4 shells on the core-only NCs (7.1 AE 0.5 nm). These core/shell NCs, with an average size of 18.1 AE 1.9 nm, can be well dispersed in nonpolar organic solvents such as cyclohexane to form a stable transparent colloidal solution (ESI Fig. S2 †). The high-resolution TEM (HRTEM) image shows a clearly observed d-spacing of 0.308 nm, which is in good agreement with the lattice spacing in the (01 11) planes of hexagonal NaCeF 4 , indicative of the high crystallinity of the as-prepared NCs. Currently, Er 3+ /Yb 3+ -doped fluorides (e.g., NaYF 4 ) with low phonon energy are frequently reported as UC materials. For the typical UC emission process, the Yb 3+ ion is usually used as the sensitizer to harvest 980 nm photons. An Er 3+ ion is then excited to its excited states via two or more successive energy transfers from Yb 3+ ions in close proximity, followed by radiative relaxation, resulting in UC emission of a higher-energy photon (Fig. 2a). 29 Nevertheless, the energy gap between the 2 F 5/2 and 2 F 7/2 levels of Ce 3+ ($2300 cm 1 ) is close to that of the 4 I 11/2 -4 I 13/2 energy gap ($3700 cm 1 ) of Er 3+ . Therefore, for NaCeF 4 :Er/Yb NCs, the 4 I 13/2 level of Er 3+ is signifcantly populated through the efficient phonon-assisted nonradiative relaxation from the 4 I 11/2 level facilitated by Ce 3+ ions. 30 Upon excitation at 980 nm, intense DS emissions centered at $1530 nm that are ascribed to the 4 I 13/2 / 4 I 15/2 transition of Er 3+ were detected for all the synthesized Er 3+ /Yb 3+ co-doped NaCeF 4 NCs (ESI Fig. S3 and S4 †). With the size increasing from 7.1 nm to 200.6 nm, the NIR-II emission intensity increased by 4.1 times, and the effective PL lifetime of 4 I 13/2 was found to increase from 1.53 to 5.60 ms (ESI Fig. S5 †). The NIR-II absolute quantum yield (QY), defned as the ratio of the number of emitted photons to the number of absorbed photons, was determined to be as high as 32.8% for NaCeF 4 :Er/ 2b). In sharp contrast, the visible UC emissions for Er 3+ were negligibly weak in NaCeF 4 :Er/Yb NCs, due to the effective depopulation of 2 H 11/2 , 4 S 3/2 , and 4 F 9/2 levels in the presence of Ce 3+ ions (Fig. 2c). The NIR-II absolute QYs were determined to be 1.9%, 5.6% and 19.5% for NaYF 4 :Er/Yb, NaCeF 4 :Er/Yb core-only, and NaCeF 4 :Er/Yb@NaCeF 4 core/shell NCs, respectively. To make the OA-capped NaCeF 4 :Er/Yb NCs hydrophilic for bioapplications, we removed the surface ligands through an acid treatment. 31 The successful synthesis of ligand-free NaCeF 4 :Er/Yb NCs was verifed by TGA, FTIR spectra and zetapotential analyses (ESI Fig. S7-S9 †). More importantly, the ligand-free NCs preserved the intense NIR-II emission from the OA-capped NCs with essentially unchanged intensity. The z potential of ligand-free NCs in aqueous solution was measured to be 21.9 AE 0.9 mV (ESI Fig. S9 †) due to the existence of positively charged Ln 3+ ions (i.e., Er 3+ , Yb 3+ and Ce 3+ ) on the surface of ligand-free NCs, which endows these NCs with excellent dispersibility in aqueous solutions. Since Ce 3+ ions in the host matrix were exposed on the surface of ligand-free NCs after the acid treatment, H 2 O 2 can directly oxidize Ce 3+ to Ce 4+ through redox reaction, 32 resulting in the quenching of NIR-II emission of Er 3+ upon 980 nm excitation. Benefting from such a redox reaction, NaCeF 4 :Er/Yb NCs can be explored as an effective bioprobe for the detection of H 2 O 2 or H 2 O 2 -generated biomolecules (Fig. 3a). In order to investigate the quenching effect of H 2 O 2 on the NIR-II emission of NaCeF 4 :Er/Yb NCs, the spectral response of ligand-free NaCeF 4 :Er/Yb NCs with a size of 25.2 AE 2.7 nm (0.5 mg mL 1 ) upon addition of different amounts of H 2 O 2 (0-10 mM) was measured upon 980 nm excitation (Fig. 3b). The integrated DSL intensity of NaCeF 4 :Er/Yb decreased gradually with increasing concentration of H 2 O 2 , due to the redox reaction between the H 2 O 2 and Ce 3+ ions. As a result, the concentration of H 2 O 2 can be quantifed by the NIR-II emission intensity of NaCeF 4 :Er/Yb NCs (Fig. 3c). In the control experiment, by utilizing NaYF 4 :Er/ Yb or NaYF 4 :Er/Yb/Ce (with a Ce 3+ content of 10 mol%) as the probe, a negligible photoluminescence (PL) quenching effect of Er 3+ was observed upon addition of different concentrations of H 2 O 2 (ESI Fig. S10 †). The LOD, defned as the concentration that corresponds to 3 times the standard deviation above the signal measured in the blank, was determined to be 41.8 nM based on NaCeF 4 :Er/Yb nanoprobes. The highly sensitive response of H 2 O 2 allows for the detection of biomarkers such as UA which can yield H 2 O 2 through the UA/uricase reaction (Fig. 4a). The level of UA, which is the end product of purine metabolism in the human body in human blood and urine, can be treated as an indicator for certain clinical criteria. Abnormal levels of UA may cause diseases like gout, arthritis, renal disorder, Lesch-Nyhan syndrome, etc. 33,34 Specifcally, excess UA in human blood is a risk factor in cardiovascular related diseases, while reduced UA levels (hypouricemia) have been found to be closely related to several diseases such as diabetes mellitus and AIDS. 35 Therefore, the accurate detection of UA is of great importance in physiological survey and clinical diagnosis. In the assay system, UA or uricase alone was not able to quench the NIR-II emission of NaCeF 4 :Er/Yb nanoprobes upon 980 nm excitation, since no H 2 O 2 was generated (Fig. 4b). However, a notable quenching in Er 3+ emission was observed with the addition of both UA and uricase in NaCeF 4 :Er/Yb solution. Meanwhile, it was found that a time of 3 h was needed to reach equilibrium for the NIR-II emission of Er 3+ (ESI Fig. S11 †). Under the optimized conditions (0.5 mg mL 1 NaCeF 4 :Er/Yb and 0.011 U mL 1 uricase), the integrated NIR-II emission intensity of Er 3+ decreased gradually with UA concentration from 0 to 900 mM (Fig. 4c), due to the gradual release of H 2 O 2 . The calibration curve for the UA concentration exhibits a linear dependence in the range of 0.411-900 mM. The LOD of UA assay was determined to be 25.6 nM, which is much lower than the UA level in the serum of healthy human beings (130-460 mM). 35 In order to verify the specifcity of the bioassay, we performed control experiments by replacing UA with other possible interfering biomolecules and electrolytes that may exist in serum samples, such as metal ions, proteins, and amino acids, under otherwise identical conditions. As displayed in Fig. 5a, the quenching of NIR-II emission of Er 3+ in the control groups was negligibly small, which is in marked contrast to the signifcant quenching effect caused by the addition of UA. Such an exclusive PL quenching in the experiment group confrms the high specifcity of the assay, thus validating the applicability of NaCeF 4 :Er/Yb nanoprobes for UA detection in complex biological matrices such as serum. For the detection of UA in human serum samples, the NIR-II signal of the serum-based detection system exhibited a linear dependence on the UA concentration ranging from 1.234 to 900 mM (ESI Fig. S12 †). To show the reliability of direct quantitation of UA in complex biological fluids by applying the NaCeF 4 :Er/Yb nanoprobes, we carried out in vitro detection of UA in 24 serum samples. The UA concentrations determined by NaCeF 4 :Er/Yb nanoprobes were compared with those detected based on a commercial kit. As shown in Fig. 5b and Table S1, † the UA levels determined from the NaCeF 4 :Er/Yb based assay are highly consistent with those from the commercial assay kit. The correlation coefficient between both kinds of assays was determined to be 0.98, demonstrating that the NC-based assay is as reliable as that using the commercial kit. Moreover, we determined the recovery of three human serum samples upon addition of UA standard solutions with different concentrations. The analytical recoveries are in the range of 93.4-108.8% (Table 1). Both the coefficients of variation (CV) and recovery are within the acceptance criteria (CVs # 15%, and recoveries in the range of 90-110%) set for bioanalytical method validation. 36 These results clearly prove that the NaCeF 4 :Er/Yb nanoprobe has high reliability and practicability for UA detection in complex biological samples. Therefore, the proposed NaCeF 4 :Er/Yb nanoprobe, exhibiting background-free NIR-II emission under NIR excitation, is highly desired as a homogeneous bioassay nanoplatform for accurate detection of UA and other H 2 O 2 -generated biomarkers in clinical bioassays. Compared to previously reported UA bioassay systems, the homogeneous assay carried out employing the NaCeF 4 :Er/Yb nanoprobe is much more convenient and cost-effective, given that the assay can be performed based on a simple mixing of the test samples with uricase and the ligand-free NaCeF 4 :Er/Yb nanoprobe, and no complicated operations are involved in either nanoprobe preparation or surface modifcation. Another important application of NIR-II emission is the deeptissue bioimaging. To make the as-prepared hydrophobic NCs biocompatible, we coated the surface of OA-capped NaCeF 4 :Er/ Yb@NaCeF 4 NCs with amphiphilic 1,2-distearoyl-sn-glycero-3phosphoethanolamine-N-[carboxy-(polyethyleneglycol)-2000] (DSPE-PEG2000-COOH) phospholipids (Lipo). 37 The resultant Lipo-modifed NaCeF 4 :Er/Yb@NaCeF 4 NCs with a hydrodynamic diameter of 22.3 AE 1.1 nm were monodisperse in water (ESI Fig. S9 †). As a proof-of-concept experiment to examine the tissue penetration ability of NIR-II emission, we covered Lipo-modifed NaCeF 4 :Er/Yb@NaCeF 4 and NaYF 4 :Er/Yb NCs with pork muscle tissue of various thicknesses, which were imaged using a modifed Maestro imaging system. As shown in Fig. 6a and b, the NIR-II luminescence of NaCeF 4 :Er/Yb@NaCeF 4 NCs was detectable even at a depth of 10 mm upon excitation at 980 nm. By contrast, the green UC luminescence of NaYF 4 :Er/Yb NCs can only be observed at 4 mm beneath the tissue surface under otherwise identical conditions. The penetration depths that correspond to 50% of the original signal of NIR-II and green luminescence were determined to be $7 and $3 mm, respectively. The higher depth of penetration of NIR-II emission is due to reduced tissue scattering of light within the NIR-II window compared with that in the visible range. Furthermore, to demonstrate their great capability for noninvasive imaging, in vivo bioimaging experiments were carried out based on the Lipo-modifed NaCeF 4 :Er/Yb@NaCeF 4 and NaYF 4 :Er/Yb nanoprobes via tail vein injection into mice with the same dosage (0.1 mg mL 1 , 1 mL). After 30 min of blood circulation, images were taken upon excitation at 980 nm with appropriately equipped flters. Fig. 6c shows the evolution of the PL signal over 24 h arising from the injection of NaCeF 4 :Er/Yb@NaCeF 4 nanoprobes. 0.5 h after injection, the NCs accumulated essentially in the hindlimbs, liver, spleen, and lungs, as can be monitored by the bright NIR-II signals of Er 3+ . Particularly, images of the mouse blood vessels of organs and hindlimbs can be clearly observed, which reveals the excellent spatial resolution of NIR-II bioimaging. After longer time periods of blood circulation, PL fading from hindlimbs was observed. 24 h later, all the NaCeF 4 :Er/Yb@NaCeF 4 nanoprobes accumulated in the liver. Note that no tissue autofluorescence signal and light scattering were detected ## Experimental Detailed experimental procedures are reported in the ESI. † ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Intense near-infrared-II luminescence from NaCeF₄:Er/Yb nanoprobes for <i>in vitro</i> bioassay and <i>in vivo</i> bioimaging", "journal": "Royal Society of Chemistry (RSC)"}

learning_to_use_the_force:_fitting_repulsive_potentials_in_density-functional_tight-binding_with_gau

## Abstract: The Density-Functional Tight Binding (DFTB) method is a popular semiempirical approximation to Density Functional Theory (DFT). In many cases, DFTB can provide comparable accuracy to DFT at a fraction of the cost, enabling simulations on lengthand time-scales that are unfeasible with first principles DFT. At the same time (and in contrast to empirical interatomic potentials and force-fields), DFTB still offers direct access to electronic properties such as the band-structure. These advantages come at the cost of introducing empirical parameters to the method, leading to a reduced transferability compared to true first-principle approaches. Consequently, it would be very useful if the parameter-sets could be routinely adjusted for a given project.While fairly robust and transferable parameterization workflows exist for the electronic structure part of DFTB, the so-called repulsive potential V rep poses a major challenge.In this paper we propose a machine-learning (ML) approach to fitting V rep , using 1 Gaussian Process Regression (GPR). The use of GPR circumvents the need for nonlinear or global parameter optimization, while at the same time offering arbitrary flexibility in terms of the functional form. We also show that the proposed method can be applied to multiple elements at once, by fitting repulsive potentials for organic molecules containing carbon, hydrogen and oxygen. Overall, the new approach removes focus from the choice of functional form and parameterization procedure, in favour of a data-driven philosophy. ## Introduction With ever-improving electronic structure codes and increasing computing power, the size and time scales accessible to computer simulations in chemistry, physics and materials science have grown tremendously. 1 Nevertheless, while semi-local Density Functional Theory (DFT) 2,3 is generally regarded as a convenient compromise between speed and accuracy, a full first-principles treatment remains prohibitive for many problems of interest today. Therefore, the application of lower levels of theory in, say, "long", or "large" simulations, as well as in hierarchical approaches, is still commonplace. One of the most important recent developments in this respect is the rise of highly accurate machine-learning (ML) potentials, most prominently the Neural Network Potentials (NNPs) of Behler and coworkers and the Gaussian Approximation Potentials (GAPs) of Csányi et al. 4,5 Here, a detailed representation of the chemical environment (within a spatial cutoff) around each atom is used to construct the potential. 6 ML potentials are highly successful in describing covalent or metallic systems such as silicon, amorphous carbon or iron. 4, However, their local nature precludes the description of long-range electrostatic interactions, important in ionic and polar systems. In these cases a separate model for longrange electrostatic interactions is required. Even in unpolar systems like graphene, long-range resonance effects can invalidate the local approximation at the heart of ML potentials. 7 Interestingly, such fundamentally non-local quantum mechanical effects and long-range electrostatics are well described even with fairly simple semiempirical electronic structure models. Consequently, these approaches occupy an important niche between empirical potentials and first-principles electronic structure theory. A wide range of semiempirical quantum chemistry methods and tight-binding based approximations are available to this end. Among these flavors, the Density-Functional Tight-Binding (DFTB) method is one of the most widely used in molecular and materials modelling, particularly in its modern self-consistent implementations. 13,24 The latter cover a broad range of bonding situations (including charge transfer in multi-component systems) at low computational cost. This renders it particularly appealing for electronic structure calculations of large systems, long time-scale dynamical simulations, extensive structure searches, or as a lower-level pre-screening technique in hierarchical high-throughput approaches. DFTB is an approximation to DFT. While it can, in principle, achieve DFT accuracy at a fraction of the cost, this is highly dependent on the applied parameters (and on the system of interest). In particular, a parameterization that is general (in terms of high coverage of the periodic table and different chemistries) should not be expected to achieve high accuracy for every individual case. This is not to say that such a method is not useful, e.g. in the context of a hierarchical high-throughput study. However, when applying DFTB in predictive simulations, a more specialized parameter-set should be used. The main goal of the present work lies in simplifying the generation of such specialized parametrizations via ML, focusing on the so-called repulsive potential within DFTB. ## DFTB in a very small nutshell The derivation of DFTB starts from the Harris-Foulkes expression for the total energy: where n is the electron density, f i are orbital occupation numbers, V H and V xc are the Hartree and exchange-correlation potentials, E xc [n] is the exchange-correlation functional and E NN is the nucleus-nucleus repulsion energy. The first term in this expression corresponds to a sum of one-particle energies and is sometimes referred to as the band-structure energy E BS . The second and third terms can be considered double counting corrections to E BS . This Harris-Foulkes energy is identical to the more commonly used Kohn-Sham (KS) formula if n is the ground state density (n GS ) of the functional. However, the formalisms differ if some approximate reference density n 0 ≈ n GS is used. Specifically, the error of eq. 1 with respect to the exact ground-state energy is O((n 0 − n GS ) 2 ). This motivates the so-called Harris approximation of using the overlap of non-interacting fragment densities in combination with eq. 1. Foulkes and Haydock argued that traditional tight-binding formalisms can be understood as approximations to eq. 1, where all terms except E BS are grouped into a repulsive potential V rep . 26 They also showed that V rep is approximately pairwise and short-ranged. Building on this idea, Seifert, Frauenheim, Elstner and coworkers developed the DFTB framework, by constructing n 0 from a superposition of isolated atom densities subject to a confinement potential. 28,29 This allows the non-empirical determination of all matrix elements required to compute E BS in the tight-binding approximation, whereas V rep remains an empirical expression. While this initial DFTB scheme (later termed DFTB0) proved quite successful for (mainly) covalent systems, it quickly became clear that the underlying approximations break down for (partially) ionic or polar systems, where constructing n 0 from isolated neutral atoms is no longer appropriate. This limitation was overcome by the introduction of the self-consistent charge (SCC) scheme, which also includes the effects of density fluctuations (δn) to second or third order (DFTB2/3). 13,30,31 The total energy in DFTB2/3 is then given by: where E coul [δn] is computed as the screened Coulomb interaction between atomic partial charges and a Hubbard-like on-site term for the energy of the partially charged atom. 27 E rep [n 0 ] is the repulsive energy, which we will discuss in more detail. ## The role of the repulsive potential In principle, E rep [n 0 ] can be derived from eq. 1, by plugging in the reference density n 0 and removing the first term: where the second term is the Hartree energy of n 0 . However, in DFTB it is not practical (nor particularly accurate) to use this formulation, as the Hartree and exchange-correlation potentials are never explicitly calculated. Instead, E rep is empirically approximated by short-ranged, atom pairwise contributions, leading to the simple model: Here, we sum over atom pairs and R IJ = |R I − R J | is the distance between the pair. The short-range pairwise potential V rep thus on;y depends on the interatomic distance and must be defined for each possible combination of elements. Usually, its short-range character is enforced via a cutoff radius, so that In the ideal case, the DFTB method should recover the suitably normalized DFT energy. Since the parameterization of the electronic terms in eq. 2 is largely non-empirical, it is common practice to first determine these parameters, and then fit V rep to minimize the difference between DFTB and DFT energies, forces and geometries for a set of reference (or training) systems. V rep thus assumes a role similar to the exchange-correlation functional in DFT, where all the complicated physics is collected into an unknown term that must be approximated. Consequently, there is no simple, physically motivated functional form that can be used for V rep . Indeed, the name "repulsive potential" is something of a misnomer. Though the potential is repulsive in the short range limit (mimicking core-core and Pauli repulsion), V rep can also display attractive parts. 31 To reflect all these effects adequately, a flexible ansatz for its functional form (such as a high order polynomial or set of splines) is key, which entails more fitting parameters. The pairwise nature of V rep causes further difficulties when fitting multicomponent systems, since the number of parameters to be fitted scales with the number of element-pairs. Overall, fitting V rep to DFT data is therefore the most time consuming step in the parametrization workflow. As mentioned above, the approximations in DFTB preclude a completely universal model with DFT accuracy. A carbon-carbon V rep fitted to diamond will significantly differ from one fitted to organic molecules. Commonly used parameter-sets are therefore often restricted to specific species and bonding situations, e.g. pure solid-state 32 or biological systems. 33 In our view, this is an unavoidable characteristic of semiempirical methods. Consequently, it is all the more important to make the parametrization of V rep as painless as possible. ## Analytical Forms of V rep As mentioned, E BS and E coul depend on non-empirical parameters, except the confinement potential. The choice of a functional form for the latter is only limited by few physically motivated requirements-it has to be zero or nearly zero until some onset radius, then increase steeply over a certain distance. Further, one may place loose assumptions on the parameters that define the onset and the width: as a rule of thumb, the pseudo-atom wavefunction should vanish completely somewhere between the covalent radius of the species and twice its value. Similarly, the repulsive potential should also vanish between the first and second nearest neighbor. However, unlike the confinement potential, one cannot make any straightforward assumption about its details, other than justifying a general, analytical pairwise functional form, as discussed above. As a result, several functional forms have been proposed over the years, and correspondingly many parametrization approaches, ranging from "fully manual" to "fully automatic", with all the nuances in between. 27,34,35 The choice of the functional form, which has to strike a balance between flexibility and facility to optimize, is to a large extent what makes current DFTB parametrization an art. An example for a "fully manual" approach with a predefined functional form is described in Ref. 27 By calculating the difference between DFT and purely electronic DFTB forces for (stretched) dimers or symmetric molecules, V rep can directly be computed by projection onto different bond types. This procedure is conceptually simple but does not lead to generally accurate parametrizations, as the type of reference structures that can be used is highly limited in this case. The first attempts to automatize the parametrization for arbitrary reference structures appeared relatively early in the history of DFTB. 34,35 This requires sufficiently flexible ana-lytical forms (typically a combination of a short-range exponential with a damped polynomial in the bonding region), with a large number of parameters. Consequently, the optimization of V rep becomes a cumbersome non-linear optimization problem in a high-dimensional space. This is particularly problematic if multiple elements are fitted simultaneously. In the state-of-the-art parameter set for organic molecules 3ob, 30 the repulsive potential is represented by a fourth-order spline, with a cutoff at which the function and its first three derivatives vanish. The cutoff is chosen to lie between first and second neighbor distances of typical organic molecules. The remaining degrees of freedom are fitted to experimental or ab initio values of atomization energies, geometries, reaction energies and vibrational frequencies, suitably expressed in terms of a linear system of equations with the spline coefficients as unknowns. Here, the overall quality of the potentials is sensitive to the choice of the division points for the spline, which thus represent the most delicate aspect of the procedure. In the context of their titanic effort at parametrizing the entire periodic table, 36 Heine and coworkers proposed an alternative formulation of the repulsive interaction. 37 Instead of expressing it (and therefore parametrizing it) as a sum of two-body potentials, they calculate it explicitly from the DFTB Kohn-Sham-like equations, using the atomic electron densities subject to a confinement. The latter is initially chosen as the optimized confinement for the electronic part 36 and subsequently adjusted to minimize the root mean square distance (RMSD) of atomic forces with respect to DFT references. Essentially, in this approach the repulsion is completely defined from one-center contributions, thus the parametrization requires only four parameters per element (two "electronic" confinement parameters and two "repulsive" confinement parameters). This approach is very general, but the accuracy of the potentials is necessarily limited by the constrained functional form. More recently, a general, property-oriented approach based on Particle Swarm Optimization 38 (PSO) proved promising in parameterizing sets of multiple atomic species at once. Here, the repulsive potential is assumed to obey a predefined analytical form, whose coeffi-cients are optimized directly in the PSO to reproduce a set of properties. Similarly, in the work of Krishnapriyan et al., 39 the electronic and repulsive parts are optimized together, rather than sequentially, through the global optimization of a single objective function, combining simulated annealing and steepest descent. In their approach, the radial dependence of the matrix elements is also represented by simple analytic functions, rather than in the conventional tabular form. Finally, Elstner and Lilienfeld 40 recently introduced so-called generalized repulsive potentials, which not only depend on the atom pair (I,J), but also on the bond type b(I, J). This allows to correct any existing parametrized repulsive potential for environment-specific effects. They then propose unsupervised ML to identify bond types, discussing different clustering algorithms and proposing the mean-shift algorithm as a viable choice. The repulsive potentials for all new bond types are then fitted as in conventional approaches. To conclude this short overview, it has to be noted that the problem of parametrizing a repulsive potential is not strictly specific to DFTB, but appears in other tight-binding formulations as well, with the same associated challenges. For instance, Grimme's xTB 20,41 is related to the SCC-DFTB scheme in its formulation, but almost completely avoids pair-specific parameters. The chosen form of the repulsive potential is similar to the pairwise potential employed in the QMDFF force field. 42 The latter only depends on the effective nuclear charges of the two atoms two further (one per atom) element-specific coefficients, as well as a global parameter. Therefore, the parametrization only requires the optimization of two parameters per atom, completely avoiding pair-specific parameters with the exception of few atom pairs for which the global parameter was finely tuned to improve accuracy. Pettifor and coworkers have proposed a number of orthogonal tight-binding (OTB) models for metals 23 and binary bulk materials. 44,45 In these formulations the repulsive energy is expressed as an embedding functional of pairwise potentials between atom pairs, fitted to reproduce cohesive energy vs. volume curves of selected structures of elemental and binary systems. The same expression for the repulsive energy is employed in the tight-binding model developed by Tang et al. for carbon. 46 The latter goes beyond the two-center approximation, allowing both the hopping parameters and the repulsion to be environment-dependent. ## V rep from Gaussian Process Regression The goal of this paper is to formulate V rep within the Bayesian machine learning framework of Gaussian Process Regression (GPR). Here, we closely follow the GAP approach of Bartók and Csányi. 5 GAP combines a local decomposition of the total energy (i.e. into n-body contributions) with the sparse GPR method. We refer to the original papers for details, but briefly summarize the method below. The GAP framework has widely been applied to fitting interatomic potentials directly, but to the best of our knowledge, this is its first application within DFTB. In GPR, V rep can be modeled as a linear combination of covariance (kernel) functions: where the sum is over all N pairs pairs in the set of reference structures {X}, with regression coefficients α IJ and the Kernel function k(R, R IJ ). We are looking to fit V rep to minimize the difference between a "repulsion-less" DFTB model and appropriately normalized DFT reference energies or forces, so that ideally: The main advantage of using GPR (relative to "conventional" potential fitting) is that there is a closed-form linear algebra expression for the coefficients α, which minimize the loss function: where t IJ are reference (target) values for the repulsive potential and σ n is a regularization parameter. The role of σ n is to provide a measure of uncertainty in the reference data. In experimental settings, it is usually related to measurement noise. While the DFT reference data used herein is obviously noise free, we cannot expect that the atom-pairwise ansatz for V rep is arbitrarily accurate. σ n then represents a measure for the overall accuracy of the approximate DFTB model. σ n also provides a means to avoid overfitting, as illustrated below. The coefficients are calculated as: where the covariance matrix K has the elements k(R IJ , R KL ) for all atom pairs I, J and K, L in {X} and y is the vector of target values t IJ . This procedure is interchangeably referred to as training or fitting. Unfortunately, it is not completely straightforward to apply GPR to V rep , since we can only compute reference values for the full E rep (via eq. 6), but not for the individual components V rep (R IJ ). An additional complication is that the matrix to be inverted in eq. 8 has the dimension N pairs × N pairs . Clearly, the number of atom pairs in a training set can become quite large, so that the fitting procedure ultimately becomes a computational bottleneck for very large training sets. At the same time, there is likely a lot of redundancy in such a dataset. In Fig. 1, a histogram of carbon-carbon distances from a set of 6800 structures of organic molecules is shown. Around the nearest neighbor distance, nearly identical pair distances appear hundreds of times. The GAP method offers an elegant solution to these issues, by using a sparse formulation of GPR and by defining target values as linear combinations of local quantities. In particular, the total repulsive energy E rep (Eq. 6) is expressed as a linear combination of (unknown) pairwise potentials V rep (R IJ ). The data redundancy issue, is overcome by defining a set of With the GAP machinery in place, it remains to define the kernel function k(R, R ). Simply put, a kernel measures the similarity between two inputs. A common choice is the "squared exponential" (SE) kernel: where θ is a length-scale parameter. If R = R , k(R, R ) = 1 and, in general, k(R, R ) ≤ 1. The length-scale θ is essentially a measure for how "generous" the kernel is when comparing two inputs. For a small value of θ, k(R, R ) ≈ 0 unless |R−R | ≈ 0. Conversely, a large value of θ means that even fairly distant inputs are still considered somewhat similar. Additionally, the kernel functions also act as the basis functions for the expansion of V rep (R). Choosing a larger value of θ therefore leads to a smoother potential. To ensure that V rep (R) smoothly decays to zero at R cut , we multiply the SE kernel with a damping function, so that with Here, d is a transition length in which the kernel is damped to zero. The parameter β enforces a smoothly decaying function. An important feature of the GAP model is that forces can be computed analytically in terms of derivatives of k damp (R, R ). This also allows using forces directly in the training. Fitting to forces is highly advantageous, as they contain much more information about the shape of the potential than energies. In the following we exclusively train on forces, but in principle combined training on forces and energies would also be possible. A final issue to note is that ML models cannot be used to extrapolate beyond the training data. At the same time, V rep (R) should be defined on all points between zero and the cutoff radius. Most importantly, spurious minima at short interatomic distances have to be avoided. At first glance, this is a conundrum, as common training sets will not contain very short bond distances (see Fig. 1). We resolve this by defining the GPR potential only in the region where data is available. For the short range part of the potential, we then fit an exponential function to the learned potential in the shortest 0.1 of the interatomic distances covered. With this, all components of the GPR approach to V rep (R) (termed GPrep in the following) are in principle defined. One of the most attractive features of GPR is the fact that the coefficients α can be determined in a closed form expression. For this reason GPR is considered a "non-parameteric" ML approach. Nonetheless, a fairly large number of "hyperparameters" must still be defined. Specifically, these are the number of sparse points N s , the kernel lengthscale θ, the regularization parameter σ n , the cutoff radius R cut and the parameters of the damping function (d and β). As shown below, for the presently studied systems the performance of GPrep potentials is largely independent of the specific choice of hyperparameters (within certain ranges). Consequently, we have not found it necessary to put much effort into the optimization of the hyperparameters. Indeed, this would defeat the purpose of the GPR approach, in our view. There is little benefit in replacing a non-linear parameter optimization problem with a similarly complex hyper parameter optimization problem. To avoid this issue, we recall that GPR is a Bayesian framework. In this context, the hyperparameters and the choice of the kernel represent our prior knowledge about the problem. The damping function can be considered a zeroth-order approximation to V rep (R), which is refined as the model is exposed to data (see Fig. 2). It is sufficient to choose the hyperparameters in such a way that the prior is a reasonable qualitative model for V rep (R). The training will then improve upon the prior, wherever it is a poor quantitative description of V rep (R). A wide range of different priors can ultimately be trained to yield potentials of similar quality. This allows us to drastically reduce the number of hyperparameters that need optimization. In practice, we fix the values of d and β to 1 and 3 −1 , respectively, throughout this work. We also found it sufficient to use 20 uniformly spaced sparse points (N s ) for each potential. Most of the remaining hyperparameters are set globally, although it would in principle be possible to choose, e.g., a separate value of θ for each pair potential. The exception to this rule are the cutoff radii R cut . Here, it clearly makes sense to define a different cutoff for, e.g., C-H and C-O potentials. Fortunately, histograms like the one shown in Fig. 1 can be used to determine the range of nearest neighbor interactions for a given element pair. We then set R cut to the onset of the second nearest neighbor distribution (see Table 1). It thus remains to optimize two global parameters, namely θ and σ n . Here, optimal values depend on the specific application and the size of the training set. Below, this will be discussed in detail for a set of organic molecules. ## Methods The GPrep method has been implemented in a Python package that is available at gitlab.com/jmargraf/gprep. All DFTB calculations were performed with the DFTB+ package. 47 DFT reference data were computed with the FHI-aims package 48 using the PBE functional, 49 tight integration settings and a tier2 basis set of numerical atomic orbitals. The atomic simulation environment (ASE) was used to process calculation results. 50 The DFT reference dataset used below consists of 68 organic molecules taken from the reaction network repository. 51 These molecules contain up to four carbon and/or oxygen atoms (and an arbitrary number of hydrogens), covering all common hybridizations and bond orders in C-C, O-O, C-O, C-H, and O-H binding. To sample diverse conformations and nonequilibrium structures, classical molecular dynamics (MD) simulations were performed with the LAMMPS code using the Dreiding forcefield and Gasteiger charge model. 52,53 Specifically, MD simulations were run in the NVE ensemble for 100 ps (after 10 ps equilibration), with average temperatures of ca. 500 K. From these trajectories, DFT forces were calculated for structures extracted every ps, leading to an overall dataset size of 6800 structures. This dataset is also available at gitlab.com/jmargraf/gprep. ## Results To test the performance of the GPrep method, we reparameterize V rep (R) for a state-of-the-art DFTB variant, namely the 3ob method of Elstner and co-workers. 31 3ob uses the third-order expansion of the DFTB equations and significantly outperforms the earlier second-order mio method for geometries and thermochemistry of organic and biological systems. 13 It was also found to be quite accurate for non-covalent interactions, when combined with dispersion corrections. 54,55 As most DFTB methods, 3ob uses a parameterization strategy that separates the optimization of the electronic parameters and the repulsive potential. The optimization of V rep (R) is the last step in this procedure. V rep (R) then acts as a catch-all term that is supposed to minimize the deviation of the final DFTB results from some reference method. Unsurprisingly, it is not possible to universally compensate all the approximations made in the DFTB formalism with a single set of pairwise potentials. Consequently, the 3ob developers actually provide two sets of parameters for V rep (R), one optimized for "general use" and one for frequency studies (3ob-f). This again underscores the advantage of being able to adjust V rep (R) in a simple way. To be clear, we do not aim to replace the original 3ob parameters with this work. Instead our goal is to show that the GPrep approach allows the simple definition of an optimized V rep (R), given a set of reference structures and suitable electronic parameters. Instead of defining a single new DFTB method, we propagate the modification of the V rep (R) for specific cases, and provide the method to do this. The application of the GPrep approach hinges on two questions: how should the hyperparameters be determined, and how much training data is required? To answer the first question, an exhaustive grid search for σ n and θ was perfomed on a subset of 1000 structures, randomly drawn from the reference data set. Fig. 3 shows a contour plot of the average RMSE on the reference forces from four-fold cross-validation. It can be seen that there is a large plateau of hyperparameter combinations that perform equally well. Importantly, potentials with similar performance can have very different shapes, however. In Fig. 4, several GPrep potentials for the C-O interaction are compared. Using a "nearest neighbor" cutoff (R cut = 2.0 ), the potential can be either fully repulsive (θ = 1 , σ n = 0.05 eV) or feature an attractive dip (θ = 0.2 , σ n = 0.0001 eV). Yet, both potentials perform equally well in cross-validation. This behaviour is frequently observed when training on forces, since the force on an atom depends on the interplay between all involved potentials. An atom with multiple neighbors can experience the same force whether the interactions are attractive or repulsive, as long as the potentials cancel each other out in the same way. In the same figure, we also illustrate the effect of changing the cutoff to include the next nearest neighbor (θ = 0.2 , σ n = 0.0001 eV, R cut = 2.5 ). Again, a well performing potential with a completely distinct shape is obtained. In this case, the function roughly resembles one of the "nearest neighbor" potentials, but is more attractive in the longer range (for the second nearest neighbor) and more repulsive in the short range. The similar performance is in this case caused by cancellation of these contributions. The relative insensitivity of the validation error to the hyperparameters is good news, because it confirms that exhaustive optimization of the hyperparameters is not necessary. However, Fig. 4 should also serve as a warning, since potentials that rely on cancellation of errors will extrapolate less well. In the worst case, oscillations or sharp features can occur, which would lead to significant problems, e.g., in MD simulations. All things being equal, one should therefore prefer a model with large values of θ and σ n , which leads to smoother potentials and less overfitting. We also strongly recommend plotting the fitted potentials, instead of just relying on a metric like the validation RMSE. In the following examples we use a value of θ = 1 (in line with the expected length scale of interatomic forces). The question of how large the training set should be can be addressed via so-called learning curves (see Fig. 5). Here, the performance of models For reference, we also include the validation error of the original 3ob parameters in Fig. 5 (ca 0.8 eV/ ). We stress again, that this is not to suggest that the GPrep parameters will be better in general applications of 3ob, not least because 3ob was trained to a different reference method. However, this does show that we can obtain a significantly improved method for a specific application, which is the goal of this paper. In this context, it is also interesting to compare the GPrep potentials with the original 3ob ones (Fig. 6). As before, this shows that two models which ostensibly describe the same underlying physics can differ quite strongly. This is for example evident for the C-H potential, which has a strong attractive feature in 3ob, while GPrep is completely repulsive. Meanwhile, the C-O potentials are practically identical. These differences should also be understood in light of the reference data used. If V rep (R) is (partially) fitted to thermochemical data, attractive features can compensate systematic underbinding of certain bond types. Meanwhile, the GPrep parameters have only been trained to forces and are therefore more consistently repulsive. ## Conclusions In this paper, we introduced a new ML-based approach to fitting repulsive potentials in DFTB. This method (termed GPrep) is based on the GAP approach of Bartók and Csányi, adapted to the special circumstances of DFTB. A reference implementation is provided at gitlab.com/jmargraf/gprep. We apply the GPrep approach to fit new repulsive potentials for the 3ob method, using a training set of 6800 non-equilibrium geometries of 68 molecules. The GPrep potentials display significantly improved accuracy for these structures. We also explore the role of the hyperparameters θ and σ n , to provide more general guidelines on how to apply the approach. Importantly, training the potential for a given set of hyperparameters only takes seconds. Overall, GPrep allows improving existing DFTB methods for specific applications in a simple manner. In our view, the approximations made in semi-empirical methods make universally accurate parameter sets impossible. This issue can be circumvented by reparametrizing DFTB "on-demand" with GPrep. This constitutes a more data-driven approach to DFTB, where the reference data and hyperparameters define the specific method, instead of a parameter file. Finally, our results also show that even such customized parametrizations cannot be arbitrarily accurate. In this context, it is interesting to note that the GAP framework could also be used to fit many-body repulsive potentials. This would be a departure from the traditional DFTB framework, but is certainly worth exploring. Indeed, the "generalized pair potential" approach of Lilienfeld and Elstner is a step in that direction. 40 ## Acknowledgements CP gratefully acknowledges funding from the German Research Foundation (DFG -Deutsche Forschungsgemeinschaft) through grant # PA 2932/1-1. We further acknowledge support by Deutsche Forschungsgemein-schaft (DFG) through TUM International Graduate School of Science and Engineering (IGSSE), GSC 81.

chemsum

{"title": "Learning to Use the Force: Fitting Repulsive Potentials in Density-Functional Tight-Binding with Gaussian Process Regression", "journal": "ChemRxiv"}

organic-inorganic-induced_polymer_intercalation_into_layered_composites_for_aqueous_zinc-ion_battery

## Abstract: Rechargeable aqueous zinc-based batteries are very attractive alternative devices for current energy storage by virtue of their low cost and high security. However, the performance of vanadium oxide cathode strongly relies on the distance of interlayer spacing. Here, we employ layered PEDOT-NH 4 V 3 O 8 (PEDOT-NVO) as a cathode material, which produces an enlarged interlayer spacing of 10.8 A ˚(against 7.8 A ˚for the single NVO) by effectively conducting polymer intercalation. This cathode material exhibits an improved capacity of 356.8 mAh g À1 at 0.05 A g À1 and 163.6 mAh g À1 , even at the highest current density of 10 A g À1 (with a high retention from 0.05 to 10 A g À1 ), and features an ultra-long lifetime of over 5,000 charge-discharge cycles with a capacity retention of 94.1%. A combination of mechanism analyses and theoretical calculations suggest that the oxygen vacancies and larger interlayer spacing through polymer assistance account for the improved electrochemical performance. ## INTRODUCTION Batteries, as energy storage systems, have been employed in electric vehicles, automotive propulsion, and stationary load leveling for electricity under the motivation of global trends. Since the utilization of lithium-ion batteries (LIBs) in 1991, they have quickly dominated the market of portable electronics, such as laptops and cellphones, by virtue of their high energy densities and long lifespan. 1,2 As they are used more widely in portable electronics, the growing safety concerns resulting from flammable organic electrolytes, the high cost of lithium sources, and the complexity of operating conditions have delayed feedback of their performance. 3,4 Aqueous electrochemical energy storage devices are endowed with several favorable characteristics over organic batteries. In addition to the flexible environments without the necessity of strictly controlling oxygen and water content, the higher ionic conductivity, greater safety, and environmentally friendly feature of water-based electrolytes are also attractive. Naturally, aqueous zinc-ion batteries (ZIBs) employing multivalent zinc anodes can not only, in particular, reach a high theoretical capacity (820 mAh g 1 ) and energy density (5,581 mAh cm 3 ) as a result of the multiple-electron transformation occurring in redox reactions but also achieve high stability and low redox potential (0.76 V versus standard hydrogen electrode [SHE]) in appropriate water-based electrolyte. Recently, various MnO 2 and Prussian blue analogs have been considered promising cathode materials with high theoretical capacities for aqueous zinc batteries. Unfortunately, Prussian blue analogs often provide limited capacities of less than ## The Bigger Picture For the high demand of largescale renewable energy storage system, zinc-ion batteries fulfill this requirement because of their abundance in Earth's total elemental reserves, higher water stability than that of alkaline (Li, Na, and K) metals, and high theoretical capacity. Here, we have developed an enlarged interlayer spacing of layered material that is capable of reversibly accommodating more Zn 2+ ion while achieving highcycle performance cathode materials for the application of zinc-ion batteries. The conducting polymer material has emerged as an important pathway for producing organic-inorganic intercalation compounds with larger interlayer spacing or oxygen vacancies, which are essential for fundamental studies as well as technique application. 120 mAh g 1 , and the full utilization of high-capacity MnO 2 cathode electrodes still faces great challenges mainly because of the hydrated H + /Zn 2+ -insertion-induced phase transformation and the subsequent structure collapse during cycling. Until now, the longest cycle life (5,000 cycles) of the MnO 2 cathode was achieved with a very low utilization of capacity (100 mAh g 1 that is 30% of the specific capacity). Vanadium-based composites have constantly been considered the host materials for traditional lithium and sodium batteries because of the multivalence state of vanadium and because different vanadium oxide frameworks arise from several coordination polyhedra. 19,20 Until 2016, Nazar's group developed a kind of layered Zn 0.25 V 2 O 5 cathode with a high capacity of 300 mAh g 1 and long life-cycling stability for 1,000 cycles in aqueous ZIBs. 21 Since then, many researchers have proposed various vanadium-based composites that serve as zinc-ion host cathodes, such as vanadium oxides containing NASICON (sodium super ionic conductor), tunnel, and layered structures. The succeeding mechanism investigation has demonstrated that the hydrated Zn 2+ prefers to be stored in the layered structure. Noticeably, it has been reported that pure V 2 O 5 cathodes present poor zinc-ion storage capacity and cycling stability, which can achieve a good rate capability through the introduction of structural water that acts as a ''lubricant'' to promote Zn 2+ diffusion. Also, metal-ion intercalation seems to be an interesting approach to improving the performance of V 2 O 5 as a cathode, such as Li + , Na + , K + , Ca 2+ , Mg 2 , and Zn 2+ ions. However, the electrochemical stability of host materials still suffers from structural degradation because the layers existing in vanadium oxides could be destroyed when guest metal species are located between layers as pillars. 24,32 In addition, the large molecular weight and volume of metal ions will lead to an unstable change upon repeated insertion and extraction of zinc ions. According to the theory that materials with extended interlayer space and reinforced layer structure could facilitate the Zn 2+ storage and improve the cycle life even at a high utilization of capacity, we wonder whether other neutral molecules or clusters will be more effective at increasing the interlayer distance. Based on the easy oxidation of neutral polymer backbone (p-doping), such as polyaniline, polypyrrole, and poly(3,4-ethylenedioxythiophene) (PEDOT) when intercalated into highly oxidizing layered oxides, the insertion of conductive polymers has been examined to change the interlayer spacing for these materials. For example, Huang et al. have inserted polyaniline, through an interface reaction, into MnO 2 nanolayers that gather together to form a mesoporous structure, which can avoid phase transformations and achieve a very stable cycling performance for ZIBs. 33 In fact, PEDOT, having been verified to display greatly enhanced stability compared to polypyrrole and polyaniline, has attracted extreme interest for application in supercapacitors in recent years. In light of such research, we designed the redox intercalative polymerization of 3,4-ethylenedioxythiophene into ammonium vanadate oxide (NVO) via a facile route. As a result, the interplanar spacing of the crystal lattice for NVO is enlarged by the successful intercalation of PEDOT between layers, giving rise to an improvement in the mobility of electrolyte cation species intercalating into the inner sites of the crystal lattice. More significantly, a comparison of the electrochemical performance between the polymer-free and intercalated NVO indicates that the hybrid of PEDOT-NVO-layered cathode promotes zinc-ion storage, resulting in a high capacity while retaining good cycling stability. ## RESULTS AND DISCUSSION The strategy for designing PEDOT-NVO is schematically depicted in Figure 1. The layered oxide was first produced by sonication treatment of V 2 O 5 and (NH 4 ) 2 S 2 O 8 (APS) solution. Soon afterward, the process of PEDOT intercalation into vanadate oxide was derived from a redox reaction. Specifically, the present EDOT monomers were oxidatively polymerized to PEDOT while the insertion of just-synthesized PEDOT into the NVO-layered structure occurred simultaneously. After PEDOT intercalation, the color of NVO changed from yellow brown to black green, as shown in Figure S1. As displayed in Figures 2A and 2B, powder X-ray diffraction (XRD) and Raman spectra were carried out to characterize the difference in the crystal structure and vibrational mode between the NVO and PEDOT-intercalated NVO. Both of the samples were found to present the monoclinic structure, which corresponds to standard data (JCPDS no. 41-0492, NH 4 V 3 O 8 $0.5H 2 O). However, the intensity of all diffraction peaks for PEDOT-NVO was lower than that of pure NVO, indicating that some feature of the NVO diffraction peaks in the composites was shielded by the polymer incorporation, which is similar to the previous polymerization of EDOT. 38 Another intriguing factor is that the strong peak at 2q = 11.3 of pure NVO corresponding to the (001) plane showed a negative shift to 2q = 8.3 of PEDOT-NVO composites. Determined from the Bragg's formula, the interplanar spacing was enlarged by intercalation of PEDOT from 7.8 to 10.8 A ˚. Further presented in Figure 2B, the low-frequency Raman signal at 143.9 cm 1 reflects the chain translation related to the layered structure. The mode centered at 993.8 cm 1 corresponds to the V=O stretching mode along the c axis, and the modes at 404.1 and 525.2 cm 1 correspond to the V-O-V stretching mode. The other peaks around 283.4 and 693.2 cm 1 were presumably produced as the signature of NVO with crystal water molecules. 39 A marked distinction of this phase is the presence of the peak at $1,444.3 cm 1 , typical for the symmetric stretching mode of the C a =C b bond in PEDOT, which is powerful evidence of successful intercalation into NVO. 40,41 Also, Fourier transform infrared (FTIR) spectra confirmed the presence of V=O stretching vibration, the V-O-V bending vibration, and edge-sharing V-O stretching vibration, which were located at $968.0, $823.2, and $727.3 cm 1 (see Figure S2). As observed in the two samples, the peak of 730.1 cm 1 is derived from a V-OH 2 stretching mode due to coordinated water, and the peak at $1,623.1 cm 1 is attributed to stretching and bending vibrations of water molecules. 42 In comparison, these vibrations are slightly shifted by the charge interaction polymerization of intercalated PEDOT to oxygen vacancy, which has been described in a later discussion. The discriminative characteristic feature at $1,116.7 cm 1 is observed for the PEDOT-NVO composite, which is ascribed to the C-S vibration, as well as a weak feature of p-doping polymer around $1,353.2 cm 1 . 41 In addition, thermogravimetric (TG) analysis was performed on oxygen atmosphere to investigate these two samples as shown in Figure S3. The weight loss occurring at the range of 0 C-300 C could be ascribed to the loss of ammonia ions and water molecules from the interlayer space. The third weight loss ($4.2%) between 300 C and 600 C was induced by the release of carbon and sulfur from PEDOT polymer. The obtained NVO and PEDOT-NVO composites were subjected to X-ray photoelectron spectroscopy (XPS) analysis, and all-element patterns of the results are shown in Figure 2C, which essentially displays the same features, except for an additional S 2p signal in PEDOT-NVO that can be attributed to PEDOT. Further processing into chemical oxidation state resulted in a comparison of nitrogen and vanadium in the two samples, as shown in Figures 2D and 2E. The N1s XPS spectrum of PEDOT-NVO (401.0 eV) shifts toward a lower binding energy in contrast to that of NVO (401.6 eV). This phenomenon might be ascribed to electron transfer between NVO and the positively charged moiety in PEDOT. The V 2p 3/2 peak of PEDOT-NVO shifts to low binding energy regions as duration extends, speculating a lower valence state of vanadium. In detail, the V 2p 3/2 and V 2p 1/2 peaks in PEDOT-NVO can be deconvoluted into two binding energies located at 517.1 and 515.4 eV (V 2p 3/2 ) and 524.4 and 522.8 eV (V 2p 1/2 ), which correspond to V 5+ and V 4+ , respectively. The mixed valence of V 4+ and V 5+ is produced because the NVO could have undergone reduction during oxidative polymerization of EDOT by using APS and it transformed from V 5+ into V 4+ . 43,44 The above analysis indicates that the polymerization and insertion of the EDOT monomer are accompanied by a sacrificial reduction of the V 3 O 8 layers and the formation of oxygen vacancy with polymer assistance in this hybrid material. On the other hand, the presence of oxygen vacancy here is confirmed by electron paramagnetic resonance (EPR) (Figure 2F). Compared with NVO, the corresponding spectra of PE-DOT-NVO have a sharp peak with an isotropic g value of 2.02 at room temperature, which matches well with the previously reposted V 4+ -containing materials. 45,46 The growth morphology of PEDOT-NVO and NVO was revealed by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). As shown in Figures 3A and 3B, both of them present the connected nanobelt structure, which is 50-200 nm wide and several micrometers long, thus demonstrating that the in situ redox intercalative polymerization is topotactic and retains the general morphology of the host material. The TEM images in Figure S4 confirm their flat and joined-together morphology, except for the residual bulk deposition of the polymer on the surface of the nanobelt for PEDOT-NVO. As observed in Figures 3C and 3D, high-resolution TEM (HRTEM) of the PEDOT-NVO samples shows a perfect interlayer distance with a lattice parameter of d = 10.8 A ˚, larger than the corresponding d-spacing (7.8 A ˚) of the pure NVO, which is well matched with the XRD result. In addition, the successful intercalation of PEDOT is further evidenced by elemental mapping analysis (Figures 3E and 3F), where V, O, C, N, and S are homogenously distributed through the whole nanobelt. As shown in Figure S5, energy-dispersive X-ray spectroscopy (EDS) revealed the atomic ratio of O to V to be 2.6, which is identical to the value of V 3 O 8 in this material. EDS analysis and scanning transmission electron microscopy (STEM)-EDS mappings further verified the uniform distribution of N, V, and O in NVO (see Figure S6). The content of N elements is lower than the pure NVO, which may be due to the polymer PEDOT occupying the former site of NH 4 + and/or the formation of a new vacant site. On the basis of these results, we can conclude that the distance of well-defined V 3 O 8 layers is markedly enlarged by the insertion of organic conducting polymer in this hybrid material. The electrochemical zinc-ion storage properties of PEDOT-NVO and NVO as cathode materials for ZIBs are evaluated in detail. Figure 4A presents the cyclic voltammetry (CV) curves of the two electrodes at a scan rate of 1 mV s 1 in a voltage range from 0.4 to 1.6 V. For the PEDOT-NVO cathode, four redox couples of peaks appeared in the cathodic scan (0.74, 0.98, 1.08, and 1.36 V) and the anodic scans (0.58, 0.78, 0.96, and 1.32 V), suggesting that the intercalation and deintercalation of zinc ions from PEDOT-NVO are achieved in multiple steps. Additionally, the major two reduction peaks located at 1.0 and 0.6 peaks could be ascribed to the possible products from PEDOT-NH 4 V 4A and 4B), which is the valence transformation of vanadium from V 5+ to V 4+ and then to V 3+ . By contrast, the NVO cathode only obtains two redox couples focused on 0.50/0.75 and 0.80/1.02 V, in agreement with many reported vanadium materials. Moreover, note that a larger peak separation of the two redox couples is observed for NVO, indicating a lower polarization of PEDOT-NVO electrode. Meanwhile, the galvanostatic charge-discharge profiles of PEDOT-NVO and NVO electrode at 0.2 A g 1 are also shown in Figure 4B, which demonstrates that the PEDOT-NVO indeed reflects a highly reversible (de)intercalation and lower polarization reaction corresponding to the CV results. The PEDOT-NVO electrode displays second discharge-charge capacities of 282.5/290.6 mAh g 1 with a high Coulombic efficiency of 96.5%, in comparison to the corresponding capacities of 174.8/160.9 mAh g 1 for pure NVO. We also prepared pure PEDOT and used it as the cathode material in our aqueous ZIBs, and the sample composition and structure were characterized by SEM, XRD, and Raman, as shown in Figure S7, which are well matched with the previously reported PEDOT samples. 40,47 The PEDOT only displayed a very small discharge capacity of 24 mAh g 1 , and capacity fade was serious in the subsequent 100 cycles (see Figure S8). In fact, the major role of PEDOT is to produce an enlarged interlayer spacing accompanied by oxygen vacancies, and its capacity can be negligible. It is speculated that the enlargement in interlayer spacing (7.8 to 10.8 A ˚) might not be the only factor that supports the huge boost in capacity, but the existence of V 5+ /V 4+ mixed valence by PEDOT incorporation is also critical. It has been reported that the oxygen vacancies caused by the mixed valence indeed alleviate the stress and strain that accompany zinc-ion (de)intercalation, thus enhancing the capacity and cycling stability. The cycling performance of samples at lower current density (0.2 A g 1 ) is observed in Figure 4C. Both of them show an increase in the discharge capacity and Coulombic efficiency during the initial cycles, which is caused by an activation process forming a fresher surface where the electrochemical kinetic reaction occurs. A maximum discharge capacity of 345.5 mAh g 1 is achieved at the 15 th cycle for the PEDOT-NVO electrode. Its capacity decreases slightly in the next 85 cycles, and the capacity retention is 94.8% over the whole 100 cycles. However, the NVO electrode suffers from severe capacity loss after reaching the maximum discharge capacity of 261.5 mAh g 1 , and only 63.5% of the maximum capacity is retained. In addition, the long-term cycling performance at 1 and 2 A g 1 is also remarkable (see Figures S9 and S10). The PE-DOT-NVO cathode at 1 A g 1 can maintain a stable capacity retention of 80.8% up to 1,000 cycles with an average capacity decay of 0.0192% per cycle, which is much higher than that of 46.5% over 1,000 cycle in pure NVO. More importantly, it shows good cycling performance with a capacity maintenance of 79.1% at 2 A g 1 , exhibiting a high Coulombic efficiency always close to 100%. More detailed electrochemical process of different cycles is provided by the differential capacity (dQ/dV) versus potential plots (see Figure S11). The four couples of distinct peaks are in accordance with CV curves, ascribed to the reversible insertion-extraction reactions cycled with a tiny change. The improved electrochemical performance of the PEDOT-NVO material is a result of both stable NVO with larger interstitial channels for zinc ions and polymer PEDOT offering great electronic conductivity. Figure 4D further depicts the rate capability of the two electrodes. The PEDOT-NVO electrode delivers a much higher discharge capacity of 356.8 mAh g 1 at 0.05 A g 1 than NVO electrode of 264.2 mAh g 1 . When the current density increases to a high rate of 10 A g 1 , the PEDOT-NVO can show a high capacity of 163.4 mAh g 1 with a capacity retention of 45.4% of its highest capacity, slightly superior to 42.7 % (113.0 mAh g 1 ) of NVO electrode. This performance of capacity retention is almost higher (45.4% at 10 A g 1 , 200 times the initial current at 0.05 A g 1 ) than that of reported advanced cathode materials, which is described in Table S1. The corresponding galvanostatic charge-discharge curves of the PEDOT-NVO and NVO electrodes at various current densities are given in Figure S12. Given the above results, it is reasonable to believe that by introducing a lower molar mass of NH 4 + into the vanadium-oxygen structure, the PEDOT and NVO electrodes are more inclined than large metal vanadates (e.g., zinc, sodium, lithium, magnesium, and manganese) to display their capacity even at high rates. Impressively, our battery could display a very high energy density of 353.1 Wh kg 1 at a good power density of 50 W kg 1 and maintain its energy density of 163.4 Wh kg 1 even at a superior density of 10,000 W kg 1 (based on the mass of cathode material, as well as the following contrast sample). Moreover, the performance of our obtained PEDOT-NVO cathode displays prominent superiority among the ZIBs of (NH 4 ) 2 V 10 O 25 $8H 2 O, 39 S13). We also estimate the long cycle life of the PEDOT-NVO cathode at a very high current density of 10 A g 1 . As shown in Figure 4E, it only shows a negligent capacity decay from 170.7 to 160.6 mAh g 1 over 5,000 cycles (94.1% of the highest capacity). To the best of our knowledge, it is rare to stabilize such high capacity for 5,000 cycles, which is superior to previous reported ZIBs (see Table S2). We measured the TEM imaged of the PEDOTintercalated NVO cathode electrode on the basis of the impressive cycling performance of 10 A g 1 (see Figure S14). As observed, the nanobelt morphology of PEDOT-NVO electrode was well maintained after 5,000 cycles without structural evolution, and the EDS mapping verified that the C and S elements of PEDOT remained in the composites. In contrast, we also observed the morphology and surface of pure NVO electrode (see Figure S15); the morphology of NVO suffered some damage, including many holes in the nanobelt after only 1,000 cycles. In order to determine whether the domination of the charge-storage mechanism is a diffusion-controlled or non-diffusion-controlled process, we measured the CV curves of the PEDOT-NVO cathode at various scan rates from 0.1 to 10 mV s 1 . As shown in Figure 5A, the main oxidation peaks in the curves shifted to higher voltage, whereas the reduction peaks shifted to lower voltage as the scan rate increased. This adopted analysis has been used to quantify the capacitive and diffusion-controlled contribution to the whole capacity. 54,55 Assume that the peak current density (i) obeys the power law relationship with the scan rate (n), which can be expressed as i = an b . By fitting the value of b according to the above equation, we can obtain the results of b from the main anodic and cathodic peaks, as displayed in Figure 5B. When b approaches 0.5, it indicates an electrochemical diffusion-controlled process; a value close to 1 suggests the domination of a capacitive process. The oxidation reaction of the PEDOT-NVO electrode had values of 0.813 and 0.933 for peaks 1 and 2, respectively, and the reduction peaks 3 and 4 had values of 0.857 and 0.847, respectively, which indicates that the charge-storage process is primarily dominated by pseudocapacitive characteristics rather than ionic diffusion behavior. The contribution of a capacitive process and a diffusion-limited redox process in overall capacity can be quantified according to the divided equation of i = k 1 n + k 2 n 1/2 . In Figure 5C, 61.7% of the total capacity corresponds to capacitive contribution at the scan rate of 0.1 mV s 1 , and the rest of the capacity contribution profiles at other scan rates are displayed in Figure S16. With the increase in scan rates, the contribution of capacitive characteristic rose to 67.7%, 72.3%, 77.8%, 84.4%, 88.9%, and 91.3% (Figure 5D). The capacitive versus diffusion-controlled contribution for the NVO electrode is shown in Figure S17, where the b values of peaks 1-4 are 0.88, 0.90, 0.90, and 0.85, respectively, indicating that the PEDOT-NVO electrode is synergistically controlled by capacitive and diffusion behaviors. Noticeably, the capacitive contribution on the NVO was found to be lower than that on the PEDOT-NVO electrode, especially at the lower scan rate, which is mainly attributed to the conductive polymer PEDOT. The results demonstrate that the surface capacitive storage could contribute to the whole capacity of PEDOT-NVO electrode, thereby leaving a large capacity at high current density. The apparent diffusion coefficients are obtained by the traditional galvanostatic intermittent titration technique (GITT) in both two electrodes (details described in the Supplemental Information). Figure S18 shows a typical amplified single titration, different parameters (DE S , DE t , etc.), and the well-fitted line among the voltage (V) and duration time (t) in this single process. According to the GITT curves (Figure 5E), the average D Zn 2+ values of PEDOT-NVO at the charge and discharge plateaus are about 5.3 3 10 9 and 6.4 3 10 9 cm 2 s 1 , respectively (Figure 4F). In contrast, the diffusion coefficient of the NVO electrode is also shown in Figure S19, in which the NVO electrode displays an approaching D Zn 2+ (5.9 3 10 9 and 4.9 3 10 9 cm 2 s 1 at the charge and discharge platforms, respectively). Therefore, the diffusion coefficient is not remarkably affected by the large interlayer distance, which is consistent with the similar rate capability among these two materials. Notice that the diffusion coefficient of PEDOT-NVO is higher than those of bulk V 2 O 5 , 26 31 The role of PEDOT was further examined by the electrochemical impedance spectroscopy (EIS), as shown in Figure S20, in which both electrodes present a semicircle in high frequency and straight line in low frequency. The charge transfer resistance (Rct) of the PEDOT-NVO electrode is about 300 U, which is obviously smaller than that of NVO ($500 U), revealing that the introduction of PEDOT is beneficial for the high electronic conductivity and efficient Zn 2+ transport. In addition, we measured the self-discharge behavior of the PEDOT-NVO and NVO electrodes in the aqueous ZIBs, and the results are shown in Figure S21. By monitoring the voltage drop over a 24 h rest test from a fully charged state, we found that 75.6% of the original voltage of the PEDOT-NVO electrode was retained, whereas the NVO electrode maintained only 68.8%. Generally, the self-discharge process is explained by the charge redistribution derived from the diffusion of ions out of electrical double layer. The C-C and C-S bonds interrupt the ionic interaction with the electrode-electrolyte interface, and the layered structure with the PEDOT-attached area results in a weak interaction between the electrode and electrolytic ions, thereby reducing the spontaneous charge redistribution of ions. 59 To further investigate the electrochemical mechanism during the subsequent discharge-charge process, we performed a series of characterizations including XRD, FTIR, and Raman analyses. Ex situ XRD patterns of the PEDOT-NVO cathode are displayed in Figure 6, and the measured electrode is made of polytetrafluoroethylene (PTFE) and acetylene black with samples pressed into a titanium mesh. As observed from the initial and final states, the crystal structure of PEDOT-NVO is completely reversible during the intercalation and extraction of Zn ions. Upon discharging to 0.4 V, a series of new diffraction peaks gradually appear at 12.9 , 33.1 , and 59.1 , which could be matched to layered materials of the Zn x NH 4 V 3 O 8 phase accompanied by the increase in the Zn 2+ content. However, a weak peak at 7.1 disappears in the discharge process and reappears when charged to 1.6 V. Because the solvation effect in the zinc ion insert process covers the interlayer electrosatic repulsion, the intensity of this peak at the final state is weakened. Interestingly, the intensity of these new peaks enhanced with increasing depth of discharge and then diminished after recharging to 1.6 V, which is an indicative of the reversible structural production and evolution of PEDOT-NVO that occurs with the insertion and extraction of Zn 2+ . Noticeably, not all of the initial phases were transformed into the intercalated new phase. Overall, two distinguished characteristic peaks are magnified in Figures 6B and 6C. The selected peaks located at 25.5 (3.49 A ˚) and 49.5 (1.83 A ˚) are shifted to lower 2q positions and then return to their original states. In detail, a gradual leftward shift (increasing d-spacing) is accompanied by discharging, whereas a rightward one (decreasing d-spacing) is accompanied by charging. At the end of the discharge reaction, the above two peaks appearing at 2q = 25.0 (d $ 3.55 A ˚) and 48.7 (d $ 1.86 A ˚) signify a slight increase in the respective interplanar distances. Such an enlargement of the peak shifts hints at the increase in the M-M bond distance within the V 3 O 8 layer when the zinc intercalation occurs, revealing the solid solution intercalation behavior of PEDOT-NVO that accommodates the ions via reversible expansion and contraction on several fronts without forming phase boundaries. Raman spectra elucidate the vibrations of the PEDOT-NVO material at the second cycle (see Figure S22A). Along the discharging process, an obvious negative shift and intensity enhancement corresponding to the four main peaks around 138.4, 281.3, 407.4, and 900.6 cm 1 are apparent, indicating that the skeleton-bending vibration is closely related to the intercalation of cations between the layers of PEDOT-NVO. With the voltage rising from 0.4 to 1.6 V, the intensity and position of these peaks gradually weaken and the position returns to the original state, which results from the extraction of Zn 2+ from the discharge product. Similarly, the FTIR spectra of PEDOT-NVO under various voltages are given in Figure S22B. With decreasing voltage, the characteristic peaks at 646.2, 662.1, and 1,034.8 cm 1 slowly come into existence as a result of the combination with the coordination of Zn 2+ . When it comes to the charging process, these peaks start to vanish gradually in the pristine state. These results reveal that the structure and bond vibration of a PEDOT-NVO-layered material can be well maintained upon the increase in or release of stresses in the layers during zinc-ion intercalation and deintercalation. On the basis of the above analysis, we can speculate the whole storage mechanism, which is shown in Figure 6D, and the electrode reaction is described below. First, discharge occurs on the cathode: PEDOT/NH To bear witness to the successful aqueous storage of Zn 2+ in PEDOT-NVO cathode, we carried out ex situ XRD and TEM analysis in the first cycle. As shown in Figure S23, the XRD patterns corresponding to the pristine and the first charged state almost present similar peaks, whereas the difference stands for the incompletely reversible intercalation and extraction of Zn 2+ in the PEDOT-NVO electrode during the first cycle. Noticeably, the peak at 8.3 weakens and decreases slightly to 6.4 (an increase in interlayer spacing), and it ultimately recovers to the peak located at 7.0 (a decrease in interlayer spacing), which is caused by the insertion of Zn 2+ upon the first discharge and the incomplete extraction of Zn 2+ (inset in Figure S23A). HRTEM images are used to characterize the structural change, as shown in Figure S23B. At the fully discharged state, the nanobelt predominantly maintains the layered structure and forms a newly discharge product with a larger interlayer spacing of 12.3 A ˚compared with the initial state (10.8 A ˚). The interlayer distance spacing is converted to 11.1 A ˚after charging, slightly larger than that of the pristine materials. Occasionally, the special intercalation mechanism involves the proton that could be inserted into some cathode materials. However, the above XRD patterns verified that no ZnSO 4 [Zn(OH) 2 ] 3 $xH 2 O or other possible phases were formed in the discharged or charged states for the PEDOT-NVO cathode, which is different from the cointercalation-deintercalation mechanism of H + and Zn for MnO 2 materials. Comparison of STEM-EDS mapping images of PEDOT-NVO cathode during the first discharge-charge process (see Figure S24) revealed an incomplete extraction of zinc ions since a small amount of zinc ions still remained in the layered material after initial charging. A comparison of the reversible stripping-plating reaction occurring on the Zn metal anode as observed by SEM is shown in Figure S25. After the first strippingplating reaction, the zinc foil had a dendrite-free morphology in which it presented a rougher surface than the pristine one as a result of the residual electrolyte. These results further illustrate that the contraction of the lattice spacing of (001) plane is not completely reversible in the first cycle, which matches the XRD result. In practical application, the issue of the Zn anode is another obstacle for Zn-ion batteries. The optimized substrate for the growth of metallic zinc achieved exceptional reversibility by Archer's group, and the concentration modulation enabled a dendrite-free Zn transportation process by Wang's group, which may be the potential solution to improve the stability of Zn. One noticeable question concerning the electrochemical properties of PEDOT-NVO is that its capacity is higher than that of pure NVO despite the fact that it is derived from the lower average oxidation state of PEDOT-NVO (mixed V 5+ /V 4+ ). To figure that out, we investigated the valence state in two electrodes by ex situ XPS spectra after full discharge at 0.4 V and charge at 1.6 V, as shown in Figure 7. As observed in Figures 7A and 7B, the newly appearing strong Zn 2p peaks clearly demonstrate the evident insertion of the Zn 2+ ion into the two materials. On the other hand, the weak intensity of Zn 2p peaks in the charged state indicates the presence of residual Zn 2+ from the first discharge products, which is consistent with the above XRD and TEM mapping results. Figure 7C displays the high-resolution V 2p spectra of the discharge products; the most consistent feature is the presence of V 4+ and V 3+ after insertion for both PEDOT-NVO and NVO. Despite the significant amount of V 4+ in the original PEDOT-NVO materials (Figure 2E), the ratio of V 3+ / V 4+ for the inserted PEDOT-NVO is 27.2%. In contrast, this ratio is only 20.6% for the NVO electrode. That is to say, there is more V 3+ in the discharged products. In addition, when recharged to 1.6 V, both electrodes return to a state similar to the original one full of V 5+ , with a small amount of V 4+ in PEDOT-NVO (Figure 7D). These results thus elaborate that the difference in capacity between PEDOT-NVO and NVO is caused by further conversion of V 5+ to V 3+ rather than just V 4+ to V 3+ , along with the insertion of Zn 2+ . If the larger interlayer distance of PEDOT-NVO leads to increased stability by right of more space accommodating Zn 2+ insertion and extraction, then this offers a feasible mechanism for reaching higher capacity in the reduced material. To investigate the function of oxygen vacancy and inserted PEDOT on the Zn-insertion performance proposed here, we applied density functional theory (DFT) calculation to compute the zinc-inserted energies (detailed calculations are shown in the Supplemental Information). The oxygen atoms at the different sites, where the OV-5 atom is formed most easily as a result of the lowest DEOV (Vo), are shown in Figure S26. In contrast, NH ## Conclusions In summary, a conducting polymer-intercalated NVO (PEDOT-NVO) with an increased interlayer spacing is prepared by a simple method and first investigated as cathode aqueous ZIBs. As a result, the PEDOT-NVO cathode presents a high capacity (356.8 mAh g 1 at 0.05 A g 1 ) and superior cycling stability (capacity retention of 94.1% after 5,000 cycles at 10 A g 1 ). A combination of TEM, XRD, XPS, and FTIR, along with ex situ Raman spectra results, reveals the successful intercalation of PEDOT and the highly reversible reaction during the chargedischarge process. This exploration of an intercalation compound developed here will show great insight into high-performance materials in aqueous ZIBs or other multivalent ion batteries. ## EXPERIMENTAL PROCEDURES Preparation of Organic-Inorganic Intercalated Composites PEDOT-intercalated NVO was prepared by a simple reflux method. In a typical synthesis, 0.4 g of V 2 O 5 powder and 0.2 g of APS were dissolved in 100 mL of distilled water and sonicated for 0.5 h to produce a light yellow homogeneous solution. Then, 120 mL of EDOT (C 6 H 6 O 2 S) solution was slowly injected into the solution under continuous sonication for another 2.5 h until the color turned grayish green. After that, the solution was loaded into a 250 mL flask and refluxed at 100 C for 12 h, after which the color changing to dark green. The polymer-free NVO was synthesized according to the same procedure without the addition of EDOT. Moreover, the pure polymer PEDOT material was prepared via chemical oxidation polymerization using EDOT as the starting monomer and APS as the oxidant. Specifically, 0.1 g of surfactant was dissolved in 100 mL of distilled water with sodium dodecylsulfate (SDS) to disperse the monomer EDOT solution (1.2 mL), and the solution was stirred for 1 h. Then, 2.0 g of APS-dissolved moderate distilled water was added dropwise into the above solution and then stirred overnight. The obtained product was filtered with deionized water and ethanol several times and dried in a vacuum oven at 80 C overnight. ## Material Characterization Powder XRD patterns were collected on a Bruker D8 Focus Power X-ray diffractometer with Cu Ka (l = 0.15405 nm) radiation (40 kV, 40 mA). The morphology of the synthesized products was observed with a scanning electron microscope equipped with a JSM-6390 microscope from JEOL working at 1 kV acceleration voltage. The structure of the samples was obtained by TEM and HRTEM with EDS mapping equipment through a Tecnai G2 F20 S-Twin (America FEI) microscope operating at 200 kV. The oxidative state of the samples and electrodes was tested by a XSAM800 Ultra spectrometer. The thermogravimetric measurements were performed on a Perkin-Elmer TGA 7 thermal analyzer at a heating rate of 10 C/min under a 40 mL/min airflow. A Raman spectrometer (LABRAM-1B) with a 514 nm laser source and FTIR spectroscopy (NICOLET 6700) were also applied. EPR data were measured at room temperature on a Bruker E500 EPR spectrometer. ## Measurements of Electrochemical Properties For electrochemical characterization, the working electrode was fabricated by mixing the active materials, acetylene black, and PTFE binder (according to the mass ratio of 7:2:1) to form a smooth film and dried in a vacuum oven at 80 C for 12 h. After that, the obtained film was pressed onto titanium mesh. The loading mass of active materials was controlled to be $2 mg cm 2 . For the full aqueous ZIBs, a 3 M ZnCF 3 SO 3 solution was used as the electrolyte, 2016-type coins were assembled by employing glassy fiber as separator between zinc foil anode and working electrode cathode. CV and EIS tests were performed on AUTOLAB PGSTAT302N

chemsum

{"title": "Organic-Inorganic-Induced Polymer Intercalation into Layered Composites for Aqueous Zinc-Ion Battery", "journal": "Chem Cell"}

real-time_synthesis_and_detection_of_plasmonic_metal_(au,_ag)_nanoparticles_under_monochromatic_x-ra

## Abstract: Plasmonic nanostructures are of immense interest of research due to its widespread applications in microelectronics, photonics, and biotechnology, because of its size and shape-dependent localized surface plasmon resonance response. The great efforts have been constructed by physicists, chemists, and material scientists to deliver optimized reaction protocol to tailor the size and shape of nanostructures. Real-time characterization emerges out as a versatile tool in perspective to the optimization of synthesis parameters. Moreover, in the past decades, radiation-induced reduction of metallic-salt to nanoparticles dominates over the conventional direct chemical reduction process which overcomes the production of secondary products and yields ultra-high quality and pure nanostructures. Here we show, the real-time/in-situ synthesis and detection of plasmonic (Au andAg) nanoparticles using single synchrotron monochromatic 6.7 keV X-rays based Nano-Tomography beamline. The real-time X-ray nano-tomography of plasmonic nanostructures has been first-time successfully achieved at such a low-energy that would be leading to the possibility of these experiments at laboratory-based sources. In-situ optical imaging confirms the radiolysis of water molecule resulting in the production of e − aq , OH • , and O − 2 under X-ray irradiation. The obtained particle-size and size-distribution by X-ray tomography are in good agreement to TEM results. The effect of different chemical environment media on the particle-size has also been studied. This work provides the protocol to precisely control the size of nanostructures and to synthesize the ultrahighpurity grade monodisperse nanoparticles that would definitely enhance the phase-contrast in cancer bio-imaging and plasmonic photovoltaic application.Plasmonic nanostructures are the promising tools for widespread application especially for phase contrast in bioimaging 1-3 and light trapping in the solar cell/photodetectors due to their conspicuous size and shape dependent localized surface plasmon resonance response 4,5 . Plasmonic nanoparticles not only produces the phasecontrast in image detection 6 but simultaneously cure the cancer tissue 7 because of its antibacterial properties 8 . On the other side, Plasmonic nanostructures are widely used in the photovoltaic cells due to its topology and morphology dependent unique optical, charge storage, absorption coefficient, good energy band-gap, and light trapping properties [9][10][11] . The unique Surface Plasmon Resonance Response of these nanostructures lies in the visible to IR-region which makes them suitable for the energetic solar photons trapping and resulting in enhancement in efficiency of the photovoltaic cells 4,9 . The properties of these polaritons could be tuned by optimizing the size and shape of nanostructures 10,11 . In order to optimize the specific response of the material, in-situ/ real-time characterizations emerge as a vital tool to probe the materials properties 12,13 . Another aspect of in-situ characterization is to avoid the material wastage and time. From the industrial application point of view, the material should be of high purity grade without any secondary product contamination along with the sustainable properties. As far as the chemical synthesis roots are concerned, various parameters have been optimized www.nature.com/scientificreports/ for the desired characteristics materials 14,15 . In this regards, real-time characterization provide the feasibility to monitor the materials characteristics to optimize the desired properties. In our previous work, the size and shape dependent properties of the plasmonic nanostructure fabricated under the irradiation of low and high-frequency radiations has been studied 16,17 . The results clearly depicted the dominant role of radiation-induced synthesis over the direct chemical reduction process along with providing the feasibility of in-situ characterization. The radiation induced synthesis protocol prevents the formation of unbidden secondary products because of the absence of external chemical reducing agents. Moreover, Material Science is a 3D-science, in which the reconstruction of microstructures incorporating anisotropic grains, intergranular phases, island formation on surfaces, crack distributions and deformed structures, etc., led to the interesting physics at the bottom 18 . In the last two decades, such material modification has been successfully achieved with the improving ion-beam driven instrumentations. So the characterization techniques which would enable researchers to investigate the 3D structures rather than a projection, are of interest in material science . In the microstructure imaging, optical coherence tomography based on the coherent properties of light to fetch structural feature of heterogeneous optically opaque samples is widely used with the advancement of its resolution ~ 2 μm and imaging speed 19 . The detection of metal oxides in the tissues has been successfully achieved using optical coherence tomography 21,22 . Metals (Au and Ag) have a strong X-ray attenuation coefficient among other elements naturally present in humans. Therefore, the accretion of Au/Ag NP in tumors would significantly enhance the X-ray attenuation, resulting in high contrast between tumor and healthy tissues on tomography images 23 . At present deoxyglucoselabeled AuNP are widely accepted as potential X-ray tomography contrast agent 24 . The 10-fold enhancement in tumor detection on X-ray tomography has been observed after the incorporation of 2.6 nm AuNP 25 . Moreover, M-NP not only enhance the image contrast also serve in treatment. The significant enhancement in radiation therapy has been observed in the presence of 1.9 nm AuNP 26 . The Auger de-excitation process in Ag-NP under gamma radiation leading to effective DNA breaking in a short distance less than the size of a single cell resulting in damage to tumor cell without harming other body-tissues 23 . This short-range therapeutic effect makes AuNP a potential element for selective tumor cells detection and treatment under photon-based radiation therapy. In case of plasmonics driven photovoltaic devices, the metallic nanoparticles/nanostructures have been employed for light trapping via interaction with surface polaritons. Such plasmonic nanostructures have been incorporated at the surface, back contact and in-bulk of the active layer of the device 9 . So it is important to consider the 3D-diffusion of particles to understand the dopant profile. In the past decades, a lot of efforts have been imparted focusing on this particular issue. The 3D doping profile of nanostructures have been studied by atom probe tomography which would be of interest in the plasmonic photovoltaic cell 20 . Another way of 3D imaging is by using discrete tomography in which 2D projection images obtained from different angles by electron microscopy can be reconstructed for the 3D tomography by applying the discrete algebraic reconstruction technique (DART) 27 . The availability of highly collimated and intense X-ray beams from the synchrotron source has begun the evolution of X-ray based characterization techniques with its widespread applications in physical, life and biological sciences 28 . Moreover, it allows the feasibility to use multiple characterization techniques simultaneously. In this work, we successfully achieve the task of fabrication of in-situ plasmonic nanostructure by the reduction of metal ions (M + ) to the zero-valent (M 0 ) nanoparticles in a liquid phase using water radiolysis along with the online characterization by X-ray nano-tomography/imaging technique under synchrotron monochromatic radiations. In these experiments the same monochromatic hard-X-ray beam has been utilized for the synthesis/fabrication and in-situ detection of plasmonic nanostructures which will definitely enlighten the concepts behind the anomalous behavior of nanomaterials. The successful outcome of this work will lead the feasibility of synthesizing the metal-polymer conjugates and in-situ probing the charge transport properties for bio-imaging and solar cell applications. ## X-ray nanotomography Experimental beamline setup and evaluation procedure. In material science, particle size and shape is an important factor to optimize their characteristics. Therefore, material imaging is a prime choice and can be executing with numerous probes, i.e., light, electrons, neutrons, ultrasound and X-rays 18 . X-rays allows the feasibility for in-situ experimentation to widespread range from thin to thick samples due to the high transmission and non-destructive property 29,30 . Although, low absorption affects the image contrast, but can be resolved by phase-contrast technique. It utilize the phase-difference between 0th and nth-order spatial frequencies. The contrast can be explained in terms of wave-optical approach as ray-refraction by the object. The small deviation in the direction ~ 10 μrad is sufficient to produce noticeable variation in intensity, after propagation over the ~ 1 m 31 . In wave-optical approach, the object is characterized by its complex transmission function T as 32 : where u and u 0 defines the monochromatic field mode (downstream and upstream respectively) of the object at (x, y) point of the object plane. The complex transmission function T includes the real and imaginary part of refractive index of material and can be defined as where A and φ are the absorption and phase modulation respectively. (1) u(x, y) = T(x, y)u 0 (x, y) (2) T(x, y) = A(x, y)e iφ(x,y) where µ and n are the absorption coefficients and real part of refractive index, respectively. The integral term over the z is due to the propagation direction in the object. In case of X-rays, the refractive index n of elements depends on their electron density as 31 where N 0 is the Avogardro's number, A m = atomic mass, Z = atomic number, ρ m = material density, r 0 = classical electron radius, f ′ disp = real part of dispersion correction and = wavelength. Due to high mobility and electron density, metals are of first choice in the imaging techniques. Moreover, In case of the metallic nanoparticles, surface electron density enhances and produce phase modulation/contrast. That is why, metallic nanoparticles, especially AuNP being antibacterial and non-toxic is widely used for image-contrast in biomedical applications. The optical layout and image showing the relative distance of the major components of 7C-XNI hard X-ray nanotomography beamline at Pohang light source are shown in the Fig. 1. The beamline emerges from the 1.4 m hybrid undulator having a period of 20 mm to produce high flux X-rays of power 2.95 kW at 3.0 GeV storage beam energy and 400 mA current. The Silicon (Si-111) double crystal monochromator (DCM-Vactron, Korea) has been used to tune and monochromatize (�E/E ∼ 10 −4 ) the X-rays beam. The DCM is allowed to work under liquid nitrogen cooling system and a thick diamond window 200 μm of 1 cm diameter is placed in front of the DCM to reduce the thermal effects generated by high flux X-rays 33 . The series of 10 parabolic beryllium crystal (RXOPTICS, Germany) lenses of each 1 mm in diameter having an effective aperture of ∼ 0.6 mm (because of the reduction due to absorption) is used to collimate the X-rays beam 34 . Each lens offers the focal length of 3.25 m at 6.7 keV beam energy. The photon flux at the focus point is estimated to be in the order of ∼ 10 11 photons s −1 . To ensure the reduction of spatial coherency and homogenous illumination a diffuser and a pinhole of 40 μm has been installed before the experimental stage. The sample is placed on the piezo-driven three-axis scanning stage (AG-LS25-Newport) and rotating (ABRS-150MP-Areotech) 6D stage. The nanotomography has been carried out using set of Zone plate, a phase plate, and scintillator detector. The objective zone plate is made up of 1.0 μm thick and 140 μm diameter tungsten (Zoneplates, UK) plate having ∼ 30% efficiency. The first order focal length comes out to be 40 mm at beam energy of 6.7 keV. In order to set the contrast of imaging, a 3.78 ± < 0.04 µm thick aluminum phase plate having a 10 µm hole is inserted at the focal point of zone plane, which offers the π/2 phase shift in the diffracted beam to produce a darker image in the bright field. The detector system contains an 18 µm thin, 10 mm diameter Tb : LSO scintillator crystal (FEE, Germany) and X20 optical microscope. In order to eliminate the blur effect, all optical components of beamline are operated under a vacuum of 10 −3 torr and placed on the 5 m long and 30cm thick vibration-free granite plate 34 . ## Target material processing All the chemicals of high purity (99.999%) grade AgNO 3 , AuCl 3 , and PVP were purchased from Sigma Aldrich and used as it is without any further purification. The precursor solution was formed by dissolving the metal-salt and capping agent (PVP) in high purity HPLC reagent water (SAMCHUN chemicals-Korea). (3) The precursor solution was filled in the Kapton-Polyimide tube (Cole-Parmer(Illinois)) having an outer diameter of 0.0615 ′′ ± 0.0005 ′′ and an inner diameter 0.0575 ′′ ± 0.0005 ′′ with the wall thickness of 0.00200 ′′ ± 0.00025 ′′ . To close the one end of Kapton tube, a closed end tip glass capillary has been injected. A dedicated glass capillary sealing system has been used to close the end of a capillary as shown in the Fig. 2. ## Results and discussion Real-time/in-situ detection of Au nanopaticles during irradiation of precursor samples under same beamline has been investigated using undulator based X-ray tomography beamline (7C-XNI) at pohang accelerator laboratory (PAL-South korea). The X-ray image of target sample at before starting the experiment (0 irradiation time) has been recorded as shown in Fig. 2b. Radiations induced material synthesis has been widely accepted for the production of high purity grade nanomaterials because of its precise control over the reaction rate by means of dosimetry 35,36 . In order to investigate the fundamental process of radiation induced reduction of metallic-salt, the monochromatic X-rays of energy 6.7 keV much less than the K − L edge-energy of Au has been selected for irradiation. Simultaneously, the imaging detector system has shown good resolution at 6.7 keV as determined by the previous beamline-imaging experiments 34 . There would be no external reducing agent involved in the system, therefore the chemical reaction will only be triggered under the effect of water putrefaction. At present, water decomposition using radiolysis process is carried out by ionizing radiation, i.e. radiation from the decay of radioactive materials, accelerated charged particles (electrons, protons and ions) and X-rays . The water putrefaction is governed around the radiations-track and directly proportional to the linear energy transfer (LET = −dE/dl) in the medium per unit path-length 38 . The mechanism is mainly categorised in three steps: physical stage (1 fs) , physico-chemical stage 10 −15υ 10 −12 s and chemical stage 10 −12υ 10 −16 s. The initial two stages involve the radiation energy deposition to decompose the water-molecule as 17 : The photon flux F Ph at 6.7 keV was in order of ∼ 10 9 ph/s , therefore integrated beam energy-flux can be estimated as E × F Ph = 1.07 × 10 −05 J/s . If beam cross-sectional lies inside the irradiated cell parameters, the expression for integrated dose (Gy) imparted to the sample is expressed as 16 : where t is the irradiation-cell thickness, ρ is the sample density and T is the irradiation time. Therefore the energy transferred to the sample is ∼ 10 −07 J/s , and the integrated dose for irradiation time of 05-60 min is about to 3.2 − 57.8 KGy with 10 Gy/s of dose rate. The radiolysis of water under the exposure of X-rays could be explained by simultaneous obtained optical imaging of the samples revealing the production of bubbles as shown in the Fig. 2c. The generation of bubbles on X-ray irradiation confirms the production of O 2 molecule, therefore the putrefaction of H 2 O . Aqueous electrons e − aq and atom radicals H • are strong reducing agents towards the metal salt with standard potential E 0 (H 2 O/e − aq ) = −2.9 V NHE and E 0 (H + /H • ) = −2.3 V NHE respectively, whereas hydroxyl radicals OH • are strong oxidant agents with standard potential E 0 (OH • /H 2 O) = +2.7 V NHE 37 . The putrefaction and recombination of radicals into water-molecule is continuously proceeding making an eight-loop reaction 17 as shown in Fig. 3. This loop is mediated by the H • and OH • radicals maintaining the equilibrium in destruction and generation rate of molecular products as: The chemical stage involves the recombination and diffusion of the generated product in the solution which reacts with solute, resulting in the reduction of metallic-ions (M + ) to zero-valent (M 0 ) particles 17,41 . Although OH • might reverse the reaction M 0 to M + , but the probability of such event is negligible due to the large concentration of e − aq /H • /O − 2 than OH •16 . Moreover the OH • can be trapped using additional isopropanol or ethanol as scavengers 42 . The radiation induced reduction of metallic-salts to nano-particles mainly follow two generation scheme as shown in Fig. 3. As the zero-valent M 0 atoms are formed, nucleation process begins to build nanostructures. The M 0 n coalesce to form the nanostructure M 0 n , which further interacts with the M 0 atom and synthesize the M − NP . If M 0 n interact with M 0 m nanostructure, it results in the growth (M 0 n+m ) of M − NP with larger size 16 . The in-situ optical imaging of kapton-polyimide tube containing precursor solution at the beginning and after the exposure of 6.7 keV monochromatic X-rays is shown in the Fig. 2d,e respectively. The optical imaging of the samples was recorded under the illumination of focused laser beam on the capillary in the dark field. At the beginning of the experiment, a clear image of the kapton tube is recorded. But after the X-ray exposure, light scattering by the tube has been observed that reveals the formation of plasmonic nanoparticles. The scattering and absorption of light is an fundamental properties of plasmonic nanoparticles due to the coexistence of size and shape dependant localized surface plasmon resonance (LSPR) response. The effect of monochromatic X-rays of energy 6.7 keV towards the reduction of metal-salts (AuCl 3 and AgNO 3 ) has also been investigated as a function of irradiation time. A polypropylene cell having kapton windows is specially designed for holding the precursor solution and to minimize the loss of X-ray dose by holder absorption 17 . As the sample holder size was small and magnetic stirring was not possible. So, the 6D sample holder stage has been utilized to uniformly scan the full area by continuous constant 2D motion 33,34 as shown in Fig. 4a. This technique produces the similar effect as of stirring during the reaction. At first, the reaction was carried out without any scavenging agent and extinction spectra of prepared samples has been investigated for their respective plasmonic effects as a function of X-ray irradiation time as shown in Fig. 4b,c. The change in the color (for Au: greenish-yellow to pink, and for Ag: transparent to yellow) during the irradiation is attributed to their respective LSPR-response and the first visible optical confirmation for the formation of plasmonic nanostructures 43 . The relative variation in the color of samples as a function of irradiation time is shown inset of Fig. 4b,c. The extinction spectra of both Au and Ag-salts irradiated with X-rays show characteristics plasmon band at ∼ 550 nm and ∼ 420 nm respectively 44 , which confirms the formation of nanostructures at 6.7 keV X-ray irradiation. The maxima in the extinction spectra is attributed to the transverse oscillation of electrons due to the absorption as well as scattering of light at particle surface, whereas minima at ∼ 500 nm in Au and ∼ 330 nm in Ag-case to the imaginary part of size dependent refractive index (n Im = Im([ǫ( )] 1/2 ) 45 . The intensity of the plasmon-band increases with the irradiation time clearly depicting the rise in concentration of nanoparticles with time 42,44,46 as shown in Fig. 4d,e. The time took to complete the reaction is depend on the sample holder size, so it is been recommended to select an appropriate design for radiation induced synthesis. The sample holder size can be made comparable to the beamsize and precursor solution must be allowed to flow at constant rate for better results. The plasmon-band position (centroid) and broadening (FWHM) in extinction spectra of M − NP belongs to particle size, shape, and surface charge 47 . In case of the Au-nanostructures, a shift in the plasmon-band towards higher wavelength is observed, whereas edge shift to lower wavelength following the similar trend as in Ag-nanostructures. The trend of shift in plasmon-frequency is attributed to the size growth alongwith the production of smaller size particle as function of irradiation time 46 , which is in agreement with the theory of radiolysis based nanostructures synthesis. Simultaneously, X-ray tomography of the irradiated sample has been recorded as shown in the Fig. 5. The monochromatic X-ray beam has been focused on a point where the high flux of formed nanoparticles was observed by in-situ optical imaging. In the scintillator based detection system, three modes of beams were detected i.e., positive first-order, zeroth-order, and negative first-order image. Since the resolution is high enough, so the arrangement is able to separate them and collect only the positive first-order image beam. The kaptonpolyimide tube is placed in the vertical position and the formed particles are highly mobile as observed by optical might be because of the effective zeta-potential at the nanoparticle surface 42 . The X-ray imaging of the reference-scale target depict the resolution of beamline and detection system as shown in Fig. 5. The image-date was calibrated using reference-target to performed particle-size and there distribution analysis. The detector system is allowed to capture the diffracted X-rays for 1 s after 5 min of exposure/irradiation. One offsite image has been captured which was subtracted from the final image to remove background effects. The obtained X-ray images were processed by the ImageJ − 1.41o software 48 . The images were treated with the bandpass filters, threshold, brightness, and contrast filters. Figure 5b,c shows the X-ray image of the reference scale target and the corresponding sectional line-profile showing 1 nm spacing in the grid-graph reveals the high resolution of system geometry. Figure 6a-c. shows the final real-time obtained X-ray image of Au-nanoparticles. The circular rings in the X-ray images (Fig. 6a,c) are observed because of the scattering by the air-bubbles produced during the irradiation as also detected by in-situ optical-imaging. The sectional line profile of the X-ray images shows the particle size of ∼ 1-2 nm, and the corresponding sizedistribution histogram reveals the average particle-size of 1.12 ± 0.29 nm with 26.58% polydispersity. The inset (Fig. 6b) shows the single Au-nanoparticle captured by X-ray imaging confirms the spherical nature of particles. The same sample has also been investigated by transmission electron microscopy (TEM), and the results shows the spherical Au-nanoparticles of averaged particle size 1.57 ± 0.30 with 19.34% polydispersity as shown in Fig. 6, which is in good agreement with X-ray tomography. As per the literature, it's the first time real-time/in-situ X-ray nanotomography of the plasmonic nanostructures has been successfully achieved using a single monochromatic synchrotron X-rays beamline. With www.nature.com/scientificreports/ the growing electronics and continuously improving instrumentation, one could precisely control the size and quality of plasmonic nanoparticles using better X-ray imaging resolving power. Therefore, for bio-applications it would be great idea to synthesise plasmonic nanoparticles at active-site to get enhanced activity to target tissues without harming the other beneficial bacterias 49,50 . The lowest and highest dose irradiated samples of Au and Ag have been studied using TEM as shown in Fig. 7. It has been observed that there were two generation of particles. The small-size particles in the vicinity of radiation path and coalesce to grow particle size as moves from the radiation track under the effect of its own zeta-potential. Therefore the average particle-size grows from 7.35 ± 2.25 nm to 11.65 ± 2.32 nm for Ag and 1.57 ± 0.30 nm to 4.61 ± 0.13 nm for Au and polydispersity decreases from 30.59% to 19.95% for Ag and 19.34% to 5.32% for Au with the increase in integrated-dose from 3.21 − 77.18 KGy . It is confirmed from the results that the reduction starts within a minute of X-ray irradiation, which would make this technique dominate over the conventional methods. Furthermore, the effect of surrounding media on the particle-size evolution has been investigated. Three precursor samples of AuCl 3 (a), AuCl 3 + PVP(b) and AuCl 3 + Ethanol in deionized-water were irradiated at 6.7 keV X-rays at constant integrated dose. The average particle size of Au-nanoparticles is 1.56 ± 0.30 nm for (a), 2.75 ± 0.74 nm for (b) and 27.63 ± 0.29 nm for (c) clearly reveals that the surrounding media significantly affect the particle-size as shown in Fig. 8. In case of sample (a) there might be a chance of oxidation (reverse reaction) under OH • radicals. Whereas PVP has enough ability for surface capping and stabilize the nanoparticles results to prevent the oxidation 16,17 . Moreover the addition of Ethanol (CH 3 CH 2 OH) not only scavenge OH • radical but also simultaneously produces the CH 3 C • radical 42 which act as strong reducing agent, therefore the reduction rate increases in sample (c) resulting in particle size growth. ## Conclusion In this work, real-time/in-situ synthesis and detection of plasmonic metal (Au and Ag) nanoparticles have been successfully achieved by Synchrotron monochromatic 6.7 keV X-rays based Nano-Tomography technique. The experiments were performed during the irradiation process using a single X-ray beamline. With the successful outcome of real-time X-ray nano-tomography of plasmonic nanostructures at such a low energy (6.7 keV) would lead to the possibility of these type of experiments at laboratory-based tabletop sources. The X-ray irradiation induced radiolysis of water molecule resulting in the production of e − aq , OH • , and O − 2 was confirmed by in-situ high resolution optical imaging detectors. The results clearly revealed the production of mono-disperse plasmonic nanoparticles and the particle size can be well-controlled by X-ray dose and surrounding media. With our best knowledge, this is the first experiment utilizing the single X-ray beam for reduction of M + to M 0 under water radiolysis and in-situ/real-time imaging characterization. This work provided the protocol for real-time detection of nanostructures and precisely control the size of nanostructures under any kind radiation ( X − rays, γ −rays, µ−wave , electron/proton etc.) or chemicals induced reduction of metallic-salts to synthesize the high-purity grade monodisperse nanoparticles that would enhance the plasmonic photovoltaic application. This work also led to the phase-contrast imaging of active-site cancer tissue without harming other body-tissues.

chemsum

{"title": "Real-time synthesis and detection of plasmonic metal (Au, Ag) nanoparticles under monochromatic X-ray nano-tomography", "journal": "Scientific Reports - Nature"}

modeling_and_optimization_of_radish_root_extract_drying_as_peroxidase_source_using_spouted_bed_dryer

## Abstract: The main advantages of the dried enzymes are the lower cost of storage and longer time of preservation for industrial applications. In this study, the spouted bed dryer was utilized for drying the garden radish (Raphanus sativus L.) root extract as a cost-effective source of the peroxidase enzyme. The response surface methodology (RSM) was used to evaluate the individual and interactive effects of main parameters (the inlet air temperature (T) and the ratio of air flow rate to the minimum spouting air flow rate (Q)) on the residual enzyme activity (REA). The maximum REA of 38.7% was obtained at T = 50 °C and Q = 1.4. To investigate the drying effect on the catalytic activity, the optimum reaction conditions (pH and temperature), as well as kinetic parameters, were investigated for the fresh and dried enzyme extracts (FEE and DEE). The obtained results showed that the optimum pH of DEE was decreased by 12.3% compared to FEE, while the optimum temperature of DEE compared to FEE increased by a factor of 85.7%. Moreover, kinetic parameters, thermal-stability, and shelf life of the enzyme were considerably improved after drying by the spouted bed. Overall, the results confirmed that a spouted bed reactor can be used as a promising method for drying heat-sensitive materials such as peroxidase enzyme.Enzymes are protein catalysts extensively used in various industries. However, their sensitivity to heat, especially in the form of liquid, is a major application problem 1 . In this regard, various methods including the use of osmolytes 2 , mutagenesis 3 , immobilization 4-7 , and drying 8 have been developed to increase the thermal as well as the storage stability of the enzymes. One of the impressive and applicable methods for improving enzyme stability is enzyme drying. This technique significantly reduces the initial solution weight as well as packaging and transportation costs 9 .One of the oxidoreductase heat-sensitive enzymes is peroxidase that catalyzes a wide variety of reactions in the presence of peroxides. According to the literature, peroxidase has various applications in industries, such as the removal of the phenolic pollution from the wastewaters, decolorization of the synthetic paints, bio pulping and biobleaching, analysis and diagnostic kits, design and construction of biosensors, and synthesis of aromatic amines, phenolic compounds and polymers 10 . Peroxidase can be extracted from plants and animals or produced by microorganisms in a fermenter 10,11 . The enzyme extraction from the plant sources has considerable advantages including low production cost and renewability 12 . Peroxidase has been extracted from plants such as Brassica rapa, Lycopersicon esculentum, Raphanus sativus L., and Brassica oleracea [13][14][15][16] .Few papers have been published about enzyme drying by spouted bed 17 . Many studies have been published on spouted bed drying in other fields of studies such as food industries and pharmaceticals 18,19 .According to relevant literature reviews, several methods have been applied for the preparation of enzyme powder, including freeze drying 20 , spray drying [21][22][23] , and spouted bed drying 17 . Spray drying of alpha-amylase 20 , protease 22 , lipase 24 , freeze-drying of alpha-amylase 20 and spouted bed of lipase 17 are examples of enzyme drying. Among the mentioned dryers, the spouted bed can be used as an effective drying equipment, due to its advantages such as low cost 25 , ability to work at low temperature, and high volumetric evaporation rates under identical thermal conditions 26 .In this study, Peroxidase was extracted from garden radish (Raphanus sativus L.) as a low-cost source 27 and a full factorial experimental design was employed to evaluate the influence of input operating variables (inlet air temperature and flow rate) on drying by spouted bed. After process optimization, an enzymatic reaction in the presence of the fresh enzyme extract (FEE) and dried enzyme extract (DEE) was carried out to evaluate the enzyme catalytic activity. Kinetic parameters of the enzymatic reaction (V max and K m ) were obtained using Michaelis-Menton (M-M) equation. Moreover, the thermal stability and shelf-life of FEE and DEE were determined and compared. To the best of our knowledge, no experimental or modeling study has been investigated on the peroxidase drying by a spouted bed. ## Materials and methods Materials. The roots of the garden radish were purchased from a local market, Kerman, Iran. Also, hydrogen peroxide (H 2 O 2 ), 3,3′,5,5′-tetramethylbenzidine (TMB), dipotassium phosphate (K 2 HPO 4 ), Monopotassium phosphate (KH 2 PO 4 ), bovine serum albumin (BSA), and Coomassie Brilliant Blue G-250 (CBBG) were all purchased from Sigma-Aldrich (St. Louis, MO, USA). Double distilled water was used in all experiments. Enzyme extraction and quantification. The enzyme extract was obtained using the technique defined by Riazi et al. 15 with a slight modification. The garden radish roots were peeled and the extract was prepared with a juicer. The slurry solution was then filtered, homogenized, and stored in a freezer (− 20 °C) until used. The total protein concentration of the extract was estimated according to Bradford's method 28 using bovine serum albumin (BSA) as standard. All experiments complied with relevant institutional, national, and international guidelines and legislation. Drying procedure. The drying procedure was performed in a spouted bed consisting of a plexiglass cylindrical column with an inner diameter of 90 mm and a height of 300 mm that was connected to the conical base of the dryer (the internal angle of 60° and the inlet orifice diameter of 15 mm). Details of the used spouted bed and experimental procedure can be found in our previous published paper 18 . The main components of the system were a heater with two elements by total power of 8 kW, a blower with a power of 2.2 kW (GREENCO 2RB, China) which was equipped with a three-phase inverter, a peristaltic pump (WPX-1, Welco Co., Japan), an air flow meter, and a cyclone (Fig. 1). The temperature and humidity of the inlet and outlet air were controlled using the sensors installed at different points. Glass granules (diameter 3 mm, sphericity 1, density 2343.1 ± 9.8 kg/ m 3 ) were considered as the inert materials to be a carrier for the liquid film, a conductive heat transfer medium, and a mechanical cleaner of the bed 29 . When the outlet air temperature showed a constant value by Data Logger (Testo, T4 176), a constant feed flow rate with 1.22 ± 0.09 ml/min was dropped on the glass beads (360 g) using the peristaltic pump. The drying process was performed according to convective and conductive heat transfer in the bed. At the end of each experiment, the produced powder at the bottom of the cyclone was collected to analyze the enzyme activity. After each test, the granules were removed from the bed, washed several times, and dried for further use. Experimental design. To evaluate the effect of drying conditions on the REA, the experimental design was conducted at different inlet air temperatures and flow rates. Minitab 17 software was used to design the experiment and analyze the obtained data. A full factorial design, considering the inlet air temperature at three levels of 50, 60, 70 °C and the dimensionless air flow rate at two levels of 1.2 and 1.4 resulted in 18 experiments that were performed with three times replication. To predict REA at different conditions, a first-order polynomial relationship relative to relevant variables has been presented in Eq. ( 2): where, REA% is the response of model, β 0 , β i , and β ij are the constant, linear, and interaction coefficients of the models, respectively. X 1 and X 2 are the inlet air temperature and dimensionless air flow rate (the ratio of Q in to Q min ), respectively 33 . To achieve the enzyme maximum residual activity, the obtained data were analyzed using RSM. Enzyme powder, obtained under the best drying conditions, was utilized in the subsequent experiments. The catalytic activity of the enzyme. To evaluate the activity of the dried peroxidase enzyme, the reaction of 0.6 mM TMB with 1 mM hydrogen peroxide in the presence of the DEE and FEE was carried out 31 . The specific activity of the enzyme can be changed by the temperature, pH, and concentration of the substrate. In order to find the optimum value of pH, peroxidase activity was measured at various pH values from 3 to 10 at ambient temperature. The optimum temperature of the reaction was also determined by the measurement of the FEE and DEE activity during the reaction at different temperature ranges (10-85 °C) under the optimum pH 34 . ## Kinetic parameters of the enzymatic reaction. Under the obtained optimum conditions, the reaction between TMB and hydrogen peroxide was carried out in the presence of the peroxidase and different concentrations of the substrate (0.06-0.6 mM TMB) and 1 mM H 2 O 2 ## 34 . The rate of the reaction at different conditions can be calculated by Michaels-Menten (M-M) equation. By inverting this equation, a linear relationship can be obtained which is called Line weaver-Burke Equation (Eq. ( 3)). The kinetic parameters of the reaction can be determined from the slope and intercept of this line 35 . (1) REA = Specific activity of the powder enzyme Specific activity of the extract enzyme × 100. ( www.nature.com/scientificreports/ In this equation, V s is the reaction rate (µM/min mg), V max is the maximum reaction rate (µM/min.mg), and K m is the M-M equation constant (mM). The product of the mentioned reaction was a colored compound and the color changed with the reaction progress. Therefore, the intensity of light absorption could be determined at different times by a spectrophotometer. Finally, the concentration of the product (C P ) at different times was determined from the Beer-Lambert equation. In the plot of C P versus time, several slopes can be obtained at different initial substrate concentrations. The reaction rate (V s , µM/min.mg) was calculated by dividing the slope of plots by the amount of the protein content. Peroxidase thermal-stability and self-life during storage. To determine the possible variation in enzyme thermal-stability, the DEE and FEE were incubated for 10 min at different temperature ranges (25-80 °C), and then after 5 min incubation at room temperature, enzyme activity was measured 2 . The shelf-life of DEE and FEE during storage was followed by the determination of the changes in the enzyme activity during storage at − 4 °C. ## Results and discussion Minimum spouting air flow rate. According to Fig. 2, the minimum spouting air flow rate was observed to be 355 l/min. The ratio of Q in to Q min in each experiment is a dimensionless parameter for investigating the effect of drying on the residual activity. ## Experimental design. Full factorial experiment design methodology and model consequence for REA. The full factorial design matrix with two independent variables and results obtained from empirical experiments and model prediction are shown in Table 1. Based on the experimental data, the following equation Eq. ( 4) can be expressed as a model that shows a correlation between the REA and the operating variables: To evaluate the importance and suitability of the obtained model, the analysis of variance (ANOVA) was carried out and the results are shown in Table 2. From the table, it can be concluded that the regression model had a high determination coefficient (R 2 = 0.95). As can be observed, all the terms of the model (the linear and interaction terms) have a p-value less than 0.05, which means that both terms are significantly effective on REA 36,37 . Moreover, the calculated F-value for the regression model is meaningfully higher than the acquired F-distribution, which demonstrates the predicted strength of the fitted model 38 . According to this result, the appropriateness and credibility of the model are confirmed for the simulation of the REA. The residual analysis was used to further check the model adequacy. This analysis indicates the difference between the experimental data and the computed data by the model. Figure 3 illustrates the residual plots related www.nature.com/scientificreports/ The higher amount of the T-value and lower amount of the p-value reveal that the corresponding coefficient is more significant and has a greater effect on the response. The results demonstrated that the inlet air temperature (p-value = 0.0), flow rate (p-value = 0.0), and their interaction (p-value = 0.005) were significant parameters in peroxidase drying process. As seen, some of the coefficients are positive and some are negative. The positive or negative coefficients display that these parameters will increase or decrease the response, respectively. Among the discussed parameters, the T had a negative effect on REA while Q had a positive effect. Effect of the inlet air temperature and flow rate on the REA. The effect of the inlet air temperature and the flow rate on the REA are illustrated in Fig. 4. As it is obvious from Fig. 4a, the flow rate has a positive effect and REA increases with the increment of the flow rate at the constant values of T. Conversely, the inlet air temperature has a negative effect on REA and an increase in the inlet air temperature from 50 to 70 °C leads to a decrease in the REA. The results in Fig. 4a demonstrate that at higher inlet air temperatures, the positive effect of the inlet air flow rate was more noticeable on REA, compared to lower temperatures. Furthermore, the slope of lines in Fig. 4b is different, which indicates the importance of the effect of each factor on the REA. To further investigate the influence of each factor on REA and to discover the relative optimum range, the contour and surface plots were obtained. Figure 4c,d depict the contour and surface plots as a function of two input operating factors. The darker area in Fig. 4c indicates the higher amount of REA. Figure 4d represents the effect of the temperature and the air flow rate on REA; the maximum of these parameters occurred at T = 50 °C and Q = 1.4. According to the obtained results, it was found that by increasing the temperature, the REA decreased, which agrees with earlier research 22,23 . Indeed, increasing the temperature can change the chemical structure of the enzyme and therefore reduce the enzyme activity. On the other hand, at the higher inlet air flow rate, due to the lower residence time of the powder, the enzyme was less affected by the heat and then the amount of the REA increased. Under the best drying conditions (T = 50 °C and Q = 1.4), the REA was obtained to about 38.7%. ## Effect of drying on the catalytic activity of the enzyme. The effect of pH on the enzymatic reaction at 25 °C has been shown in Fig. 5a. By increasing the pH value from 3 to about 5.75 for FEE and from 3 to 5 for DEE, the specific activity of the enzyme was enhanced. Further increase in pH value reduced the residual www.nature.com/scientificreports/ enzyme activity. As the result, it was inferred that the optimum values of 5.75 and 5 can be considered for FEE and DEE, respectively. The effect of temperature on the enzymatic reaction at the obtained optimum pH has been shown in Fig. 5b. The optimum temperature of the reaction catalyzed by FEE and DEE was observed at about 35 °C and 65 °C, respectively. As can be seen in this figure, although the reduction in activity owing to drying is noticeable, the optimum temperature of the enzymatic reaction has been increased in the presence of peroxidase as powder form. As shown in Fig. 5, the activity of the enzyme has been affected upon drying. The optimum pH shifted to the acidic pH in DEE compared with the FEE. Interestingly, the optimum temperature of the reaction for DEE is increased by more than 1.8-folds. ## Effect of drying on the kinetics of the peroxidase reaction. To evaluate the influence of initial concentrations of TMB on the product concentration of FEE and DEE, the experiments were carried out in the presence of different concentrations of TMB (0.06-0.6 mM) and the obtained results are presented in Fig. S1 (Supporting file). As evident, at initial times of reaction (about until 3 min), the slope of each line is constant and independent of the substrate concentration. Therefore, for each line, an initial rate was obtained in accordance with the initial substrate concentration. The reaction rates were obtained by dividing the obtained slope by the amount of protein content. The obtained results from Fig. S1, the initial slopes, and the reaction rates are presented in Table 4 for the dried enzyme and the fresh extracted one. According to Eq. ( 4), by drawing the inverse of the reaction rates versus the initial substrate concentrations, the M-M parameters (K m and V max constants) were obtained (Fig. 6a) as presented in Table 5. After obtaining the constants, the rate of the reaction could be calculated from the M-M equation for different concentrations. The calculated rates from the M-M equation and the experimental data are presented in Fig. 6b. As can be seen, a clear decrease in enzyme activity was observed in DEE. The calculations showed that the FEE activity was reduced by 48.9%, on average, after the drying process. In addition to the physicochemical properties, the kinetic parameters of peroxidase were also affected upon enzyme drying, which may be due to peroxidase structural changes. Based on the results, because of drying, the affinity of the enzyme to substrate increased and the maximum reaction rate (V max, DEE ) decreased. Indeed, amino hydrophilic acids are in the exterior structure of the enzyme, creating hydrogen bonds between the enzyme and the water molecules 39 . As a result of drying, the aqueous medium around the enzyme is removed, causing hydrogen bonds to break down 40 , which leads to an increase in the salt concentration followed by a change in the electrostatic interaction between charged amino acids 41 . These changes may cause a variation in www.nature.com/scientificreports/ the enzyme. Changes in kinetic properties as a result of enzyme structure variation during drying have been reported previously . Effect of drying on the thermal-stability and shelf-life during storage. Relative activity at each temperature indicates the ratio of the specific peroxidase activity at a specific temperature to the maximum specific peroxidase activity which is the specific peroxidase activity at 45 °C42 . As can be seen in Fig. 7, the enzyme www.nature.com/scientificreports/ powder showed more stability toward heating compared to the extract one. For instance, at 80 °C, enzyme extract activity tended to be zero, while more than 30% of the activity of enzyme powder was remained. Moreover, the shelf-life of the powder and extract was examined during storage. The peroxidase powder maintained 87.24% of its initial activity after storage at − 4 °C for 9 months while the peroxidase extract lost about 20% of its initial activity after one month of storage at the same temperature. In accordance with Fig. 7, the thermal stability of the enzyme powder was improved during drying in the spouted bed. Although the increase of enzyme thermal-stability was reported by several methods 2,41 , to the best of our knowledge, this is the first time that the enhancement of peroxidase thermal-stability was reported by a drying process. The shelf-life of the DEE during storage was also improved owing to drying in the spouted bed, which is in agreement with the results of the enzyme drying in spray dryer 43 . Generally, improvement of optimum temperature, thermal stability, and shelf-life during storage has several advantages for the utilization of dried enzymes in industries which results in low-cost storage and long-time preservation as well as its use at higher temperatures 1,9 . As shown in the results, the enzyme powder is active and more stable in the presence of heat, but the residual activity was relatively low. Therefore, further studies are needed to improve the residue activity as well as the other physicochemical properties of the dried enzyme by the use of protective or stable agents. Until now, several studies have been conducted and used some protective agents that are presented in Table 6. ## Conclusion The enzymes can be extracted directly from the natural resources of plants and animals; however, due to the short shelf-life of fresh enzymes, these materials should be dried. In this research, the root ingredients of garden radish which is a rich source of peroxidase were extracted and the resulting paste was dried using the spouted bed dryer with inert glass beads. The effect of important operating factors (the inlet air temperature and the air flow rate) and their interaction on the residual activity of enzyme were established by full factorial experimental design methodology for the first time. The ANOVA analysis indicated that the inlet air temperature, contrary www.nature.com/scientificreports/ to the air flow rate, had a negative effect on the residual activity of the enzyme. The obtained results indicated a reduction of the enzyme activity after the drying process which can be due to a change in peroxidase structure. After evaluating the optimum values of operating factors (T = 50 °C and Q = 1.4), the effect of the drying process on the thermal-stability and kinetics of the peroxidase was investigated. The kinetic studies were carried out according to the Michaelis-Menton (M-M) equation, and V max and K m of enzymatic reaction were obtained for different TMB concentrations. According to the stability studies, it was confirmed that spouted bed drying of peroxidase had an impressive effect on the thermal-stability and shelf-life which improved up to 50% compare to the fresh enzyme. Overall, it could be concluded that the spouted bed dryer can be a promising method -as a successful unit operation for drying the heat-sensitive materials such as enzymes. Received: 1 February 2021; Accepted: 28 June 2021

chemsum

{"title": "Modeling and optimization of radish root extract drying as peroxidase source using spouted bed dryer", "journal": "Scientific Reports - Nature"}

geminal_dimethyl_substitution_enables_controlled_polymerization_of_penicillamine-derived_β-thiolacto

## Abstract: To access infinitely recyclable plastics, one appealing approach is to design thermodynamically neutral systems based on dynamic covalent bond, the (de)polymerization of which can be easily manipulated with low energy cost. Here, we demonstrate the feasibility of this concept via the efficient synthesis of polythioesters PN R -PenTE from penicillamine-derived b-thiolactones and their convenient depolymerization under mild conditions. The gem-dimethyl group adjusts the thermodynamics of (de)polymerization to near equilibrium, confers better (de)polymerization control by reducing the activity and conformational possibilities of the chain-end thiolate groups, and stabilizes the thioester linkages in the polymer backbone. PN R -PenTE with tailored properties is conveniently accessible by altering the side chains. PN R -PenTE can be recycled to pristine enantiopure b-thiolactones at >95% conversion from minutes to a few hours at room temperature. This work highlights the power of judicious molecular design and could greatly facilitate the development of a wide range of recyclable polymers with immense application potentials. Polythioesters are promising candidates as sustainable polymers, but their controlled and selective (de)polymerization remains a significant challenge. In the current study, we addressed this problem by designing a thermodynamically neutral and kinetically trapped system based on penecillamine-derived b-thiolactones, opening the possibility of establishing a circular economy for plastic materials. Wei Xiong, Wenying Chang, Dong Shi, ..., Xuhao Zhou, Er-Qiang Chen, Hua Lu eqchen@pku.edu.cn (E.-Q.C.) chemhualu@pku.edu.cn (H.L.) HIGHLIGHTS Highly controlled polymerization using gem-dimethyl substituted b-thiolactones Efficient depolymerization due to a conformationally constrained terminal thiol Semicrystalline polythioesters with high molar mass and tunable mechanical properties Thorpe-Ingold effect for tuning the reversibility of ring-opening polymerization ## INTRODUCTION The annual global production of plastics has increased more than 20-fold since 1964, reaching 348 million metric tons in 2017. 1 The rapid accumulation of petroleumbased plastic wastes has created one of the greatest environmental crises in the world. 2 Single-use plastics have been banned in Europe, while other countries (e.g., China) are expected to enact similar regulations in the near future. In conjunction to developing methodologies for the degradation of existing plastics, the need for infinitely recyclable new plastics from renewable feedstock has received vast global attention. As previously suggested by Endo, Albertsson, and others, one of the many appealing strategies is to design thermodynamically near-equilibrium systems for easy manipulation of the polymerization and reverse depolymerization. Along this direction, many biomass and CO 2 derived synthetic polymers, mainly polyesters and polycarbonates, have shown their capability of establishing a circular economy of monomer-polymer-monomer. Despite the advancement, the development of recyclable polymers under mild condition and at low energy cost have been limited. Some of the thermodynamically neutral systems may suffer from low selectivity in monomer recovery as a result of the generation of oligomeric mixtures, thus necessitating extra work to push the equilibrium towards monomer formation. Kinetically trapped polymers such as poly(g-butyrolactone) can be completely recycled, elegantly demonstrated by Chen et al., but the polymerization ## The Bigger Picture Chemically recyclable polymers from renewable feedstock hold great promise for solving the imminent global plastic-waste crisis. Despite recent advances, challenges including high energy consumption, the limited choices of sustainable polymers, and side reactions that could hamper complete monomer recovery are still yet to be addressed. By introducing a gem-dimethyl substitution, we succeeded in regulating both the forward polymerization of b-thiolactones and the backward depolymerization of polythioesters in a highly controlled fashion and with low energy input. Moreover, our strategy allows facile access to semicrystalline plastics with tailorable mechanical and thermal properties from biorenewable penicillamine feedstock. We envision that our design principle can facilitate the development of a new generation of chemically recyclable polythioester-based polymers, providing a sustainable solution to the ongoing environmental and economic challenge caused by nondegradable plastics. of the non-strained g-butyrolactone requires very low temperatures (30 C to 60 C) because of unfavorable thermodynamics (DG P = + 6.4 kJ/mol). Moreover, both the polymerization and depolymerization need strong catalysts and demanding conditions because of the relatively high energy barrier for ester activation. 19,31 Thus, introducing more dynamic bonds to thermodynamically near-equilibrium system could be a viable but underdeveloped approach to design sustainable polymers. Polythioesters (PTE) are one such intriguing example because of the dynamic thioester bonds in their backbones, which are more reactive than their oxoester analogues. However, PTEs are significantly less explored than polyesters due to a lack of controlled ring-opening polymerization (ROP) methods that convert thiolactones to high-molar-mass (M n ) polymers with narrow dispersity (Ð). A few recent advances by the Bowman, 37 Lu, 40 and Gutekunst 39 groups achieved modest-to-good control of PTEs, but the recycling of the polymers were not investigated. The use of amino acid as the feedstock is a smart approach to access chiral and semicrystalline polymers. We have recently reported the controlled synthesis of PTEs from 4-hydroxyproline-derived thiolactone (ProTL) monomers and demonstrated that the polymers can be conveniently depolymerized. 48 Nevertheless, these polymers are found to be brittle due to their prolyl backbone with relatively restricted conformations, necessitating further optimization of appropriate side chain and new backbone design. 49 An interesting study by Suzuki and coworkers reported the synthesis of PTEs from a cysteine-derived b-thiolactone (CysPTE). 50 Unfortunately, the resulting polymers are characterized by relatively low molar mass (typical M n < 10 kg/mol), broad dispersity (Ð $ 1.6-2.4), and mixed linear and cyclic topologies. The underlying challenges include undesirable chain transfer, reshuffling, and backbiting, all or at least partially attributable to the extensive transthioesterification side reactions. CysPTEs are also difficult to depolymerize for monomer recycling owing to the highly strained 4membered b-thiolactone ring. One classical strategy of accelerating ring closure and stabilizing strained rings in physical organic chemistry is the gem-disubstituent effect, so called Thorpe-Ingold effect. 51,52 We thus hypothesize that the introduction of a geminal dimethyl (gem-DM) group on the four-membered ring could tune the thermodynamics to near equilibrium for improved propensity of depolymerization, and in the meantime mitigate the reactivity of both the chain ends and PTE backbone for reduced transthioesterification (Scheme 1). Notably, such monomers can be easily produced and tailored with various side chains starting from a naturally occurring amino acid, D-penicillamine. 53,54 ## Controlled Polymerization We first synthesized five penicillamine-derived b-thiolactone monomers (N R -PenTL) with different side-chains (Scheme 1B), including N Ac -PenTL and N Boc -PenTL as white crystals, as well as N C8 -PenTL, N ene -PenTL, and N EG4 -PenTL as colorless oils ( 1 H and 13 C NMR, high-resolution mass spectrometry, and X-ray diffraction in Figures S1-S16). Notably, the monomer synthesis is a simple and robust onepot process and can be easily scaled up to ten-gram scale per batch in the laboratory. The ROP of each substrate was then initiated by benzyl mercaptan and catalyzed by an organobase of suitable basicity (Scheme 1B). No ROP of N Ac -PenTL or N Boc -PenTL was observed (Table 1, entries 1 and 2), most likely because of their limited solubility (less than $90 mg/mL in tetrahydrofuran [THF]), which is lower than the equilibrium monomer concentration. We therefore focused our effort on the ROP of the three liquid monomers because they were substantially more soluble in common organic solvents or could be even executed for bulk polymerization. We measured the [M] eq of N ene -PenTL at various temperatures to draw the Van't Hoff plot (Figure S17). According to the linear regression, the enthalpy (DH P ) and entropy (DS P ) changes of the ROP were calculated to be 9.4 kJ mol 1 and 28.1 J mol 1 K 1 , respectively. This, in turn, gave a DG P of 1.0 kJ mol 1 (0.24 kcal mol 1 ) at 25 C and a ceiling temperature (T c ) of 61 C at an initial monomer concentration ([M] 0 ) of 1.0 M. To increase reaction efficiency, we conducted bulk polymerization of N ene -PenTL and N C8 -PenTL at room temperature and a feeding monomer/initiator ratio (M/I) of 100/1. To our gratification, the ROP of N ene -PenTL, catalyzed by a weak organobase, triethylamine (TEA, pK a DMSO = 9.0), 55 afforded the desired polymer product PN ene -PenTE ( 1 H NMR in Figure S18) with a considerably larger M n and narrower dispersity (entry 3, Table 1; M n = 19.4 kg/mol, Ð $ 1.10) than those of similar PTEs synthesized previously from CysTLs (M n $8.8 kg/mol, Ð $ 2.4) at the same M/I ratio. 50 Replacing TEA (1.0 equiv relative to initiator) with 1,8-diazabicyclo(5.4.0)undec-7ene (DBU, 0.1 equiv), a stronger base with a pK a DMSO of 12, 55 greatly accelerated the ROP reaction (Table 1, entries 4-7; Figure 1A) while preserving the controllability, as evidenced by the unimodal peaks in size-exclusion chromatography (SEC) analysis. Increasing the M/I ratio resulted in a corresponding linear elevation in the M n of PN ene -PenTE (Figure 1A). The DBU-catalyzed ROP of N ene -PenTL also demonstrated other typical features of controlled polymerization, such as the observation that the monomer conversion displayed a linear relationship with M n (Figure 1B). The ROP of N C8 -PenTL showed very similar controllability to that of N ene - S20) also with satisfactory control (Table 1, entry 12). Notably, all polymerizations became viscuous immediately after the addition of base and gelized eventually as a result of the high concentration. It is also noteworthy that the polymerizations needed to be carefully quenched before subsequent processing because of the reversibility. ## Facile Functionalization Next, we examined the chain-end group of the resulting PN ene -PenTE and its ability to undergo post-polymerization modification. Matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrum of benzyl mercaptan-initiated PN ene -PenTE 15 contained only one set of molecular ion peaks with a spacing of 347 Da between two adjacent peaks, which corresponded to the molar mass of the monomer (Figure 1C). Moreover, the end groups were unambiguously assigned to the initiating PhCH 2 S-group on the a end and free tertiary thiol on the u terminus (Figure 1C, plus Na + or K + ). When the ROP was quenched with a small molecular capping agent, such as iodoacetamide, MALDI-TOF analysis gave exclusively PN ene -PenTE bearing PhCH 2 S-and -CH 2 CONH 2 as the a and u end groups, respectively (Figure S21). Similarly, PN C8 -PenTE also gave well-defined chain end groups in the MALDI-TOF analysis (Figures S22 and S23). Moreover, PN ene -PenTE (u end capped) was found to withstand typical UV-triggered thiol-ene reactions, and the side chain alkenes were converted to long alkyls (Figure S24) or anionic sulfate (Figure S25) groups almost quantitatively. Together, these results indicate that not only the chain ends but also the side chain groups are easily tunable, allowing facile introduction of a variety of functionalities. ## Thermal and Mechanical Properties Next, we studied the thermal properties of PN C8 -PenTE (M n $70.6 kg/mol) via thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC). PN C8 -PenTE showed a 5%-weight-loss decomposition temperature (T d ) of $192 C regardless of the capping status at the u end (Figure S27). DSC depicted a weak glass transition at $50 C (T g ), and a large endotherm peaked at $100 C, which corresponds to crystal melting in the heating scan (Figure 2A). Upon cooling, an exothermic peak was observed at a temperature slightly lower than the melting temperature (T m ). Dilatometry testing suggested that the T g and T m of the same polymer were $45 C and $100 C (Figure S28), respectively, which agreed well with the DSC results. In the tensile test using dynamic mechanical analysis (DMA), PN C8 -PenTE showed a Young's modulus of 300 MPa at 30 C and a catastrophic facture before yielding with a strain of 2.8% (Figure 2B). Above T g , the Young's modulus reduced to 110 Mpa, and on the other hand, the breaking strain increased to 170% at 60 C ## Controlled and Complete Depolymerization To investigate the depolymerization of the u-end-uncapped PN C8 -PenTE, we tested all reactions in diluted CDCl 3 (initial polymer concentration = 5.0 mg/mL) by employing various bases and temperatures. When PN C8 -PenTE was mixed with 0.05 equiv DBU (relative to the number of polymer chains) at 65 C, the gradual regeneration of N C8 -PenTL was confirmed by 1 H NMR spectroscopy and SEC. The conversion of depolymerization exhibited an inverse linear relationship with the remaining M n of the polymer (Figure 3B). Moreover, only unimodal peaks were observed in the SEC analysis of the depolymerization of PN C8 -PenTE, implying that no oligomerization occurred (Figure 3C). Meanwhile, as shown in Figure S29, the DBU (0.5 equiv)mediated depolymerization of PN C8 -PenTE (degree of polymerization $ 40) at room temperature gave a linear correlation of the conversion as a function of time (zero order kinetics). All these data offer convincing evidence for a domino-like unzipping depolymerization process. However, [a] D testing of the 65 C recycled monomer indicated racemization (Table 2, entries 1 and 2). Interestingly, at reduced temperatures such as 25 C, PN C8 -PenTE 40 was completely depolymerized (>95%) into enantiopure monomers (Figure 3D; Table 2, entry 3) within 4 h, catalyzed by 0.1 equiv DBU. Thus, it appears that lower temperature could prevent racemization effectively. This notion was further confirmed by the fact that complete depolymerization without racemization of PN C8 -PenTE 40 was achievable within 10 min with 1 equiv DBU at 25 C (Table 2, entry 4). PN ene -PenTE showed a very similar result to PN C8 -PenTE in depolymerization (Table 2, entries 5 and 6; Figure S30). Of note, the u-end-capped PN ene -PenTE remained almost unchanged when mixed with 0.1 equiv DBU at 25 C for 12 h (Figure S31). The depolymerization could also be catalyzed by 1.0 equiv sodium thiophenolate (PhSNa), a weaker base but stronger nucleophile than DBU, which gave >95% conversion in less than 2 h at ambient temperature, again with no detectable racemization (Table 2, entry 5). More interestingly, even the u-end-capped PN C8 -PenTE could be completely depolymerized when treated with 1.0 equiv PhSNa at room temperature (Figure S32), implying a complementary (mechanistic) approach to the previously described DBU-catalyzed depolymerization (Figure S33). ## Article Mechanism of ROP and Depolymerization We further studied the ROP and depolymerization of N R -PenTL by density functional theory (DFT) calculation and molecular dynamics (MD) simulation. To simplify the calculation, we used N Ac -PenTL as a model monomer and considered the reactive chain end to consist of a dissociated anionic thiolate. The free energies of key intermediates (INTs) and transition states (TSs) in both the chain propagation and chain transfer are summarized in Figure 4A. The change in free energy of the ROP was calculated to be 0.8 kcal/mol (INT 1 to INT 2 ), agreeing well with the previous Van't Hoff plot (0.24 kcal/mol). The energy barrier for the chain propagation (INT 1 to TS 1 ) and chain transfer (INT 1 to 2TS 1 ) was 6.6 and 19.2 kcal/mol, respectively (Figure 4B). The $12.6 kcal/mol difference in energy barrier, according to the Eyring equation, suggested that the rate constants of the two pathways differed from each other by an order of magnitude of $6, which ensured that the ROP would proceed in a highly controlled fashion as we observed. Interestingly, the energy barrier of chain propagation and chain transfer in the ROP of N Ac -CysTE were substantially lower, 0.5 and 2.2 kcal/mol, respectively (Figure S34). Such low energy barriers suggest that the rate of both propagation and transfer reactions was fast and that it was difficult to minimize chain transfer. For depolymerization of PN Ac -PenTE, the energy barrier was 5.8 kcal/mol (INT 2 to TS 1 ; Figure 4B), which served as a key contributor to the amenability of PN R -PenTE to depolymerization. On the other hand, for N Ac -CysTE, the change in free energy was favored for the ROP (-14.4 kcal/mol), making the reverse depolymerization pathway highly unfavorable thermodynamically (C-INT 2 to C-TS 1 ; Figure S34). DFT calculation further suggested that in the low-energy conformation of PN R -PenTE, the terminal tertiary thiolate and the adjacent thioester took a gauche conformation with a S-C=O distance of $2.96 A ˚and a S-C-C-CO dihedral angle (J) of 66.4 (Figure 4C), which required no extra rotary energy for the ring-closing depolymerization. On the contrary, the low-energy conformation of N Ac -CysTE was the staggered conformation in which the same S-C=O distance was $4.17 A ˚(Figure 4C). All-atom MD simulation (Figures 4D ansd 4E) of the 20-mer of PN Me -PenTE also suggested that the most populated conformation (J = 67.7%) is characterized by an average S-C=O distance of 3.4 A ˚between the terminal thiol and its adjacent thioester carbon, whereas the staggered conformations with a longer distance (4.20 A ˚in average) were less populated (18.2%). This restrained conformation is a clear indication of the Thorpe-Ingold effect induced by the gem-DM. ## Article DISCUSSION The unique thermomechanical, optical, and dynamic properties of PTE polymers, coupled with a global emphasis on environmental sustainability, have propelled a resurgence in their popularity. 35,56,57 Although chemical production 47 and biosynthesis 58 of PTEs were first reported in 1968 and 2001, respectively, controlled chemical synthesis of high M n PTEs remains a technological bottleneck because of the dynamic nature of thioesters. In this work, we achieved the controlled ROP of b-thiolactones by introducing a gem-DM group on the four-membered ring (Figure 1; Table 1). The facile access to high M n PTEs under mild conditions could provide a significant boost to their industrial and biomedical application. ## Article Functionalization of both the termini and side groups in PN R -PenTE opens up opportunities to create novel, high-performance materials by introducing different combinations of substituents to the polymer chain. As demonstrated earlier, the N-octanoyloxy groups in PN C8 -PenTL confer semicrystallinity and increase processability and durability (Figure 2). On the other hand, PN EG4 -PenTE (Table 1, entry 12) exhibits excellent water solubility and degradability, making it an attractive alternative to PEG and a promising high-value biomaterial for temperature-induced self-assembly and/or therapeutic protein conjugation. 59,60 The physicochemical properties of PN R -PenTE could be further expanded or fine-tuned by copolymerizing several types of monomers with different side chains (Table 1, entry 11). Previously reported strategies for polymer recycling predominantly involved reverting back to cyclic monomers bearing a five-or six-membered ring, which are relatively unstrained. 17,19,22,31,61 We recently succeeded in extending the scope of such regenerative building blocks to bridged bicyclic thiolactones. 48 In the current study, we further demonstrated, on the basis of both experimental data and theoretical calculations, that the presence of gem-DM played a key role in enabling fast, selective, and highly controlled depolymerization of PN R -PenTEs (Figures 3 and 4). Of note, although depolymerizable polymers were not uncommon, the domino-like unzipping fashion in the depolymerization of PN R -PenTEs without the generation of thermodynamic equilibrium mixture of oligomeric intermediates (Figures 3A, 3B, and S29) was relatively rare. This control was again most likely empowered by the presence of gem-DM (Figures 4C-4E). It is also worth pointing out that similar ring-closure reactions that are driven by the Thorpe-Ingold effect have been frequently employed. Thus, we envisage that our current strategy can be broadly applied to the design of various recyclable polymers beyond PN R -PenTEs. The PhSNa-mediated complete depolymerization of both u-end-capped and -uncapped PN R -PenTEs allows more versatile end-group functionalization without jeopardizing the recycling capability of PTEs and thus opens new opportunities to a broader scope of applications. It should be noted that the relatively high price of penicillamine could pose a challenge to our method when applied on an industrial scale. Nevertheless, cost mitigation could potentially be achieved by further optimizing the chemical and/or biosynthetic-based routes to produce similar substrate monomers in a more affordable manner. 65 Overall, the strategy can be utilized to rapidly and efficiently generate a wide range of high-value, recyclable polymers with immense application potential as self-immolative materials, 66,67 covalent adaptable networks, 68 sacrificial domain for composites and nanolithography, and responsive biomaterials. 56 ## EXPERIMENTAL PROCEDURES Resource Availability Lead Contact Requests for further information should be directed to and will be fulfilled by the Lead Contact, Hua Lu (chemhualu@pku.edu.cn). ## Materials Availability This study did not generate new unique materials. ## Data and Code Availability The accession number for the crystallographic data reported in this paper is CCDC: 2004873. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre at https://www.ccdc.cam.ac.uk/structures/.

chemsum

{"title": "Geminal Dimethyl Substitution Enables Controlled Polymerization of Penicillamine-Derived \u03b2-Thiolactones and Reversed Depolymerization", "journal": "Chem Cell"}

high_nitrogen-containing_cotton_derived_3d_porous_carbon_frameworks_for_high-performance_supercapaci

## Abstract: Supercapacitors fabricated by 3D porous carbon frameworks, such as graphene-and carbon nanotube (CNT)-based aerogels, have been highly attractive due to their various advantages. However, their high cost along with insufficient yield has inhibited their large-scale applications. Here we have demonstrated a facile and easily scalable approach for large-scale preparing novel 3D nitrogencontaining porous carbon frameworks using ultralow-cost commercial cotton. Electrochemical performance suggests that the optimal nitrogen-containing cotton-derived carbon frameworks with a high nitrogen content (12.1 mol%) along with low surface area 285 m 2 g −1 present high specific capacities of the 308 and 200 F g −1 in KOH electrolyte at current densities of 0.1 and 10 A g −1 , respectively, with very limited capacitance loss upon 10,000 cycles in both aqueous and gel electrolytes. Moreover, the electrode exhibits the highest capacitance up to 220 F g −1 at 0.1 A g −1 and excellent flexibility (with negligible capacitance loss under different bending angles) in the polyvinyl alcohol/KOH gel electrolyte. The observed excellent performance competes well with that found in the electrodes of similar 3D frameworks formed by graphene or CNTs. Therefore, the ultralow-cost and simply strategy here demonstrates great potential for scalable producing high-performance carbon-based supercapacitors in the industry.The depletion of fossil fuels and the increasingly environmental pollution has required the scientific community to develop new class of clean and sustainable energy sources. The fast-growing market for portable electronic devices and hybrid electronic vehicles has also demanded innovative energy storage materials of high power density and efficient energy conversion. Due to the large amount of energy stored in a very short time and long-cycle life, supercapacitors have been projected to be the most common energy storage devices 1-3 , with particular advancements achieved in flexible all-solid-state supercapacitors in recent years [4][5][6][7][8] .Among the well-developed electrical double layer capacitors (EDLCs) and pseudo-capacitors, porous carbon nanostructure-based electrode materials have attracted increasing interests for supercapacitors owing to their high surface area, sufficient electrical conductivity, excellent chemical stability and low cost 9 . A remarkable variety of carbon sources with various nanostructures or unique morphologies have been widely explored, aiming to improve the electrochemical capacitance and power density. For such purpose, three-dimensional (3D) carbon porous nanostructures have been recently pursued because of the advantageous features of enhanced ion and electron transport, high specific capacities, superior electrochemical stability [10][11][12] . As typical 3D carbon nanostructures, carbon nanotube (CNT) aerogels and graphene aerogels (GAs) of high electrical conductivity, large surface area and interconnected porous structures have been largely studied . For examples, Robert and coworkers have demonstrated two-electrode supercapacitor cells using 3D CNT-based materials, showing an area specific capacitance of 1 mF cm −2 in 1 M LiPF 6 at a current of 10 μ A 18 . On the other hand, Duan and coworkers have fabricated flexible GA-based solid-state supercapacitors, which demonstrated enhanced specific capacitance up to 186 F g −1 with area specific capacitance of 372 F cm −2 (current density of 1 A g −1 ) in two-electrode polyvinylalcohol (PVA)/H 2 SO 4 gel system 19 . Further improvement includes simply employing heteroatoms into such 3D carbon frameworks, and thus additional pseudo-capacitance could be achieved in EDLCs. For instance, Müllen and coworkers have fabricated 3D nitrogen and boron co-doped graphene into all-solid-state supercapacitors, which showed high specific capacitance of 62 F g −1 at a scan rate of 5 mVs −1 in the two-electrode PVA/H 2 SO 4 system 20 . Moreover, the nitrogen-doped 3D graphene framework fabricated by Qu and coworkers exhibited high specific capacitance, approaching 484 F g −1 at a current density of 1 A g −1 in 1 M LiClO 4 aqueous solution (three-electrode system) 21 . Although significant progresses have been made in the EDLCs based on 3D graphene-and CNT-based porous frameworks, the critical issues associated with high cost and low yield of such novel carbon-based materials still limit their applications in the industry of energy storage devices 22 . It is noticed that great challenges remain in the exploration of sufficient and simple procedures for scalable preparing CNTs and graphene . As a consequence, the high price/cost of commercially available high-quality CNT and graphene-based materials (Table S1) only allows them to be used in the laboratories thus far. In the present work, we demonstrate a simple strategy for large-scale preparing aerogel-like 3D carbon frameworks using ultralow-cost commercial cotton (0.01 USD/g, much lower than the commercially available CNTs and graphene shown in Table S1). Since the aerogel-like 3D porous framework of the cotton has been well preserved in the convenient thermal treatment, the as-prepared cotton-derived carbon frameworks (CCFs) have delivered excellent cycle stability with high specific capacities of 182 F g −1 at the current density of 0.1 A g −1 in 6 mol L −1 KOH aqueous solution. Further electrochemical improvements have been achieved by introducing low-cost N-doping sources (urea and melamine), and the resulting N-doped cotton-derived carbon frameworks with subsequent acid treatment (NCCFs) presented much enhanced specific capacities of 308, 240 and 200 F g −1 at current densities of 0.1, 1 and 10 A g −1 , respectively. Moreover, the highest capacitance up to 220 F g −1 at 0.1 A g −1 and negligible capacitive loss were also observed in the flexible all-solid-state supercapacitors fabricated by NCCFs. Direct comparison indicates that such NCCFs of much lower cost offer very competitive electrochemical performance to the 3D graphene-and CNT-based frameworks in the literature. Implication of the results suggests that the use of extremely simple and easily scalable strategy with very cheap commercial cotton is highly promising for large-scale producing high-performance supercapacitors in the industry. ## Results In the typical preparation (Fig. 1), certain amount of commercial cotton was used as the raw materials and the corresponding size and amount could be easily scale-up according to the chamber of furnace for carbonization. For preparing CCFs, the cotton was directly heated up for carbonization under N 2 protection, and further treated under sonication in the mixed acid solution to improve the hydrophilicity via introducing carbonyl and hydroxyl groups 27 . NCCFs were obtained by the same conditions except for the presence of urea and melamine (N-doping sources) in the carbonization, allowing the nitrogen-containing radicals in the decomposition of melamine and urea to reacts with the carbon radicals upon high-temperature treatment. Figure 2a shows a large piece of NCCFs prepared in this work, indicating the easy scalability of the approach. Similar to the observation in the recently reported graphene and CNT-based 3D frameworks 28,29 , the as-prepared NCCFs show light weight (Fig. 2b) and mechanical robust with excellent flexibility (Fig. 2c), which has also been confirmed by the stress-strain results of the static tensile tests (Figure S1). Figure 3 exhibits typical scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of raw cotton (Fig. 3a-d) and NCCFs (Fig. 3e-h). According to the direct comparison of the morphologies, the raw cotton and NCCFs both present similar 3D frameworks, indicating that the well-formed 3D frameworks in the raw cotton has been well reserved in the high-temperature carbonization. As the basic structure for the 3D frameworks, the carbon stripes in the NCCFs exhibit much curved and twisted (Fig. 3f) in comparison with those in the raw cotton (Fig. 3b,c). Such morphological changes should be associated with the mass loss upon the high-temperature carbonization, leading to shrinkages in the curved cotton stripes as well. Representative SEM and TEM images suggest the wrinkles along with rough and porous surface of the CCFs (Figure S2) and NCCFs (Fig. 3g,h). The elemental mapping results also suggest the uniform elemental distribution of N and O in the NCCFs, due to the N-doping and acid treatments (Figure S3). According to Fig. 4, further details of the microstructures have been observed in the CCFs and NCCFs, suggesting no meaningful difference has been found in these two samples. The observed mesopores and micropores in both of the CCFs and NCCFs would provide effective porosity and improved surface area for charge storage. Figure 5 shows the N 2 adsorption/desorption isotherms of cotton-derived carbon materials. The pore parameters calculated from N 2 adsorption/desorption isotherms are summarized in Table 1. According to Fig. 5a, the N 2 adsorption/desorption isotherms of both samples exhibit type I isotherms with H4 type hysteresis loops. The rapid increase of the adsorption isotherms in the low pressure region (P/ P 0 = 0 ~ 0.1) and the hysteresis loops indicates a large amount of micropores and presence of mesopores, respectively (Fig. 5a) . The results of comparison show that CCFs possess slightly larger surface area (373 m 2 g −1 ) than NCCFs (285 m 2 g −1 ), which is attributed to the reduced micropores caused by the micropore collapse and merge during the acid treatment. The observed larger loops suggest increased meso/macropores surface area in the NCCFs (128 cm 2 g −1 ), compared to 88 cm 2 g −1 in the CCFs. These changes are mainly attributed to the surface functionalization during the carbonization process 34 . On the other hand, the distribution of average pore diameter was obtained by the density functional theory method. As exhibited in Fig. 5b, the pores size of CCFs was mainly around 0.5 nm, while decreased peaks of 0.5 nm coupled with pronounced peaks were observed around 1.4 nm, 1.7 nm and 2.7 nm in NCCFs. Such changes indicate that the surface functionalization has not only effectively enhanced the contents of heteroatoms, but also introduced pores in the 3D frameworks as well. Raman spectra (Figure S4) illustrate two prominent bands around 1360 and 1600 cm −1 , which are denoted as the D and G bands of carbon materials, respectively 27 . No significant difference was found in the ratios of I d /I g for CCFs (0.94) and NCCFs (0.99). X-ray photoelectron spectroscopy (XPS) was carried out to investigate the chemical composition of carbon materials (Fig. 6). As expected, pronounced C peaks have been observed in both samples (Fig. 6a), showing dominant carbon-carbon species (284.6 eV) (Figure S5). Compared to CCFs, enhanced C-N (286.33 eV) and C = O (287.9 eV) peaks have been found in NCCFs, which is attributed to the doping of nitrogen and acid treatment. It is noticeable that the substantial increase of nitrogen content in NCCFs (~12.1 mol%) listed in Table 1 suggests the sufficient N-doping with the presence of urea and melamine in the carbonization process, where nitrogen-containing precursors would react with the carbon sources to form C-N bonding under an inert atmosphere . As shown in Fig. 6b, the results exhibit different types of nitrogen-containing groups in the NCCFs, and the peaks at 400.6 eV (quaternary N (N-Q) species, 48.8%) and 398.1 eV (pyridine (N-6) species, 26.2%) 38,39 were generated with the presence of melamine and urea, respectively. The peak associated with nitro-type complexes NO 2 -(5.5%) at 406.5 eV should be associated with acid treatment. It is well known that the positively charged N-Q and nitrogen oxide (N-X) (401.8 eV, 7.8%) could improve electron transfer, resulting in enhanced capacitive performance under larger current densities. The negatively charged pyrrole or pyridine (N-5) and N-6, on the other hand, could offer pseudo-capacitive interactions, leading to further enhancement in the specific capacitance . The results of above characterizations imply that the as-prepared NCCFs present similar morphological and structural features to the typical 3D graphene-and CNT-based frameworks. Owing to such porous 3D carbon frameworks, they should be also ideal to serve as electrode materials for supercapacitors. First, the cotton-derived 3D carbon frameworks were fabricated into a three-electrode system (6 mol L −1 KOH aqueous electrolyte) for electrochemical measurements. Figure 7a shows the typical cyclic voltammetry (CV) curves of CCFs and NCCFs at 5 mV s −1 . The curves demonstrate quasi-rectangular shape with slight distortion, which is mainly induced by the pseudo-capacitance due to the oxygen and nitrogen functional groups. The redox signals of nitrogen-doped carbon electrodes are unobvious especially in alkaline electrolyte, which is in accordance with many reported works 39, . Apparently, NCCFs show much larger area than CCFs. As illustrated in Fig. 7b, the electrical resistance and ion transfer of the as-prepared supercapacitors were characterized on the electrochemical impedance spectroscopy (EIS) over a frequency range from 10 −2 to 10 5 Hz. According to the diameters of the semicircles at high frequency range, NCCFs exhibit lower resistance (~0.5 Ohm) than that of CCFs (~1.5 Ohm), which may be related to the promoted electron transfer with presence of N-Q and N-X. At lower frequency, all the samples present nearly vertical line, which suggests that the aerogel-like 3D porous frameworks facilitate the electrolyte transport in the cotton-derived carbon electrodes. Figure 7c exhibits the galvanostatic charge-discharge curves of NCCFs at different current densities. The almost isosceles triangles indicate an excellent double-layer capacitance with no obvious voltage drop (IR drop) observed, which is in good agreement with the low resistance in Fig. 7b. The specific capacitance was acquired according to the equation of C S = I × t/V/m, where C s (F g −1 ) is the specific capacitance, I (A g −1 ) the response current density, t (s) the discharge time, V (V) the potential and m (g) the mass of active material. As demonstrated in Fig. 7d, the specific capacities of the NCCFs present the highest capacities, approaching 308 F g −1 and 200 F g −1 at current densities of 0.1 A g −1 and 10 A g −1 , respectively. As exhibited in Fig. 7e, NCCFs shows a high retention up to 97% after 10,000 cycles at a current density of 5 A g −1 , suggesting excellent cycle stability. It is noticed that the NCCFs with surface area only up to 285 m 2 g −1 show very high capacities and excellent retention in the cycles, which should be mainly due to the high-concentration doped nitrogen (12.1 mol%) that not only introduces sufficient pseudocapacitance but also improves the electronic structure of graphitic carbon as well 45 A symmetrical flexible all-solid-state supercapacitor was further fabricated using the as-prepared electrodes coupled with PVA/KOH gel electrolyte (Fig. 1) 46 . The electrochemical properties of the all-solid-state supercapacitors are exhibited in Fig. 8. Figure 8a exhibits the CV curves at different current densities, and no pronounced redox peaks have been observed, suggesting typical double-layer capacitive behavior. The impedance in Fig. 8b shows only slight increase of impedance (~2 Ohm) in all-solid-state supercapacitors compared to the values in the three-electrode system. However, the vertical line in the low frequency region also indicates the fast ion transport in the PVA/KOH electrolyte (Fig. 8b). According to the galvanostatic charge/discharge curves in Fig. 8c, similar isosceles triangles were also observed except for the presence of limited IR drop. The specific capacitance of the electrodes in a two-electrode supercapacitor was achieved by the formula that C s = 4I × t/V/m, where m is the mass of the active material in both electrodes. Figure 8d displays the rate stability of the all-solid-state supercapacitor, showing the highest capacitance up to 220 F g −1 . After 10,000 cycles at a current density of 5 A g −1 , the all-solid-state supercapacitor presented excellent cycle ability with 98% retention (Fig. 8e). In order to measure the capacitance retention under stress, the flexible supercapacitors were bent with different angles and the results show negligible effects on the capacitance upon bending (Fig. 8f). ## Discussion The results demonstrate that the cotton-derived 3D carbon frameworks present graphene aerogel-like morphologies and structures, which allow the electrolyte to deliver fast ion transport in the porous frameworks and meanwhile enable carbon interconnected networks to serve as the effective pathways for electron transport. On the basis of the sufficient surface area of the conductive porous carbon frameworks, further enhanced capacitance could be easily achieved by the introduced N-doped functional groups. In a typical work by Ruoff and coworkers, it is suggested that the increasing N content (0.7~2.3 wt% N) would contribute to greater enhancement in the capacitive capability of the N-doped graphite oxide (Table S2) 42 . Similar results in the work by Chen and coworkers also suggest the carbon with 11.89% N content substantially enhanced the specific capacitance 43 . Therefore, such NCCFs with highly concentrated N-containing groups (12.1 wt% N) are expected to exhibit promising electrochemical performance. In the three-electrode configuration, it is interesting to find that the capacitance in NCCFs is well competitive to that found in the graphene-and CNT-based 3D frameworks (Table S2). According to the price listed in Table S1 and direct comparison of typical performance of other carbon-based electrodes, the results suggest that such low-cost 3D carbon sources with the facile strategy provide sufficient capacitance similar to the fashion carbon materials of high expense. Unlike relatively high surface areas based on the presence of CNTs and graphene in the aerogel-like structures, NCCFs here based on simply doped with high-concentration N-functional groups, which are responsible for both enhancing pseudocapacitance and facilitating electron transfer in the NCCF frameworks, have exhibited the comparable charge storage to those based on large surface areas and sufficient pore sizes. It may be argued that the NCCFs should be directly fabricated into the binder-free electrodes for taking their advantages. In comparison, the binder-free electrode was fabricated via directly compressing the NCCF onto the Ni foam. In the same three-electrode system of 6 M KOH aqueous electrolyte, the electrochemical performance of the binder-free electrode was 173, 124 and 52 F g −1 at current densities ## Pore parameters Compositions S total S micro (m 2 g −1 ) S meso V total V micro (cm 3 g −1 ) V Total surface area (S total ) and surface area of micropores (S micro ) were obtained from multi-point Brumauer-Emmett-Teller plot and V-t plot, respectively. Surface area of meso/macropores (S meso ) was calculated by subtracting S micro from S total . of 0.1, 1 and 5 A g −1 , respectively (Figure S6). Apparently, the binder-free systems were far behind the binder systems (308 and 200 F g −1 at 0.1 and 10 A g −1 , respectively). Therefore, the binder systems can deliver more promising performance for electrochemical storage. This observation is highly usual in the biomass-based active materials 31,32,45 , which are general required to be mixed up with binders and conductive agents to fully realize the optimal electrochemical performance. Furthermore, the mass loading of the binder electrode could reach ~9 mg/cm 2 , which is more close to the practical supercapacitors. However, the mass loading of the binder-free electrode was only up to 4 mg/cm 2 due to the highly porous framework, which may not meet the practical applications. Although the mixing process with binder and conductive agent might change the framework morphology, it is believed to be the optimized approach to fully use the capacitive capability. Apparently, there are more opportunities for further improving the device performance because related enhancements including optimization of cotton carbonization for tuning the sp2 carbon, enlargement of the surface area and porosity or adjustment of the heteroatomic doping (N-and B-doing for examples) could be performed on the current stage, by which the fundamental understanding of the impacts on the device performance would be also achieved. Compared to the mostly attractive graphene and CNT-based 3D frameworks, the cotton-derived configurations appear to be more promising in the large-scale applications because of the much lower cost and easier scalability. Therefore, the commercially available cotton coupled with simply strategies has shown a new stage for scalable fabrication of high-performance supercapacitors. ## Conclusions In conclusion, a facile and low-cost approach was demonstrated for scalable fabricating high-performance electrodes for supercapacitors. The commercially available cotton was used as the carbon source for preparing aerogel-like 3D nanostructures, which are favorable for both the ion and electron transport. Therefore, the resulting N-doped cotton-derived frameworks present effective capacitance, highly competitive to the performance observed in the other graphene-and CNT-based 3D frameworks. The strategy demonstrated here is very simply, low cost, easy scale-up and efficient, showing great potential for wide applications in the energy storage industry. ## Methods Synthesis of CCFs and NCCFs. The commercial cotton was directly used as the starting material without any further pre-treatment. Typically, a piece of cotton was cut into a certain shape (determined by the furnace size) and mixed with melamine and urea with mass ratio of 1:2:2. Subsequently, the mixture was subjected to carbonization at 800 o C for 1 h with a heating rate of 5 o C min −1 under N 2 atmosphere. The resulting samples were further treated under ultrasonication for 1h with the presence of mixture of HNO 3 /H 2 SO 4 (v/v = 1/3, 70% HNO 3 and 98% H 2 SO 4 ). The resulted samples (NCCFs) were then washed with distilled water and dried at 90 o C overnight. For comparison, the reference samples (CCFs) were prepared under the same conditions except for the absence of melamine and urea. Field-emission scanning electron microscopy (FESEM) was conducted on ZEISS supra 55 system. Transmission electron microscope (TEM) was performed on JEOL JEM-2010. Raman spectra were carried out on HR800 (Horiba JobinYvon) with a 514.5 nm Ar-ion laser. The nitrogen absorption/desorption isotherms associated with specific surface area and pore diameter distribution data were investigated on an Autosorb-iQ2-MP (Quantachrome) analyzer under 77k. X-ray photoelectron spectroscopy (XPS) was acquired on PHI-5300. Static tensile tests were performed with a mechanical analyzer (TA-XT Plus system, SMS). Characterization. The electrochemical performance of the carbon materials was determined in a three-electrode cell with basic aqueous solutions. The working electrode was prepared by mixing the carbon samples, acetylene black with polytetrafluoroethylene in a weight ratio of 80:15:5, and then the mixture was pressed onto a nickel foam. The typical mass and dimensions of the working electrodes are 10 mg and 1 cm 2 . For the three-electrode system, Pt and Hg/HgO electrode were used as the counter electrode and reference electrode, respectively, while the KOH solution (6 mol L −1 ) was employed as the electrolyte. Electrochemical evaluation. The electrochemical performance of NCCFs was further determined in a gel electrolyte by using a two-electrode cell. Similar to the three-electrode system, the working electrode for two-electrode system was also fabricated with the carbon samples, acetylene black and polytetrafluoroethylene (80:15:5 in weight ratio), followed by being pressed onto a nickel foam. For the two-electrode system, an all-solid-state supercapacitor was fabricated as depicted in Fig. 1c. In a typical preparation, PVA powder (2 g) was dissolved in distilled water (20 mL) under vigorous stirring at 85 o C, followed by immersing two pieces of NCCFs electrodes in the PVA solution. Then, excessive KOH solution (6 M, 200 mL) was added and the mixture was kept for 24 h to form all-solid-state supercapacitors, where the PVA/KOH gel served as both electrolytes and separators. Galvanostatic charge/discharge was tested at various current densities using LAND-CT2001A (Wuhan Jinnuo Electronics. Ltd.). Cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were carried out using a CHI660C electrochemical workstation (CH Instruments, Inc.). 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": "High nitrogen-containing cotton derived 3D porous carbon frameworks for high-performance supercapacitors", "journal": "Scientific Reports - Nature"}

a_highly_soluble,_crystalline_covalent_organic_framework_compatible_with_device_implementation

## Abstract: Covalent organic frameworks (COFs) have emerged as a tailor-made platform for designing nextgeneration two-dimensional materials. However, COFs are produced as insoluble and unprocessable solids, which precludes the preparation of thin films for optoelectronic applications. Here, we report designed synthesis of a highly soluble yet crystalline COF material through the regulation of its interlayer interactions. The resulting COF is remarkably soluble in a variety of organic solvents and forms stable true solutions with retention of its layered structure. These unique features endow the COF with solution processability; high-quality, large-area COF films can be produced on various substrates in a high-throughput and efficient manner, with good control over the film thickness, making this material compatible with a variety of device applications. The films are electrically anisotropic; the intra-layer carrier conduction is inhibited, while the inter-layer carrier migration is outstanding, showing the highest conductivity among all reported COF materials. Our highly soluble and processable COF may open new pathways for realising high-performance COF-based optoelectronic devices with diverse functions. ## Introduction Covalent organic frameworks (COFs) are a new class of crystalline porous materials that allow the atomically precise integration of desirable p units to create predesigned skeletons and nanopores, and their unique features of well-defned crystalline porous structure, high stability together with tailored functionalities make them promising materials for many applications. Recently, signifcant progress has been made in the structural diversity and complexity of COFs for functional exploration. Nevertheless, a large number of challenges remain in the COF feld including their poor processability. COFs are inherently crosslinked and are produced as insoluble and unprocessable powders. The limited utility of this form prevents the use of COFs in many applications. For example, two-dimensional (2D) COFs are ideally suited for electronic devices by virtue of their well-defned lattice with p functionality. 12,13 However, they are difficult to form highquality thin flms, and cannot be reliably interfaced onto electrodes. Thus COF materials-based optoelectronic devices usually exhibit inferior performance. Monolayer investigation and flm preparation are highlighted as bottleneck issues of "critical importance" in recent reviews of the COF feld. 1 Currently, there are several approaches to prepare COF nanosheets or flms. A typically used method is substrate-assisted deposition, 14,15 in which flm-like COFs are deposited onto certain substrates during their organic synthesis. Nevertheless, this method is not a highthroughput one as the majority of COFs remain as powders in solutions. A few physical approaches, such as ultrasonication, 16 ball milling, 17,18 and mechanical delamination, 19,20 have been utilised to exfoliate COF powders into nanosheets. However, these physical methods usually break the COFs into small pieces, decreasing the quality of the materials. A facile synthetic method for the preparation of COF flms is interfacial polymerisation in which the condensation reaction occurs at the liquid-liquid interface, 21,22 allowing the formation of flm-like COF materials; unfortunately, this method is difficult to scale up. A newly developed method to prepare colloidal COF suspensions is to add acetonitrile as a stabilizer. 23,24 However, this strategy was only appropriate for boronate ester-linked COFs, which were less useful for optoelectronic devices due to the stability issues of these materials. Recently, a bulky synthesis of ionic COFs was developed via the incorporation of cationic or anionic species into the frameworks. Several ionic COFs could be exfoliated into nanosheets upon mixing with solvents, 25,26 but these COFs are not really soluble in solvents and show rather limited solubility and small sizes of the exfoliated nanosheets; the ionic COF suspensions could not be employed for flm preparation by solution-processing methods. The development of a methodology for the direct synthesis of solution-processable COFs remains a substantial challenge but is of great importance for the further advancement of this feld. Herein, we report a highly soluble yet crystalline COF material by encoding high-density electrostatic repulsion into the COF skeleton to withstand the inter-layer p-p stacking. The resulting COF features obvious crystallinity, while it could readily dissolve in various organic solvents to form true solutions and self-exfoliate into large-area monolayer or multilayer nanosheets. Importantly, the high solubility of this COF allows us to simply prepare high-quality COF flms, which allows the investigation of its intra-layer and inter-layer properties, such as its electronic properties. In this work, we reveal for the frst time the high, yet anisotropic, conductivity of the COF flms. ## Results and discussion We chose two building blocks featuring opposing properties affecting the crystallinity of the COFs. The knot monomer 4,4 0 ,4 00 ,4 000 -(pyrene-1,3,6,8-tetrayl)tetraaniline (Py) exhibits strong p-p interactions and tends to form crystalline stacks, while the linker monomer 1,1-bis(4-formylphenyl)-4,4 0 -bipyridinium dichloride (Vg 2+ $2Cl ), with its inherently high charge density, provides strong electrostatic repulsion and hampers the stacking of COFs. By integrating both of these factors into the COF skeleton, the PyVg-COF (Fig. 1a) was synthesised via reticular polycondensation of Py with Vg 2+ $2Cl under solvothermal conditions. The optimal reaction conditions for the synthesis of the PyVg-COF were N,N-dimethylacetamide (DMAc)/mesitylene/6 M AcOH (1/ 9/1 v/v) at 120 C for 7 days, and the desired product was obtained as a brownish-red solid in a 38% isolated yield. The PyVg-COF was unambiguously characterised by various analytical methods (Fig. S1-S5 and Table S1 †). The crystal structure of the PyVg-COF was determined by using synchrotron powder X-ray diffraction (PXRD) measurements in conjunction with structural simulations. The synchrotron PXRD pattern of the PyVg-COF revealed an obvious peak at 2q ¼ 2.54 and several small peaks centred at 2q ¼ 1.55 , 1.95 , 4.99 , 7.20 , 8.07 and 14.34 (Fig. 2). To determine the optimal crystal structures, crystal models were generated based on the geometry of the building blocks. The optimal cell dimensions were a ¼ 56.16 , b ¼ 47.87 , c ¼ 6.48 , and a ¼ ß ¼ g ¼ 90 (Table S2 †). Using the structures of the monolayers, two stacked confgurations, that is, eclipsed AA and staggered AB modes, were generated and optimised (Fig. 1b and c and Tables S3 and S4 †). The simulated PXRD pattern of the staggered AB model was in good agreement with the experimental pattern and all the peaks in the experimental pattern could be well indexed, which confrmed that the PyVg-COF adopted a staggered AB stacking arrangement. The porosity of the PyVg-COF was characterised by N 2 and CO 2 adsorption isotherms (Fig. S6a †) from which the Brunauer-Emmett-Teller (BET) surface area of the PyVg-COF was calculated to be 348 m 2 g 1 from the CO 2 adsorption isotherm curve at 273.15 K, and the pore sizes were determined to be 1.20 and 1.56 nm, respectively (Fig. S6b †). The pore size distribution corresponded well with the two pore sizes generated by AB stacking. These results demonstrated that the PyVg-COF is highly porous with a large surface area. Surprisingly, while the PyVg-COF adsorbed very little N 2 at 77.35 K and 273.15 K, it adsorbed a substantial amount of CO 2 (196 mg g 1 ) at 273.15 K (Fig. S6a †), which is among the highest performances of COFs reported to date, and is also comparable to those of other top-class members. The CO 2 /N 2 selectivity calculated from the ideal absorbed solution theory (IAST) reached as high as 96 at a relative pressure of 0.2 (Fig. S6c †). In the PyVg-COF, the staggered structure reduced the pore size, and the strong polarity of the pores caused them to selectively adsorb CO 2 and exclude nonpolar N 2 . At 298.15 K, the PyVg-COF adsorbed less CO 2 (134 mg g 1 ). To clarify the nature of CO 2 adsorption, the isosteric heat of adsorption (Q st ) was calculated using the Clausius-Clapeyron equation from the CO 2 adsorption isotherms measured at 273.15 and 298.15 K (Fig. S6d †). The Q st value was as high as 36.1 kJ mol 1 , which further confrmed the strong interactions between the framework and CO 2 . Thus, by virtue of its highly polar pores, which form strong interactions with polar gas molecules, the PyVg-COF adsorbs a signifcant amount of CO 2 and a negligible amount of N 2 , making it an outstanding CO 2 adsorbent. Remarkably, by simple manual shaking, the PyVg-COF readily dissolves in various organic solvents (ESI Video †), such as N-methyl pyrrolidone (NMP), dimethyl sulfoxide (DMSO), N,N-dimethylformamide (DMF), N,N-diethylformamide (DEF), DMAc, and 1,3-dimethyl-2-imidazolidinone (DMI) (Fig. 3a). Among these solvents, the PyVg-COF was most soluble in DEF (8.3 mg mL 1 ), and it was least soluble in DMAc (2.1 mg mL 1 ) (Fig. S7 and Table S5 †). Such high solubility was in sharp contrast to traditional COFs which were not soluble. The concentrated PyVg-COF solutions were rather stable, and no sediment formed even when the solutions were left under ambient conditions for more than six months, in sharp contrast to traditional COF dispersions, which show low stability and a rapid settling process. To gain insight into the solubilisation mechanism, we calculated the inter-layer interactions and the interactions of the COF skeletons and the solvents (Table S6 †). For PyVg-COF, the inter-layer interaction was as low as 2.22 eV, while the interactions of COF skeletons and THF and water solvents were 1.36 and 1.95 eV, respectively. Since the solubility is the consequence of the difference value of the skeleton-solvent interactions and the inter-layer interactions, it is comprehensible that the PyVg-COF could not dissolve in these two solvents. In sharp contrast, the skeleton-solvent interactions for DMAc, NMP, DMF and DEF were as high as 5.74, 4.86, 4.87 and 5.12 eV, which were much higher than the inter-layer interaction in terms of the absolute value. Thus, PyVg-COF could readily dissolve in the above solvents, while the solubility enhancement for the four solvents also matched well with the experimental trend. As a control, we also calculated the skeleton-solvent interactions and the inter-layer interaction for a neutral COF which has almost the same structure but in AA stack mode (Table S6 †). In this case, the inter-layer interaction was too large compared with the skeleton-solvent interactions to allow the COF to be soluble. These results clearly revealed that the essence of the solubility was the weak inter-layer interactions and remarkable skeleton-solvent interactions. After dissolving, the PyVg-COF could self-exfoliate into monolayers or multilayers that were well dispersed in solution and exhibited a clear Tyndall effect (Fig. 3a). Since the fluorescence of the Py monomer was quenched upon aggregation, we could determine the critical aggregation concentration (CAC) of the PyVg-COF solution by taking advantage of the concentrationdependent fluorescence quenching, which revealed that the CAC of the PyVg-COF solution was 0.03 mg mL 1 , below and above which monolayers and multilayers were formed, respectively (Fig. S8 †). To investigate the aggregation behaviour of the PyVg-COF in solution, we conducted small-angle neutron scattering (SANS) measurements of a concentrated solution of the PyVg-COF in DMSO-d 6 (Fig. 3b). By virtue of the nature of neutron scattering, a large difference of scattering length density, namely, high contrast, between deuterated solvent and nanosheets could be realised. Therefore, SANS techniques can provide more convincing scattering data for these COF materials compared to small angle X-ray scattering methods. The data were well ftted with a typical 2D planar model in which nanosheets with sizes larger than 1 mm could be confrmed from data ftting in the low Q range (0.001 to 0.02 1 ). The SANS data ftted in the high Q range (0.03 to 0.4 1 ), especially considering the match of the frst oscillation peak at ca. 0.06 1 , suggested that the thicknesses of the nanosheets are approximately 10 to 15 nm. These results confrm the large sizes of PyVg-COF and the formation of stacking aggregates of the monolayer nanosheets at high concentration. It can be concluded from the SANS studies that the PyVg-COF could dissolve in solvents and form true solutions, while it retained the 2D skeleton and formed aggregated structures with thickness of nanometre precision. The structure of the PyVg-COF nanosheets was observed by high-resolution transmission electron microscopy (HR-TEM). The TEM image clearly revealed large-area nanosheets with size up to 10 mm (Fig. 3c and S9 †). TEM of the nanosheets also allowed the direct visualisation of the ordered rhombus polygon textures (Fig. S10 †). Notably, the selected-area electron diffraction (SAED) pattern showed a highly ordered arrangement with independent diffraction points, which could only be observed in the case of highly crystalline materials, thus further con-frming the high intra-layer crystallinity of the PyVg-COF (Fig. 3d). Therefore, the present PyVg-COF simultaneously possesses two distinct features, namely, crystallinity and solubility, which are highly desired for solution-processable 2D polymers. The atomic force microscopy (AFM) images revealed that the PyVg-COF nanosheets possessed smooth morphologies with heights ranging from 0.65 to 1.5 nm (Fig. S11 †), indicative of monolayer and multilayer formations. The high solubility of the PyVg-COF allowed us to prepare its flms in a solution-processed manner. The simplest method involves drop-casting the PyVg-COF solution onto substrates, followed by thermal evaporation of the solvents. Scanning electron microscopy (SEM) images revealed continuous flms with smooth morphology (Fig. S12 †). Grazing incidence wide-angle X-ray scattering (GIWAXS) of the flms indicated the "face-on" orientation of the COF skeletons parallel to the substrate (Fig. S13 †). This facile method allows the preparation of thin flms on conductive or insulating substrates and can be used for the fabrication of optoelectronic devices with various confgurations (vide infra). Upon further evaluation of the structural features of the PyVg-COF, we developed the frst electrophoretic deposition (EPD) method for preparing ionic COF flms by taking advantage of its highly charged skeleton (Fig. 4a). An EPD method for preparing metal-organic framework (MOF) flms has been reported; 38,39 however, unstable MOF suspensions that competitively form sediments must be used during the EPD process. By virtue of the high solubility and stable dispersibility of the PyVg-COF, EPD of the present COF system reproducibly provides high-quality flms. We set up a two-electrode EPD system (Fig. S14 †) in which various conductive substrates, such as indium tin oxide (ITO), gold and stainless steel, could serve as the working and counter electrodes. Under the optimal reaction conditions, DMI was used as the solvent, and the PyVg-COF concentration was 4.9 mg mL 1 . EPD was conducted by applying a constant DC voltage. As the EPD proceeded, the working electrode connected to the cathode of the power supply gradually turned brown, as identifed by the naked eye. After the EPD process, the PyVg-COF thin flms were transparent and brown, while the thick flms appeared almost black (Fig. 4b). Infrared spectra of the PyVg-COF flms revealed that the structures were stable during the EPD process (Fig. S15 †). SEM images of the flms showed that their morphologies were substrate-dependent (Fig. S16 †). The EPD flms retained their solubility, porosity and crystallinity (Fig. S17-S19 †). A distinct feature of EPD is the ability to produce COF flms with good control over their thickness by regulating the electric feld intensity and the EPD time. We prepared flms at various electric feld intensities and EPD times and measured the flm thicknesses. For instance, at an EPD time of 20 s, the thickness of the flm was linearly dependent on the electric feld intensity, and the slope of these data showed that the thickness increased by 1.4 nm/(V cm 1 ) (Fig. 4c). On the other hand, the thickness of the EPD flms could also be regulated by controlling the EPD time under constant electric feld intensity (e.g., 33.3 V cm 1 ), and the thickness increased linearly by 1.1 nm s 1 (Fig. 4d). Notably, the developed EPD method was a rapid and highthroughput method for flm preparation; for instance, the production of a 70 nm-thick flm required only 20 s even with a low electric feld intensity (33.3 V cm 1 ) (Fig. S20 †). The tunability of the flm thickness determines the compatibility with various device applications. The flms were only deposited on the electrodes and had the same size and shape as the employed electrodes. Thus, the present EPD method produces high-quality flms, with synthetically controlled thickness, size, and shape. The solution processability of the PyVg-COF provides an opportunity to explore some of its basic optoelectronic properties, such as electrical conductivity, which is important for COF materials, but there is lack of reliable investigations. The most common method for measuring this parameter is to use pressed pellets made by compressing COF powders. However, this method has large grain-boundary resistance and can be poorly reproducible. 40 Furthermore, the random orientation of crystallites in pressed pellets implies that, for materials exhibiting anisotropic conduction, the observed bulk conductivity of the pellet is a weighted average of the electrical conductivity in each crystallographic direction. To measure the intra-layer (horizontal direction) and inter-layer (vertical direction) conductivity, we prepared the PyVg-COF flms by drop-casting the COF solution onto ITO and glass substrates, respectively. The horizontal and vertical conductivities were measured in two types of device structures in a glovebox at 25 C (Fig. S21 †). The vertical conductivity of the COF flm exhibited a linear current-voltage (I-V) profle, indicative of ohmic conduction (Fig. 5a, red curve). The slope yielded an exceptional conductivity of 0.4 S m 1 , which is the highest value among the reported conductive COFs (Table S6 †). Notably, the reported conductive COFs in their assynthesised states are usually insulators, and their conductivity could be improved just by oxidation with iodine or doping with redox-active molecules, which occupy the pore space and cause the doped COFs to be nonporous (Table S6 †). In comparison, the PyVg-COF flms in the as-synthesised states exhibited inherently high vertical conductivity, and the pores were not occupied by oxidants or redox-active molecules; thus, the COF flms described herein are more useful for further functional exploration. In sharp contrast to the vertical measurements, the horizontal measurements of the COF flms showed negligible current with a conductivity of only 1.8 10 10 S m 1 (Fig. 5a, black curve). To verify these results, we used a four-contact probe method to measure the horizontal conductivity. 40 Unfortunately, the current was below the detection limit of the machine. To clarify the origin of the insulating behaviour in the horizontal direction, we utilised a two-contact probe method to measure the conductivity of several single COF nanosheets with large sizes and different thicknesses that had been prepared by drop-casting from a dilute solution (ESI Fig. S22 †). 40 Again, no current was observed even over a much wider voltage range. This result indicated that the insulating behaviour of the PyVg-COF in the horizontal direction is an intrinsic property rather than arising from grain-boundary resistance or edge defects in the nanosheets. All these results confrmed the electrically anisotropic nature of the PyVg-COF; holes or electrons could hop through the individual inter-layer donor or acceptor p-columnar arrays, while intra-layer conduction is inhibited because of charge recombination in the interlaced donor-acceptor networks. In principle, 2D COF materials should be anisotropic because they are crosslinked through covalent bonds in two dimensions and stack with each layer via p-p interactions in the third dimension. However, to date, the anisotropic property has not been confrmed experimentally due to the poor processability of COF powders, although theoretical simulations have previously predicted this property. 41 By virtue of its solution processability, the present COF system allows the anisotropic properties to be reliably measured and could thus serve as a model for investigating other basic behaviours of 2D porous materials. To gain insight into the anisotropic conductivity, we measured the carrier mobility of the COF flms. The horizontal mobility was measured by constructing a feld-effect transistor in which a COF flm acted as the active layer. However, no reliable current was observed, indicating that intra-layer carrier transmission was inhibited. On the other hand, the vertical mobility was measured by fabricating a sandwich-type diode in which a COF flm served as the active layer via a time-of-flight (TOF) method, which is widely recognised as the most accurate method to determine the carrier mobility. Upon increasing the electric feld intensity, the transient time of the holes increased signifcantly, while that of the electrons increased slightly (Fig. S23 and S24 †). Thus, both the hole and electron mobilities exhibited linear yet negative dependencies on the logarithm of the electric feld intensity, and the hole mobility drastically decreased when the electric feld intensity was increased (Fig. 5b). These results suggested that the built-in electric feld of the COF flms had negative effects on carrier migration, and hole transport was more likely to be hindered. We also calculated the intrinsic carrier mobility at zero electric feld (m 0 ), which was 9.0 10 4 and 1.4 10 5 cm 2 V 1 s 1 for holes and electrons, respectively. These results revealed that the COF flms presented high yet balanced carrier mobility, which was the essence of the exceptional vertical conductivity. ## Conclusions In summary, our fndings represent a new strategy for the direct preparation of COF solutions and anisotropic flms. The design principle of incorporating crystalline and ionic building blocks affords the resulting COFs with simultaneous crystallinity and remarkable solubility, enables solution processability of COFs into high-quality large-area flms in a high-throughput and efficient manner, and allows control over the flm thickness. For the frst time, the anisotropic electric properties of COF materials were determined. Our conductivity results are encouraging and will inspire further structural explorations. We anticipate that solution-processable COFs and their semiconducting flms will fnd a wide variety of applications in transistors, diodes, photocatalysis and energy storage. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "A highly soluble, crystalline covalent organic framework compatible with device implementation", "journal": "Royal Society of Chemistry (RSC)"}

d–a–d-type_orange-light_emitting_thermally_activated_delayed_fluorescence_(tadf)_materials_based_on_a

## Abstract: The design of orange-light emitting, thermally activated, delayed fluorescence (TADF) materials is necessary and important for the development and application of organic light-emitting diodes (OLEDs). Herein, two donor-acceptor-donor (D-A-D)-type orange TADF materials based on fluorenone and acridine, namely 2,7-bis(9,9-dimethylacridin-10(9H)-yl)-9H-fluoren-9-one (27DACRFT, 1) and 3,6-bis(9,9-dimethylacridin-10(9H)-yl)-9H-fluoren-9-one (36DACRFT, 2), were successfully synthetized and characterized. The studies on their structure-property relationship show that the different configurations have a serious effect on the photoluminescence and electroluminescence performance according to the change in singlet-triplet splitting energy (ΔE ST ) and excited state geometry. This indicates that a better configuration design can reduce internal conversion and improve triplet exciton utilization of TADF materials. Importantly, OLEDs based on 2 exhibited a maximum external quantum efficiency of 8.9%, which is higher than the theoretical efficiency of the OLEDs based on conventional fluorescent materials. ## Introduction Since multilayered OLEDs were first reported by Tang in 1987 , organic light-emitting diodes (OLEDs) have been a research focus due to their applications in display devices and general lighting. The efficiency of OLEDs was previously limited by the statistic rule of spin multiplicity. For conventional fluorescent materials, only singlet excitons are involved in electroluminescence, leading to a theoretical maximal internal quantum efficiency (IQE max ) of 25% and a theoretical maximal external quantum efficiency (EQE max ) of 5%, when assuming the out-coupling efficiency to be 20%. On the other hand, phosphorescent materials could utilize triplet excitons in electroluminescence processes to achieve 100% IQE max . However, the utilization of metals like iridium and platinum, which are expensive and nonrenewable, inevitably increase the cost of the final OLEDs. Alternatively, a thermally activated delayed fluorescence (TADF) material is a kind of noble-metalfree fluorescent material able to transform triplet excitons into singlet excitons through reverse intersystem crossing (RISC) to achieve 100% IQE max in theory . On the basis of the previous considerations, for TADF materials, the energy difference (ΔE ST ) between the first singlet excited state (S 1 ) and the first triplet excited state (T 1 ) must be small enough to enable the RISC process with the activation of environmental thermal energy . To achieve this, electron donors (D) and electron acceptors (A) are introduced into the molecule to form an intramolecular charge transfer (ICT) state with a large twisting angle between the donor and the acceptor to achieve the separation of highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) , which is the key to reduce the ΔE ST . Therefore, D-A-type or D-A-D-type molecules are the most classical TADF molecular structures . Although there have been numerous TADF materials synthesized and reported , to the best of our knowledge, orange and red TADF materials are still rarely reported in comparison with blue and green TADF materials . It is difficult to achieve TADF in orange and red fluorescent materials not only because red TADF materials require a strong ICT state, which strongly facilitates nonradiative transition processes, but also because the energy gap law generally results in a low radiative rate constant (k r ) to compete with a large nonradiative rate constant (k nr ) . The increasing nonradiative transition processes and large k nr play a role in competition with RISC and radiative transition processes and seriously restrict the development of orange and red TADF materials . Therefore, further attempts and new designs towards orange and red TADF materials are necessary. In this work, we designed and synthetized two novel D-A-Dtype orange TADF materials, namely 2,7-bis(9,9-dimethylacridin-10(9H)-yl)-9H-fluoren-9-one (27DACRFT, 1) and 3,6bis(9,9-dimethylacridin-10(9H)-yl)-9H-fluoren-9-one (36DACRFT, 2, Scheme 1). The compounds are isomers with different donor-accepter bonding positions, where the fluorenone unit is a strong electron acceptor, which has not been reported in the field of TADF materials before, while acridine, one of the most commonly used donors in TADF materials, has strong electron-donating and hole-transport ability. The combination of the strong acceptor and strong donor can give a narrow energy gap and thus longer wavelength emission. Compounds 1 and 2 were thoroughly characterized by 1 H NMR, 13 C NMR and electron ionization (EI) mass spectrometry. Both of them show TADF behavior with orange emission color according to the photoluminescence spectra and time-resolved transient photoluminescence decay measurement. EQEs of 2.9% and 8.9% were achieved for the OLED devices based on 1 and 2, respectively, which are higher than the theoretical efficiency of the OLEDs based on conventional fluorescent materials. ## Results and Discussion 27DACRFT 1 and 36DACRFT 2 have similar thermal properties according to thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) measurements. They have high decomposition temperatures (T d , corresponding to a 5% weight loss) of 361 and 363 °C, respectively. In addition, no glass-transition temperature (T g ) was found according to their DSC curves. Thanks to their amorphous characteristics, the stability of their morphology and chemical composition can be expected during the evaporation processing fabrication of OLEDs. In order to characterize their electrochemical properties, cyclic voltammetry (CV) measurements were conducted to measure The molecular geometry of 1 and 2 in the ground state and excited state were simulated by density functional theory (DFT) and time-dependent density functional theory (TD-DFT) calculations, respectively. The ground state (S 0 ) geometries were optimized on B3LYP/6-31G* level in gas phase, while the lowest triplet excited state (T 1 ) energy levels and the singlet excited state (S 1 ) geometries of those molecules were optimized by TD-DFT on m062x/6-31G* level based on the optimized ground state geometries. The optimized geometries of S 0 and S 1 are shown in Figure 1. The optimized geometries in S 0 are shown in Figure 1a, and all the data are summarized in Table 2. Large twisting angles (θ) of 89.33° and 88.80° between the donor units and the accepter units were estimated for compound 1 and 2, respectively. As shown in Figure 1b, HOMOs and LUMOs are mainly located on the acridine unit and the fluorenone unit, respectively, which contribute to small ΔE ST . The existence of a very small overlap of HOMOs and LUMOs is advantageous to retain high photoluminescence (PL) quantum yields . The calculated ΔE ST of 1 and 2 are 0.33 and 0.27 eV, which are small enough to achieve TADF behavior. As shown in Figure 1c, the twisting angle (θ') of 1 in S 1 is 63.74°, which is much smaller than its θ in S 0 , meanwhile, the conformation of the acridine units in 1 is also changed in S 1 as a result of vibrational relaxation and internal conversion (IC), which means the S 0 geometry of 1 becomes unstable when the molecule is excited and the wave function distribution is changed. The different twisting angles between S 0 and S 1 may reduce its PL property according to the energy gap law as vibrational relaxation and intersystem crossing (IC) processes can consume the energy in S 1 , leading to increased nonradiative deactivation , reduced PL quantum yield, and thus reduced singlet exciton utilization. On the contrary, the geometry of 2 is hardly changed when excited. Thus, compound 2 shows more potentiality in the application of OLEDs for its better configuration. Ultraviolet-visible (UV-vis) absorption and PL spectra in dilute solutions of 1 and 2 (10 −5 M) are presented in Figure 2. Both compounds 1 and 2 have similar absorption peaks at around 345 and 456 nm. The peaks at around 456 nm result from their ICT states from the donor to the acceptor, while the absorption below 380 nm is caused by their short π-conjugation. It is obvious that 2 has not only a higher oscillator strength (f) than 1 from its transition of charge-transfer states, but also a weaker oscillator strength from its local excited (LE) states. It could be considered that 2 has a better configuration, which is advantageous to intramolecular charge transfer compared with 1, which coincides with the conclusion from DFT calculation. The PL spectra of the materials in different solvents were also measured. However, no emission was observed in the dilute solutions of dichloromethane (DCM) and tetrahydrofuran (THF) because vibrational relaxation and internal conversion are promoted to reduce the PL intensity. Both compounds 1 and 2 show almost the same PL spectra in dilute solutions of toluene and n-hexane. The photoluminescence spectra of the n-hexane solutions show a peak at 517 nm with a shoulder at 545 nm, which can be considered as the radiative transition of 1 LE states. Noticeably, the charge-transfer process is limited in n-hexane because of its lower polarity. Only one peak at 593 nm was observed for the dilute toluene solutions of both molecules with the typical PL spectra from the radiative transition of ICT states, which could be the evidence of the existence of strong ICT states of both molecules. More importantly, both materials achieve orange luminescence in a dilute solution of toluene, which could be attributed to the strong electron-withdrawing ability and excess conjugation length of fluorenone plane compared with conventional benzophenone acceptor . In addition, low temperature photoluminescence (LTPL) spectra of the materials in toluene at 77 K were measured. The energy levels of S 1 and T 1 were determined from the onset of the prompt and delayed emission peaks, respectively. As shown in Figure 3, both T 1 states of the materials could be confirmed as 3 CT character from their delayed photoluminescence spectra without any well-defined vibronic structure . The ΔE ST of 1 and 2 are 0.19 and 0.09 eV, respectively, indicating that compound 2 may have a much more efficient RISC process than 2 (Table 3). To gain a further understanding of the photophysical properties of 1 and 2 in solid state, two doped films in 4,4'-dicarbazolyl-1,1'-biphenyl (CBP) were vacuum co-deposited at a concentration of 8 wt % for photoluminescence quantum yield (PLQY) and time-resolved transient photoluminescence decay measurements. The concentration of the doped films was optimized to ensure complete energy transfer between the host and the guest. PLQY measurements of 1:CBP and 2:CBP are 7% and 26%, respectively. The PLQY measurements of the doped films with lower concentration show varying degrees of deviation due to the incomplete energy transfer and the obvious luminescence from CBP (PLQY of 1 and 2 doped in CBP with 1 wt % are 2% ## and 10%, respectively). As shown in Supporting Information File 1, both PL spectra of the doped films of 1:CBP and 2:CBP show red-shift from their PL spectra in n-hexane, which could be considered as the influence from aggregation. As mentioned above, 1 and 2 show nearly the same PL spectra in their dilute toluene solution. However, the PL spectrum of 2 is slightly blue-shifted from its PL spectrum in toluene, while 1:CBP shows alike spectra with 1 in toluene. It could be considered as the solid-state solvation effect , as 2 and 1 have different dipole moment of 1.814 D and 3.501 D, respectively from DFT calculation, owing to their different configurations. The doped film 2:CBP shows a typical TADF behavior as shown in Figure 4b, according to the time-resolved transient photoluminescence decay measurement. The proportion of delayed fluorescence increases rapidly with improved temperature from 77 to 250 K and slowly by acceleration of the nonradiative transition rate when the temperature is higher than 250 K. On the other hand, 1:CBP hardly shows a TADF behavior when the temperature is below 300 K, as shown in Figure 4c. The signals are characterized by noise rather than delayed fluorescence when the temperature is lower than room temperature due to its low PLQY. Delayed fluorescence can be only observed when the temperature is above 300 K. This could be attributed to the large ΔE ST and low PLQY of 1 which requires more energy to achieve RISC process from T 1 to S 1 . According to the integration and the lifetime of the prompt and delayed components of the time-resolved transient PL decay curves at room temperature, the PLQY of their respective components and rate constant of different kinetic processes were calculated, as shown in Table 4. The rate constants were calculated following Equations 1-4 below . ( (2) (3) (4) where k r , k nr , k isc , and k risc represent the rate constant of radiative, nonradiative, intersystem crossing and reverse intersystem crossing, respectively; Φ, Φ PF , Φ TADF , τ PF and τ TADF represent the photoluminescence quantum yield, quantum yield of the prompt component, quantum yield of the delayed component, and lifetimes of the prompt and delayed components, respectively. As shown in Table 4, 2 has a significantly larger k nr than 2, which is consistent with the DFT simulation. On the other hand, a much lower k risc and longer τ TADF was acquired by 1:CBP than 2:CBP, as a result of the blocked reverse intersystem crossing and the large ΔE . Further, the existence of strong IC and vibrational relaxation processes of 1 is proved by its large k nr and low PLQY. In contrast, owing to the relatively small ΔE ST , k risc of 2 is higher and τ TADF is relatively shorter than 1. The short τ TADF not only signifies efficient utilization of singlet excitons, but is also advantageous in reducing the triplet exciton concentration and efficiency roll-off in the OLED devices. Finally, electroluminescent properties of 1 and 2 were characterized in a device structure of ITO/TAPC (25 nm)/1 wt % emitter in CBP (35 nm)/TmPyPB (55 nm)/LiF (1 nm)/Al, where 1,1-bis(4-(di-p-tolylamino)phenyl)cyclohexane (TAPC), 4,4'bis(9H-carbazol-9-yl)biphenyl (CBP), 1,3,5-tri[(3-pyridyl)phen-3-yl]benzene (TmPyPB) and LiF play the roles of hole transport layer, host material, electron transport layer and electron injection layer, respectively . The energy level diagrams and the chemical structures of the materials utilized are shown in Figure 5. TAPC and TmPyPB also play the role of exciton blocking layer at the same time because of their high T 1 energy level. Carriers will also be trapped by the emitter directly because of the energy level difference between CBP and the emitter, which makes it possible for the OLEDs with such a low emitter concentration to achieve complete energy transfer. The performance of the fabricated devices is summarized in Table 5 while the J-V-L (current density-voltage-luminance) and EQE-current density characteristics of the devices are shown in Figure 6. A significantly higher performance was observed from the device based on 2 with a maximal current efficiency (CE max ) of According to the previous study, triplet-triplet annihilation (TTA) might be the main cause of efficiency roll-off in the TADF-OLEDs when the triplet exciton concentration increases with brightness and current density . The efficiency roll-off caused by the TTA process of TADF-OLEDs could be analyzed by the TTA model using Equation 5 below: ( where η 0 represents the EQE without the influence of TTA, and J 0 represents the current density at the half maximum of the EQE; η and J represent the EQE with the influence of TTA and the corresponding current density, respectively. As shown in Figure 7, both devices show good agreement with the TTA model fitted curves at low current density because TTA process is the leading factor to the efficiency roll-off of TADF-OLEDs when the exciton concentration is low. With the increase of exciton concentration, singlet-triplet annihilation (STA), singlet-polaron annihilation (SPA) and triplet-polaron annihilation (TPA) may also have serious impact to the efficiency rolloff, which cause the TTA model fitted curves to deviate from the actual value. The device based on 2 shows a better agreement with the fitted curve in higher current density while the device based on 1 does not. In addition, 2 has a better triplet exciton utilization ability to reduce the efficiency roll-off, which comes to the same conclusion with the analysis of their photophysical properties. ## Conclusion In summary, two novel D-A-D-type orange-emitting TADF materials, namely 2,7-bis(9,9-dimethylacridin-10(9H)-yl)-9Hfluoren-9-one (27DACRFT, 1) and 3,6-bis(9,9-dimethylacridin-10(9H)-yl)-9H-fluoren-9-one (36DACRFT, 2), with the fluorenone unit as acceptor and the acridine as donor, were synthetized. Compounds 1 and 2 are isomers but show greatly different performance in terms of both photoluminescence and electroluminescence. It has been shown that the fluorenone unit is a promising acceptor for orange TADF materials, which aids in the design of the TADF behavior and luminescence color of 1 and 2. Owing to the strong electron-withdrawing ability and extended conjugation length of fluorenone unit, the emission peaks of both materials show obvious red-shifts from other TADF materials based on carbonyl acceptor . According to the DFT and TD-DFT simulation and photophysical characterization, 2 shows a smaller singlet-triplet energy difference (ΔE ST ) and a larger radiative rate constant (k r ) to give reduced internal conversion, promoted RISC process, and thus a better triplet exciton utilization ability. Maximum EQE values of 8.9% and 2.9% were achieved for the OLED devices based on 2 and 1, respectively. Efficiency roll-off, which is considered to be the result of TTA, is also reduced more effectively for the OLEDs based on 2. Experimental

chemsum

{"title": "D\u2013A\u2013D-type orange-light emitting thermally activated delayed \ufb02uorescence (TADF) materials based on a fluorenone unit: simulation, photoluminescence and electroluminescence studies", "journal": "Beilstein"}

universal_endogenous_antibody_recruiting_nanobodies_capable_of_triggering_immune_effectors_for_targe

## Abstract: Developing monoclonal antibodies (mAbs) for cancer immunotherapy is expensive and complicated.Nanobodies are small antibodies possessing favorable pharmacological properties compared with mAbs, but have limited anticancer efficacy due to the lack of an Fc region and poor pharmacokinetics. In this context, engineered universal endogenous antibody-recruiting nanobodies (UEAR Nbs), as a general and cost-effective approach, were developed to generate functional antibody-like nanobodies that could recapitulate the Fc biological functions for cancer immunotherapy. The UEAR Nbs, composed of the IgG binding domain and nanobody, were recombinantly expressed in E. coli and could recruit endogenous IgGs onto the cancer cell surface and trigger potent immune responses to kill cancer cells in vitro.Moreover, it was proved that UEAR Nbs displayed significantly improved half-lives in vivo. The in vivo antitumor efficacy of UEAR Nbs was demonstrated in a murine model using EGFR positive triple-negative breast cancer (TNBC). ## Introduction Nanobodies, also known as single-domain antibodies, are heavy-chain-only antibodies with a single variable antigenbinding domain derived from camelids. 1 Compared to conventional monoclonal antibodies (mAbs), nanobodies have unique characteristics including small size, high specifc affinity, outstanding stability and solubility, easy production and manipulation and low immunogenicity. 2,3 Now, it is well recognized that nanobodies are a promising alternative to conventional mAbs from basic research to clinical application, where nanobodies or chemical functionalized nanobodies are extensively explored as imaging, diagnostic and therapeutic agents in diverse areas. 4,5 For example, the frst nanobody, caplacizumab, was approved by the EU and FDA for acquired thrombotic thrombocytopenic purpura treatment recently, 6,7 and more than dozens of nanobody-based medicines are in different stages of clinical trials. However, for cancer immunotherapy, the capability of nanobodies was largely limited because of the lack of an Fc region and the poor pharmacokinetic profle. Detailed mechanism studies have established that the Fc region in mAbs could trigger a potent innate immunity to kill cancer cells and prolong the serum halflife of mAbs, through antibody-dependent cellular cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC), antibodydependent cellular phagocytosis (ADCP), and Fc receptormediated recycling mechanism, respectively. 8 Therefore, a general and cost-effective technology that is able to generate antibody-like nanobodies that could recapitulate the Fc biological functions for cancer immunotherapy is required. Currently, nanobody fusion with anti-HSA nanobody and modifcation with a half-life extension moiety are two effective technologies to improve the pharmacokinetics of nanobodies, but these are unable to reinstate the Fc-mediated biological functions. 9,10 Engineered nanobody fusion with the Fc portion is a popular technology, whose product has advanced into clinical trials for cancer immunotherapy. 11 However, Fc-fused nanobodies suffer from similar shortcomings to the conventional mAbs, such as complicated and expensive production, 12 potential immunogenicity caused by human allotype of the Fc portion, 13 and reduced ADCC or CDC activity due to the low affinity to FcgRIIIa on immune cells or the complement. 14,15 Antibody-recruiting molecules (ARMs) are bifunctional molecules composed of a cell-targeting moiety and an antibodybinding component, which could bridge the target cells and immune system and induce downstream immunity to eliminate the target cells. 16,17 Many rationally designed ARMs, where preferred haptens that could be recognized by natural occurring endogenous antibodies, such as dinitrophenyl (DNP), galactose-a-1,3-galactose (aGal) and rhamnose (Rha), as an antibodybinding component, have been successfully achieved for cancer, viruses, 23,24 bacteria 25,26 and others. 27 For example, we and others demonstrated that nanobody-DNP conjugates could form an in situ immune-complex with specifc endogenous anti-DNP antibody existing in human serum and subsequently provoke potent ADCC and CDC cytotoxicity to target destructing cancer cells in vitro and exhibit in vivo antitumor activity in mouse xenograft models. 28,29 Notably, it is observed that the pharmacokinetic profle of nanobody-DNP conjugates was improved more than 20-fold in the presence of anti-DNP antibodies. 28 This strategy provides a novel and effective solution to reconstitute the missing Fc-mediated anticancer biological function and extend the half-life of the nanobody simultaneously, which could potentially avoid the adverse immunogenicity issues associated with nanobody-Fc fusion protein. However, the success of the specifc endogenous antibodyrecruiting strategy for cancer immunotherapy, including ARMs, nanobody-DNP conjugates and others, was heavily dependent on the amount of anti-hapten antibodies in the human system and their affinity to ARMs, given the fact that only approximately 1-3% of hapten-specifc antibodies pre-exist in human blood. Although pre-vaccination could generate high titers of the hapten-specifc antibody for this purpose, the complicated treatment procedure as well as the unknown side effects may attenuate the possibility of clinical transformation of this technology. To address the short supply of specifc endogenous antibody, relative abundant human IgGs (10-20% in total serum protein), 33 a product of delayed immune response to infections in life, attract our attention. We speculated that an engineered nanobody that could bind with the target cancer cells and recruit the universal endogenous IgGs in human serum without compromising the capability of Fc biological function may solve the above challenge. In this case, a synthetic IgG-binding domain derived from domain B in Staphylococcus aureus protein A, 34,35 also designated as the ZZ domain, which has been successfully applied in IgG purifcation, 36,37 detection 38,39 and immunological therapy, 40 seems to ft these requirements. Importantly, crystal structure studies demonstrated that this domain and the human Fc receptor on immune cells could bind at different regions of an IgG Fc portion, 35,41 respectively, indicating that the fusion of this domain with the nanobody may not affect the required Fc biological functions. We report here the design and construction of an engineered universal endogenous antibody-recruiting nanobody (termed UEAR Nbs to describe this technology) that was recombinantly expressed in E. coli by fusing the IgG binding domain with the nanobody through a GS linker. As shown in Fig. 1, the UEAR Nbs could recognize the cancer cells and recruit general endogenous IgGs onto the surface of cancer cells, followed by triggering immune responses, such as ADCC, CDC and ADCP, to eliminate tumor cells in a precise manner by the engagement of IgG Fc-mediated mechanisms. Furthermore, the pharmacokinetics of the UEAR Nbs as well as the anti-cancer efficacy in triple-negative breast cancer (TNBC) xenograft mice was evaluated in vivo. ## Results and discussion Design and expression of UEAR Nbs 7D12-ZZ and ZZ-7D12 in E. coli The epidermal growth factor receptor (EGFR) is a transmembrane glycoprotein belonging to a family of four receptor tyrosine kinases. Ligand-induced activation of the EGFR and its downstream signaling pathways are involved in many cellular processes, such as cell proliferation, mobility and differentiation. 42 Overexpression of the EGFR has been observed in a variety of cancer types, including breast cancer (especially in TNBC), lung cancer, metastatic colorectal cancer, and head and neck cancer. 43 Consequently, the EGFR is a validated therapeutic target as well as a diagnostic biomarker in clinic. In this work, an EGFR targeting nanobody 7D12 was selected as the tumor-targeting molecule. Previous studies have confrmed that 7D12 binds to domain III of human EGFR with high affinity 44 and has the potential to overcome EGFR ecto-domain mutantmediated primary and secondary resistance to clinically approved full mAbs. 45 For a universal antibody-recruiting moiety, the ZZ domain was selected as the Fc-binding molecule because of its high binding affinity and specifcity. Two UEAR Nbs, 7D12-ZZ and ZZ-7D12, where the ZZ domain was placed at the N-and C-terminus of nanobody 7D12 via a GS linker (GGGGSGGGGS), respectively, were designed (Fig. 2A). In addition, a His6 tag was genetically engineered to the Cterminus for affinity purifcation purpose. After the recombinant plasmids of pET22b-7D12-ZZ and pET22b-ZZ-7D12 were constructed, they were transformed into E. coli BL21 (DE3) cells, which were cultured at 37 C in Terrifc-Broth (TB) medium and induced with isopropyl-b-D-thiogalactopyranoside (IPTG) to express proteins. The recombinant UEAR Nbs were harvested and purifed by Ni-resin affinity chromatography to give 7D12-ZZ and ZZ-7D12 in a yield of 50 and 100 mg L 1 , respectively. Nanobody 7D12 and the ZZ domain were recombinantly expressed similarly in E. coli BL21 (DE3) with a yield of 40 and 90 mg L 1 . All proteins were then characterized by SDS-PAGE. 7D12 or ZZ domain protein appeared as a single band around 16 kDa (Fig. 2B, lane 1 and 2), while 7D12-ZZ and ZZ-7D12 appeared as a single band with a molecular weight of approximately 32 kDa (Fig. 2B, lane 3 and 4). These proteins were unambiguously confrmed by MS, which was consistent with the calculated molecular weight (Fig. S1 †). ## Characterization of UEAR Nb binding with IgGs and cancer cell overexpressing EGFR With UEAR Nbs, 7D12-ZZ and ZZ-7D12, in hand, we frst evaluated their binding affinity to IgGs from different mammalian species using an ELISA method. The experiments were performed by coating the plates with PBS buffer, 7D12, ZZ domain, 7D12-ZZ and ZZ-7D12 proteins, respectively, and then incubated with IgGs derived from mouse, rabbit and human, followed by HRP-modifed secondary antibodies. The results were determined by using a TMB kit. As shown in Fig. 3A, no appreciable signal was observed in PBS and 7D12 coated wells. By contrast, strong signals were observed in the ZZ domain, 7D12-ZZ and ZZ-7D12 coated wells incubated with three IgG samples. The intensities of absorption at 450 nm from 7D12-ZZ and ZZ-7D12 groups were comparable to those of the ZZ domain group. This result suggested that placing the ZZ domain at the C-or N-terminus of nanobody 7D12 did not alter its structure signifcantly; both formats retained their binding capability to IgGs after fusion with nanobody 7D12. Then, to evaluate the binding affinity of various Nbs to the EGFR, a standard competitive cell-based ELISA was performed. After being fxed with EGFR positive A431 cells, the wells were treated with 7D12, 7D12-ZZ or ZZ-7D12 in the presence of recombinant human EGFR with different concentrations (0.016 to 500 nmol L 1 ). Thereafter, wells were incubated with anti-His6 tag antibodies and HRP-modifed anti-mouse IgG antibodies. The results were fnally determined using a TMB kit. As shown in Fig. 3B, 7D12, 7D12-ZZ and ZZ-7D12 all presented an EGFR dose-dependent binding efficiency, with an IC 50 of 6.3, 3.8 and 9.7 nmol L 1 , respectively. This result indicated that both of UEAR Nbs remained at a nanomolar level of binding affinity to the EGFR. However, compared with 7D12, the binding affinity of 7D12-ZZ was improved approximately 1.7 fold while ZZ-7D12 exhibited slightly decreased binding affinity. Next, we need to prove whether UEAR Nbs could bind specifcally to EGFR positive cancer cells and recruit endogenous IgGs onto cell surfaces. Accordingly, three cell lines, namely human TNBC MDA-MB-468 cells, human squamous carcinoma A431 cells with high EGFR expression, and MCF7 cells with low EGFR expression, were used for this study. The cells were seeded in 24-well plates and treated with 7D12, ZZ domain, 7D12-ZZ and ZZ-7D12, respectively, followed by incubation with mouse IgGs and Dylight 488-conjugated anti-mouse IgG antibodies. The cells were then imaged using a fluorescence microscope. As presented in Fig. 4A Furthermore, we conducted a quantitative evaluation of cellbinding and IgG-recruiting ability of UEAR Nbs using flow cytometry. As shown in Fig. 4B and C, only EGFR positive cells (A431 and MDA-MB-468), but not EGFR negative cells (MCF7), displayed signifcantly IgG anchoring when treated with UEAR Nbs. The mean fluorescence intensity (MFI) detected on A431 after incubation with 7D12-ZZ or ZZ-7D12 was 6.8-and 7.3-fold higher than that of 7D12, respectively. Notably, the MFI detected from the MDA-MB-468 cell group treated with 7D12-ZZ or ZZ-7D12 was 20.4-and 20.9-fold higher as compared with the result of the 7D12 control group. To further reveal the effect of In vitro cytotoxicity evaluation induced by UEAR Nbs ADCC, CDC and ADCP are three known crucial anti-cancer mechanisms engaged by the Fc-portion of therapeutic mAbs. ADCC critically relies on the binding of Fc with the specifc FcgRIIIa receptor on innate immune cells, such as NK cells and dT cells existing in human peripheral blood mononuclear cells (PBMCs). Based on the different binding profles of the ZZ domain and FcgRIIIa receptor to Fc-portion, 46,47 it was expected that UEAR Nbs may bridge the cancer cells and immune cells through the Fc-terminus of the recruited universal endogenous IgGs to trigger ADCC cytotoxicity. To verify this hypothesis, we frst utilized A431 cells and MDA-MB-468 cells to assess ADCC. Following incubation with different concentrations (0.2, 2, 20 and 50 nmol L 1 ) of UEAR Nbs and mouse serum (as the source of endogenous IgGs), cells were co-cultured with freshly isolated PBMCs (as the source of immune cells). Then, cell lysis was determined using a lactate dehydrogenase (LDH) cytotoxicity kit. As presented in Fig. 5A, a dose-dependent ADCC activity in the experiment group was observed, where cell lysis increased with the increase of UEAR Nb concentrations. By using a dosage of 50 nmol L 1 of UEAR Nbs, over 25 and 44% of A431 and MDA-MB-468 cells were killed, respectively. However, the viabilities of A431 and MDA-MB-468 cells were not influenced (<10%) when treated with 7D12. To induce an effective ADCC in this system, an immune complex, formed by cancer cells, UEAR Nbs, endogenous IgGs and FcgRIIIa receptor on immune cells, is required. Immunofluorescence image and flow cytometry experiments have demonstrated that a ternary complex by cancer cells, UEAR Nbs and endogenous IgGs was formed successfully. The ADCC results further demonstrated that immune cells were involved in this system. Although we were unable to isolate the active immune complex, the result clearly indicated that UEAR Nbs, both 7D12-ZZ and ZZ-7D12, could bridge the cancer cells and immune cells by the engagement of endogenous IgGs effectively, thereby triggering ADCC to destruct target cells. CDC is a classic complement pathway initiated by the formation of the antibody-complement component 1q (C1q) complex and a series of subsequent cascade reactions. 48 We next evaluated the CDC cytotoxicity mediated by UEAR Nbs. To this end, cells were incubated with 50 nmol L 1 of UEAR Nbs in the presence of mouse serum and rabbit complement (RC); then, the cell viability was measured using a CCK8 kit. As shown in Fig. 5B, distinct cytotoxicity was observed when cancer cells were treated with 7D12-ZZ or ZZ-7D12. For example, approximately 53.2 and 61.2% of MDA-MB-468 cell lysis were observed when incubated with 7D12-ZZ and ZZ-7D12, respectively. The potency was 10.5-and 12.2-fold higher as compared with the results of the 7D12 treatment group. Additionally, no obvious cell lysis was observed when the heat inactivated rabbit complement (HIRC) was applied in UEAR Nb groups, further demonstrating that the observed cell-killing effect was indeed induced by CDC. The similar potency induced by UEAR Nbs in the two cancer cell lines proved that the fusion sites of the ZZ domain had no signifcant impact on the CDC activities. ADCP is another important mechanism to induce target cell phagocytosis through the activation of FcgRs on macrophages by the Fc-moiety of mAbs. 49 THP-1 is a human monocyte cell line that has been successfully used in phagocytosis assays. 50 In this regard, we employed THP-1 as effector cells to investigate the ADCP level induced by UEAR Nbs. To do this, target cells and THP-1 cells were frst stained with DiO (green cell membrane probe) and DiI (orange red cell membrane probe), respectively. Then, target cells were treated with UEAR Nbs and human serum (as the source of endogenous IgGs), followed by incubation with THP-1 cells. As directly observed by confocal fluorescence imaging (Fig. S6 and S7 †), both cancer cell A431 and MDA-MB-468 cells were clearly phagocytosed by THP-1 cells in the presence of UEAR Nbs and human serum, indicating that UEAR Nbs were capable of evoking ADCP. We further quantitatively analysed the phagocytosis efficiency by counting the cell numbers in R1 (double-positive cells) and R2 (remaining target cells) in flow cytometry experiments. As shown in Fig. 6A, considerable amounts of double-positive cells were only observed in the groups of 7D12-ZZ and ZZ-7D12, whereas nearly no double-positive cells were observed in the groups of PBS and 7D12. Interestingly, the potency of 7D12-ZZ-and ZZ-7D12-mediated ADCP displayed discriminated profles. The phagocytoses of A431 and MDA-MB-468 cells induced by 7D12-ZZ could reach 45.3 and 57.6%, respectively, which were signifcantly higher than those induced by ZZ-7D12 (A431, 21.5%; MDA-MB-468, 22.2%) (Fig. 6B). The underlying mechanism is unknown at this stage. It is speculated that the difference may be caused by the different interactions of the UEAR Nb-IgG immune complex with multiple FcgRs involved ADCP activation. ## Pharmacokinetics evaluation of UEAR Nbs One of the main challenges for nanobody-based therapeutics is their relatively short half-life in vivo. For example, the half-life of 7D12 in mice was less than 10 min. To investigate whether ZZ domain fusion could prolong the serum half-life of UEAR Nbs, we performed in vivo experiments to determine their half-lives. Thus, 7D12-ZZ and ZZ-7D12 were intravenously (i.v.) injected into the healthy mice and the blood samples were collected at different time points. The corresponding plasma concentrations of 7D12-ZZ and ZZ-7D12 were determined according to a standard curve obtained by ELISA assay (Fig. S8 †). As shown in Fig. 7, the half-life of 7D12-ZZ and ZZ-7D12 could reach 25.9 and 29.5 h, respectively, which were approximately 160.9-and 183.4fold improved compared to the previously reported 0.16 h halflife of 7D12. 28 Additionally, no signifcant difference in half-life was observed between 7D12-ZZ and ZZ-7D12. The signifcant improvement of the pharmacokinetics of UEAR Nbs could be attributed to the in situ formed immune complex of the ZZ domain with endogenous IgGs, whose sizes were increased to above the renal threshold, thereby avoiding waste-pass clearance. ## In vivo antitumor efficacy evaluation of UEAR Nbs TNBC is a notorious breast cancer subtype that lacks the overexpression of the estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor-2 (HER2), which comprises 10-24% of all breast cancers. 51 However, previous studies using human TNBC tissue demonstrated that EGFR overexpression frequently occurred in TNBC. MDA-MB-468 cells are validated triple-negative basal-A mammary carcinoma with positive EGFR overexpression. Thus, MDA-MB-468 cells were s.c. injected into Balb/c nude mice to create a TNBC xenograft mice model; then, these mice were randomly divided into four groups which were treated by i.p. injection of PBS, 7D12, 7D12-ZZ, or ZZ-7D12 (Fig. 8A), respectively. Treatments were given every two days in 10 days using 50 mL of nanobodies in PBS (30 mmol L 1 ) and 50 mL of pooled normal mouse serum as the source of endogenous IgGs. During the course of experiments (18 days), we monitored the tumor size in mice every two days. As described in Fig. 8B, the tumor volume in groups treated with 7D12-ZZ or ZZ-7D12 was signifcantly reduced and persistent tumor regression was observed in the treatment period as well as after treatment cessation. To further ascertain the antitumor efficacy of UEAR Nbs, we excised tumors from each mouse and conducted tumor weight measurement at the experimental endpoint. As shown in Fig. 8C-E, for PBS and 7D12 groups, the mean tumor weight was 419.2 AE 122.7 mg, 255 AE 62.5 mg, indicating that nanobody 7D12 exhibited a limited tumor growth inhibition effect (TGI, 39.2%). By contrast, for 7D12-ZZ and ZZ-7D12 treatment groups, the mean tumor weights were 27.8 AE 9.0 and 13.4 AE 2.2 mg. The TGI was 89.1 and 94.7% (7D12-ZZ and ZZ-7D12 vs. 7D12), respectively. This result clearly suggested that UEAR Nbs could signifcantly enhance the antitumor efficacy of nanobodies by reconstituting the Fc functions in the presence of endogenous IgGs. The improved pharmacokinetics of UEAR Nbs may also contribute to this outstanding antitumor activity. To evaluate the toxicity of UEAR Nbs, the mouse weight was monitored and no obvious weight loss was observed in mice treated with either nanobody 7D12 or UEAR Nbs (Fig. S9 †). In addition, hematoxylin and eosin (H&E) staining results presented in Fig. S10 † revealed that none of the collected tissues showed acute or chronic inflammation or necrotic regions. Moreover, to evaluate the anaphylactic shock risk of UEAR Nbs, C57BL/6J mice were i.p. injected with 100 mL of 7D12-ZZ or ZZ-7D12 (1 mg mL 1 in PBS) at day 0, followed by day 7 to 15 with a frequency of every two days. During the course of experiment, no anaphylactic shock was observed. In addition, the nanobody-specifc IgE level of serum samples at day 15 was as low as that at day 0 (Fig. S11 †), indicating the low anaphylactic shock risk of UEAR Nbs. All the above results suggested that UAER Nbs were essentially nontoxic. ## Conclusions mAb-based cancer immunotherapy has achieved great success in the past few decades. 52,53 However, it is widely recognized that developing therapeutic mAbs as drugs is complicated and expensive. Nanobodies, as a new type of antibody fragment derived from camelids, offer competitive advantages over full length mAbs and other antibody fragments, including physiochemical properties, discovery, optimization and production. The ARM strategy is a promising modality for cancer immunotherapy. However, currently, only hapten-specifc naturally occurring antibodies were harnessed to induce humoral and cellular immune responses to fght cancer cells. Recently, Fc-ARMs, composed of folic acid and a Fc binding peptide, were nicely constructed and could redirect endogenous antibodies to trigger an anti-cancer immune response. 54 However, this molecule exhibited marginal in vivo anti-tumor efficacy. In this work, to make a functional antibody-like nanobody for cancer immunotherapy, a simple and robust UEAR Nb was constructed based on the ARM concept and our previous studies, to reinstate the missing Fc biological functions. We demonstrated that the UEAR Nbs could recruit abundant universal endogenous IgGs, independent of hapten-specifc antibodies in human serum, onto cancer cell surfaces and exert the Fc-mediated anticancer biological mechanisms to eliminate cancer cells in vitro and in vivo. A benefcial pharmacokinetic improvement of UEAR Nbs was achieved simultaneously with this strategy. Considering the superior targeting capability of nanobodies, this strategy will provide a general and economic approach to generate antibodylike nanobodies for cancer and other disease applications. ## Ethical statement All animal experiments were performed according to the guidelines and protocols approved by the Institutional Animal Care and Use Committee of the Jiangnan University (JN. No. 20190315b0650625 and 20201130c0650220). ## Conflicts of interest The authors declare no conflict of interest.

chemsum

{"title": "Universal endogenous antibody recruiting nanobodies capable of triggering immune effectors for targeted cancer immunotherapy", "journal": "Royal Society of Chemistry (RSC)"}

internal_extractive_electrospray_ionization_mass_spectrometry_for_quantitative_determination_of_fluo

## Abstract: Antibiotics contamination in food products is of increasing concern due to their potential threat on human health. Herein solid-phase extraction based on magnetic molecularly imprinted polymers coupled with internal extractive electrospray ionization mass spectrometry (MMIPs-SPE-iEESI-MS) was designed for the quantitative analysis of trace fluoroquinolones (FQs) in raw milk samples. FQs in the raw milk sample (2 mL) were selectively captured by the easily-lab-made magnetic molecularly imprinted polymers (MMIPs), and then directly eluted by 100 µL electrospraying solvent biased with +3.0 kV to produce protonated FQs ions for mass spectrometric characterization. Satisfactory analytical performance was obtained in the quantitative analysis of three kinds of FQs (i.e., norfloxacin, enoxacin, and fleroxacin). For all the samples tested, the established method showed a low limit of detection (LOD ≤ 0.03 µg L −1 ) and a high analysis speed (≤4 min per sample). The analytical performance for real sample analysis was validated by a nationally standardized protocol using LC-MS, resulting in acceptable relative error values from −5.8% to +6.9% for 6 tested samples. Our results demonstrate that MMIPs-SPE-iEESI-MS is a new strategy for the quantitative analysis of FQs in complex biological mixtures such as raw milk, showing promising applications in food safety control and biofluid sample analysis. Antibiotics have been widely used for decades to effectively treat a variety of bacterial infections, and great contributions have been made in human health protection. Unfortunately, because of worldwide overuse and misuse of antibiotics in planting and breeding production process , bacteria are often becoming strongly resistant to hospital treatment 4,5 . Thus the antibiotics contamination in food products is of increasing concern due to their hazardous effects on human health and ecosystem 1, , which include but not limited to the infections caused from antibiotic-resistant bacteria and possible carcinogenicity . Among series of antibiotics, fluoroquinolones (FQs) are one kind of broad-spectrum antibiotics, which are ubiquitously used in human health care and veterinary applications 13 . Conventional analytical methods including microbiological methods 14 , electrochemical method 15 , fluorospectrophotometry 16 , high performance thin layer chromatography (HPTLC) 17 , high performance liquid chromatography with ultraviolet detector (HPLC-UV) 18 , high performance liquid chromatography mass spectrometry (HPLC-MS) 19 and enzyme immunoassay 20 have been applied to the detection of FQs in environment water, foodstuffs, and biofluid samples, etc. Although spectroscopy detection methods (e.g., ultraviolet detector) have been widely used in the determination of FQs benefited by the chromophore or fluorophore groups in the FQs molecule 21 , its limited sensitivity may be a challenge in specific application. Moreover, tedious sample pretreatments (e.g., centrifugation, diluting, and multistep chemical extraction, etc.) for the matrix clean-up are routinely required, which prevents the high-throughput analysis of FQs in practical samples. Thus, there is an urgent demand for the development of highly efficient analytical methods of sensitive and selective identification or quantification of FQs in samples with complex matrices. Recently, ambient mass spectrometry (AMS) allows the direct analysis of complex samples with high speed, high selectivity, and high sensitivity . Charged droplet generated by electro-spray or sonic spray is a common ionization reagent, which is widely used in various ambient ionization technologies such as desorption electrospray ionization (DESI) 25 , probe electrospray ionization (PESI) 26 , extractive electrospray ionization (EESI) 27 , laser ablation electrospray ionization (LAESI) 28 , and easy ambient sonic spray ionization (EASI) 29 , etc. Benefited by the high ionization energy, the primary ions generated by electric field (electron/plasma) have been employed in many ambient ionization technologies including direct analysis in real time (DART) 30 , low temperature plasma (LTP) 31 , microwave plasma torch (MPT) 32 , plasma assisted laser desorption ionization (PALDI) 33 , dielectric barrier discharge ionization (DBDI) 34 , desorption atmospheric pressure chemical ionization (DAPCI) 35 , etc., which are of unique advantages for the preparation of specific analytes ions from raw samples. Great convenience has been provided by these versatile ambient ionization technologies owing to the direct sampling or ionization of raw samples. To date, efforts are still devoting to improve the analytical performance of AMS facing highly complex matrices. In recent years, fast and facile sample pretreatment methods (e.g., solid-phase microextraction (SPME) 36,37 , magnetic solid-phase extraction (MSPE) 38 , thin-layer chromatography 39 , solid phase mesh enhanced sorption from headspace (SPMESH) 40 , etc.) combined with AMS has been developed for direct analysis of trace target analytes in various highly complex samples (e.g., biological, environmental, food, forensic samples, or even individual small organism), which greatly improved the sensitivity and selectivity of AMS. Given raw milk as a typical example of extremely complex matrix, a facile method of solid-phase extraction based on magnetic molecularly imprinted polymers combined with internal extractive electrospray ionization mass spectrometry (MMIPs-SPE-iEESI-MS) was designed for the quantitative analysis of FQs in raw milk samples. FQs in the raw milk samples were selectively captured by the MMIPs for subsequent iEESI-MS interrogation. Overall, the established method showed a high sensitivity in the determination of three kinds of FQs (norfloxacin, enoxacin, and fleroxacin) in raw milk samples. Our results demonstrate that the established MMIPs-SPE-iEESI-MS is a powerful method for the quantitative analysis of FQs in raw milk samples, providing potential application value in other biofluid sample analysis. ## MMIPs-SPE-iEESI-MS Analysis of FQs in raw milk samples. To exclude false positive result for the analysis of FQs in milk samples, collision-induced dissociation (CID) experiments were performed for all the suspected FQs protonated molecule ions of m/z 320, m/z 321, and m/z 370. Figure 1 shows the MS/MS spectra of precursor ions of m/z 320, m/z 321, and m/z 370 collected from raw milk samples with authentic FQs (10 μg L −1 ). 44,45 . All the protonated molecule ions of [norfloxacin +H] + , [enoxacin +H] + , and [fleroxacin +H] + were easily to occur neutral loss of H 2 O and CO 2 under the CID conditions. The loss of CO 2 (−44) got characteristic fragment ions of m/z 276, m/z 277, and m/z 326 by precursor ions of norfloxacin, enoxacin, and fleroxacin, respectively. These CO 2 (−44) lost fragment ions should provide higher significance for the identity check of these three FQs. Thus, the signal intensities of fragment ions of m/z 276, m/z 277, and m/z 326 were selected as analytical response to establish the quantitative methods for norfloxacin, enoxacin, and fleroxacin, respectively. As a result, the FQs in the raw milk samples were successfully detected using MMIPs-SPE-iEESI-MS. Optimization of MMIPs-SPE-iEESI. For better performance during MMIPs-SPE-iEESI-MS analysis, analytical parameters including sorbent amount, composition, volume of extraction solvent, and the flow rate of extraction were optimized using FQs spiked raw milk as samples. The concentration of each FQs (i.e., norfloxacin, enoxacin, and fleroxacin) was set at 10 μg L −1 in all the milk samples. MMIPs material was simply fabricated by co-mixing of Fe 3 O 4 magnetic nanoparticles (MNPs) and a commercial molecularly imprinted polymers (MIPs) products in methanol. As shown in the SEM image of MMIPs material (Fig. 2), the MNPs were coated on the surface of the MIPs after the co-mixing preparation. Additional elemental analysis of the MMIPs, Fe 3 O 4 MNPs, and MIPs also imply the assembly of Fe 3 O 4 MNPs and MIPs (Supplementary Fig. S1). A comparison experiment of the Fe 3 O 4 MNP material (without MIPs) and the MMIP material (with MIPs) was carried out. As expected, the target FQs signals were remarkably increased when using MMIP material (Supplementary Fig. S2). To achieve high adsorption performance for the FQs, different amounts of MIPs material (i.e., 0, 0.5, 1.5, and 2.0 mg) were experimentally investigated for FQs adsorption, while the amount of MNPs was kept at 2.0 mg. The signal intensities of the three FQs notably increased with the increase of MIPs amount from 0 to 1.5 mg, and showed a decreasing trend when the MIPs amount increased to 2.0 mg (Fig. 3a). As a result, 1.5 mg MIPs and 2.0 mg MNPs were used for the preparation of MMIPs material. As shown in the SEM image of MMIPs material (Fig. 2), the MNPs were coated on the surface of the MIPs material. Considering the extraction solution was acted as both the elution solution for FQs desorption and the solution for electrospray, the extraction solution was also investigated. Methanol containing with different proportion of ammonia 0%, 0.5%, 1.0%, 2.0%, 4.0%, 6.0%, and 8.0% (w/w) were applied for the MMIPs-SPE-iEESI-MS analysis. As a result, 2.0% ammonia in methanol (w/w) was the optimal extraction solution (Fig. 3b). The increased ammonia proportion in methanol should be helpful for the desorption of FQs, while excessively high concentration of ammonia (e.g., 8.0%, w/w) may suppress the ionization efficiency of FQs. Moreover, the volume of the extraction solution for the elution of FQs from the MMIPs material and the flow rate of the solution were also optimized to achieve better elution and ionization efficiency. Higher FQs signal intensity was obtained under a volume of 100 μL and a flow rate at 8 μL min −1 (Fig. 3c and d). Finally, optimized conditions showed satisfactory performance for the determination of three kinds of FQs in raw milk samples. ## Quantitative analysis of FQs in milk samples using MMIPs-SPE-iEESI-MS. Three kinds of FQs standard solutions (i.e., norfloxacin, enoxacin, and fleroxacin, respectively) were spiked in blank raw milk samples (2 mL) to make a series of working solutions containing 0.1-500.0 μg L −1 of FQs for MMIPs-SPE-iEESI-MS/ MS analysis. In the case of norfloxacin, the signal intensity of m/z 276 was linearly responded with norfloxacin concentrations over the range of 0.1-500.0 μg L −1 (R 2 = 0.9999) (Fig. 4a). The LOD of norfloxacin defined by a signal-to-noise ratio (S/N) of 3 was estimated to be 0.019 μg L −1 . The relative standard deviations (RSDs) of six replicates for the norfloxacin concentrations ranging from 0.1-500.0 μg L −1 were less than 8.7% (detailed in Supplementary Table S1). For the quantitative analysis of enoxacin and fleroxacin, the linear responding ranges and relative standard deviation values (n = 6) were 0.1-100.0 µg L −1 (R 2 = 0.9999) and less than 7.5% for enoxacin (Fig. 4b and detailed in Supplementary Table S2), and 0.1-500.0 µg L −1 (R 2 = 0.9995) and less than 8.4% for fleroxacin (Fig. 4c and detailed in Supplementary Table S3), respectively. The LODs defined by a signal-to-noise ratio (S/N) of 3 were estimated to be 0.022 μg L −1 for enoxacin and 0.024 μg L −1 for fleroxacin (Table 1), respectively. A short time estimated less than 4 min (exclude the time of MMIPs preparation) was taken for each measurement. Recoveries of all the three FQs from raw milk samples were also estimated by analyzing spiked samples. Acceptable recoveries from 82.5% to 110.0% were obtained for all the samples, and RSDs (n = 6) of all spiked samples were less than 9.4% (Table 1). Furthermore, intra/inter-day precision and accuracy of the method were carried out with the FQs spiked at three different concentrations in milk samples. The intra-day precision and accuracy were determined on the same day and consisted of six replicates at each of three concentration levels, and the inter-day precision and accuracy were carried out with a continuous fourteen days. The results obtained are shown in Table 2. The intra-and inter-day RSDs were less than 8.2% and 10.9%, respectively, while the intraand inter-day recoveries ranging from 84.7 to 104.8% and from 85.9 to 105.6% were obtained, respectively. ## Method validation. Validation of the analytical results of MMIPs-SPE-iEESI-MS for detection of trace FQs in milk samples were performed using the conventional off-line LC-MS/MS method according to a standard operation procedure recommended on National Standard of China (GB/T 22985-2008) (detailed in Supplementary). As summarized in Table 3, the MMIPs-SPE-iEESI-MS results were all in good agreement with those obtained by LC-MS/MS. The good recovery rates (94.2-106.9%) and relatively low relative errors (−5.8% to +6.9%) confirmed that the MMIPs-SPE-iEESI-MS perfectly meet the requirement for the quantitative determination of FQs in raw milk samples. ## Discussion In the optimization of the MMIPs amounts, the signal intensities of three kinds of FQs were increased with the increase of MIPs amounts from 0 to 1.5 mg, indicating that more FQs molecules in the complex milk sample were captured by the MMIPs material, which is consistent with higher ratio of MIPs. Interestingly, the signal intensities were decreased by using 2.0 mg MIPs. The preparation of MMIPs by co-mixing method was interpreted as "aggregate-wrap" process, i.e., the MIPs were likely to be wrapped by MNPs and aggregated to form a magnetic composite 46,47 . In this respect, sufficient MNPs were necessary to ensure all the MIPs material could be magnetic coated for the milk matrix separation. As the amount of MNPs was fixed at 2.0 mg, the magnetism of the MMIPs particles was decreased when more mass of MIPs (e.g., 2.0 mg) added for the assembly, resulting part of the MMIPs material loss during the solid-liquid separation. Also, a higher mass of MMIPs material might cause serious aggregation effect, which hindered the elution of FQs with a fixed volume of elution solution. Thus, a lower FQs signal was obtained. Of course, more detailed material properties of the MMIPs and the spontaneous assembling mechanism of MNPs and MIPs will subject to our further studies. Matrix effects from highly complex samples are a great challenge on the quantitative analysis of AMS because of serious ion suppression. To achieve highly sensitive and selective determination of trace analytes in complex samples, coupling simple, rapid and sensitive sample pretreatment methods to AMS is a promising strategy to improve the performance of AMS . Raw milk is a typical extremely complex sample which cannot introduce to MS analysis directly. To address this problem, a facile method of solid-phase extraction based on magnetic molecularly imprinted polymers (MMIPs) combined with iEESI-MS was designed for the quantitative analysis of FQs in raw milk samples. The FQs molecules in the milk were selectively adsorbed by MMIPs and the MMIPs (together with the adsorbed FQs) was separated from the milk matrix. Thus, the majority of the milk matrix was cleaned up. Additionally, to avoid the milk residues interference, the separated MMIPs material was washed three times using 1 mL deionized water, acetonitrile, and 15% acetonitrile in deionized water (v/v), respectively. As a result, the matrix of the milk was largely cleared. The target analytes are sequestered by MMIPs and directly analyzed by iEESI-MS. Due to the highly selective extraction of MMIPs, ionic suppression is minimized; hence no chromatographic separation is necessary, which greatly increases analytical speed and sensitivity. Moreover, during the MS interrogation, CID experiments were carried for the suspected FQs ions, i.e., the FQs were identified based on their characteristic fragment ions, which practically avoid false positive result. Our results demonstrate that MMIPs-SPE-iEESI-MS enables direct quantification of sub-ppb level of FQs in raw milk samples without tedious sample pretreatments (e.g., centrifugation and chemical extraction). Furthermore, a comprehensive analytical performance comparison of the proposed MMIPs-SPE-iEESI-MS method with those of previous reported methods 46, in the analysis of FQs is presented in Table 4. The data showed that the method established in this work was of higher speed and better sensitivity than those previously reported methods. Combination of MMIPs-SPE with iEESI-MS was benefited by the high performance of MMIPs material in the capture of FQs from milk (i.e., fast and easy sample matrix clean-up step), as well as the specially designed sample loading/ionization process of iEESI. Molecularly imprinted polymers (MIPs) are a class of material engineered to bind one target compound or a class of structurally related compounds with high selectivity 57,58 . Due to the highly selectivity of MIPs, ionic suppression during ESI could be minimized, e.g., Figueiredo et al. employed MIP-SPE in ESI-MS for analysis of drugs in human plasma 59 . Although merits such as no chromatographic separation and minimized ionic suppression were achieved in the combination of MIP-SPE-ESI 59 , but tedious and laborious sample pretreatments including liquid-liquid extraction (for proteins elimination), centrifugation, preconcentration, sample re-dissolution, etc. were still needed on the account of highly complex of the plasma sample. In this respect, ambient ionization technologies provide a unique strategy for direct sampling/ionization analytes from the sample with no/minimum sample pretreatments 22,60,61 . Undoubtedly, the combination of facile sample pretreatment strategies (e.g., SPE, SPME, etc.) and ambient ionization methods is of promising when facing highly complex samples such as plasma and milk samples . iEESI belongs to the ambient ionization methods family, which has been developed as a direct and fast sampling and ionization method for mass spectrometric analysis of complex samples 41,62 . Combination of SPE method and iEESI is a promising strategy to improve the analytical performance of iEESI. In a previous study, coupling of magnetic solid-phase extraction with iEESI was developed to study 1-hydroxypyrene in undiluted human urine samples with the assistance of polypyrrole-coated Fe 3 O 4 magnetite nanocomposites (Fe 3 O 4 @Ppy nanocomposites) 42 . Due the low polarity of the coated polypyrrole on the surface of Fe 3 O 4 magnetite nanocomposites, chemicals of low polarity (e.g., 1-hydroxypyrene and 3-hydroxybenzeo[a]pyrene, etc.) were easily captured for subsequent iEESI-MS analysis. The selectivity of Fe Ppy nanocomposites is a notable drawback when treating chemicals with similar polarity 38,42,63 . To address this concern, highly selectivity and specificity could be introduced by MIPs material. Selectivity of MIPs is introduced during MIPs synthesis in which a template molecule, designed to mimic the analyte, guides the formation of specific cavities that are sterically and chemically complementary to the target analytes 64,65 . Strong retention is offered between a MIP phase and its target analyte(s) based on multiple interactions (e.g., Van der Waals, hydrogen bonding, ionic, hydrophobic) between the MIP cavity and analyte functional groups 64,65 . As a result, even trace FQs in the raw milk were captured and subsequently subject to iEESI-MS. To conclude, combination of fast and easy-to-use sample pretreatment with mass spectrometry is a promising strategy for high throughput quantitative detection of trace analytes in highly complex samples. As a typical example of the analytical strategy, MMIPs-SPE-iEESI-MS was designed for the confidently quantitative analysis of FQs in raw milk samples. As a result, FQs in the raw milk sample were selectively enriched by the MMIPs and then directly eluted by the electrospraying solvent to produce protonated FQs ions for mass spectrometric interrogation. The LOD of ≤0.03 µg L −1 and the high speed of 4 min for per sample were achieved. The analytical performance for real sample analysis was validated by a nationally standardized protocol using LC-MS, resulting in acceptable relative errors from −5.8% to +6.9% for 6 tested samples. Our results demonstrate that MMIPs-SPE-iEESI-MS is a facile method for the high throughput quantitative analysis of FQs in raw milk samples, which shows promising applications in food safety control and biofluid sample analysis. ## Methods Materials and chemicals. Water-compatible commercial molecularly imprinted polymers (MIPs) material, named SupelMIP TM SPE-Fluoroquinolones, was purchased from Sigma-Aldrich (St.Louis, MO, USA). Fe 3 O 4 magnetite nanocomposites (MNPs) were prepared according to previous studies 46,66 (detailed in Supplementary). Fluoroquinolones (norfloxacin, enoxacin, and fleroxacin, purity 98%) were purchased from J&K Scientific Ltd. (Shanghai, China). The individual stock solutions of norfloxacin, enoxacin, and fleroxacin were prepared in methanol at a concentration of 0.1 mg mL −1 and stored at 4 °C before use. Both Methanol and acetonitrile were HPLC grade and purchased from Merck KGaA (Darmstadt, Germany). Ammonium hydroxide solution (suitability for use in UPLC/LC-MS, NH 3 , w/w 20%) was bought from CNW Technologies GmbH (Düsseldorf, Germany). Ethylene glycol, ethylene diamine, ferric trichloride hexahydrate (FeCl 3 •6H 2 O), and sodium acetate (NaAc) were purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). Ultrapure water obtained from a Millipore water purification system (Milli-Q, Millipore; Bedford, MA, USA). Milk samples. Milk samples were purchased from local market and directly used in all the experiments without any pretreatment. In a prior trial of MMIPs-SPE-iEESI-MS, none of FQs such as norfloxacin, enoxacin, and fleroxacin were found in the blank milk samples. A series of standard solutions containing 0-500 µg L −1 of FQs including norfloxacin, enoxacin, and fleroxacin were prepared by serial dilution from 0.1 mg mL −1 stock solution of FQs in methanol. Note that, to ensure mixed evenly, all the FQs-spiked milk samples were vigorously shaken using a test tube shaker (2800 rpm, Lab Dancer S25, IKA, Germany) before MMIPs-SPE-iEESI-MS analysis. ## SPE based on MMIPs coupled with iEESI-MS analysis. The schematic illustration of MMIP-SPE-iEESI-MS for quantification of FQs was shown in Fig. 5. MMIPs were obtained from a simple co-mixing procedure according to the previous literature 46 , i.e., 2.0 mg Fe 3 O 4 magnetite nanocomposites (MNPs) and 1.5 mg molecularly imprinted polymers (MIPs) were co-mixed in 1.0 mL methanol by vigorously vortexing for 1 min in a 5-mL glass vial. Then, the methanol was removed from the MMIPs with the assistance of an external magnet and residues of methanol in the MMIPs material volatilized away after about 1 min. The obtained MMIPs were used for the extraction of FQs from milk samples. A 2 mL aliquot of raw milk sample was added into the 5-mL glass vial containing MMIPs material (3.5 mg) and vortexed for 1 min. The suspension mixture was loaded in a 1 mL syringe (Hamilton company, Nevada, USA), and MMIPs captured with FQs were magnetically gathered to the inner wall of the syringe with an external magnet. The milk waste was discharged into a glass beaker. After twice repeats of the MMIPs collection, all the FQs captured MMIPs were gathered on the inner wall of the syringe. To avoid the milk residues interference during the ionization of FQs, the FQs captured MMIPs inside the syringe were washed using 1 mL deionized water, acetonitrile, and 15% acetonitrile in deionized water (v/v), respectively. After loading with 100 μL extraction solution (2% ammonia in methanol, w/w), the syringe was shaken for 20 s to allow the FQs eluted to form a FQs solution which is suitable for electrospray purpose. The FQs solution was pumped through a capillary for ESI at flow rate of 8 μL min −1 , a strong magnet was placed outside of the capillary to prevent the MMIPs material from reaching to the ESI nozzle. Thus, all the MMIPs material was purposely held by the external magnet and no particles reached to the ion entrance of the mass spectrometer instrument. The MS/MS signal collection duration was 1 min and the average signal intensities of fragment ions were selected from a 30 s window. The average signal intensities of fragment ions of m/z 276, m/z 277, and m/z 326 were selected as analytical response to establish the quantitative method for norfloxacin, enoxacin, and fleroxacin, respectively. It is noted that the lifetime of the MMIPs was about 3 times performances of MMIPs-SPE-iEESI-MS and the performance would decrease significantly over 3 times due to the matrix contamination. All the experiments were carried out using an Orbitrap Fusion ™ Tribrid ™ mass spectrometer (Thermo Scientific, San Jose, CA, USA). Mass spectra were collected at the mass range of m/z 50-500 under positive ion detection mode. The electrospray solution was pumped at a flow rate of 8 μL min −1 using a syringe pump (Harvard Apparatus, Holliston, MA, USA). The ionization voltage was set at +3.0 kV, and the heated LTQ capillary was maintained at 250 °C. The pressure of nitrogen sheath gas was 60 Arb. CID experiments were carried out for MS/MS analysis. During the CID experiments, precursor ions were isolated with a window width of 1.0 Da, and normalized collision energy (NCE) was set to 30-40%. Other parameters were set to instrument default values. Scanning electron microscopes (SEM) and energy dispersive X-ray analysis (EDX) were performed to investigate the morphological, size and elements of MIPs, MNPs, and MMIPs materials by the FIB-SEM instrument (Helios Nanolab 600i from FEI Co., USA). The electron beam and working distance of the instrument were set to 10-20 kV and 4 mm, respectively.

chemsum

{"title": "Internal Extractive Electrospray Ionization Mass Spectrometry for Quantitative Determination of Fluoroquinolones Captured by Magnetic Molecularly Imprinted Polymers from Raw Milk", "journal": "Scientific Reports - Nature"}

advances_in_automated_transition_state_theory_calculations:_improvements_on_the_autotst_framework

## Abstract: Kinetic modeling of combustion chemistry has made substantial progress in recent years with the development of increasingly detailed models. However, many of the chemical kinetic parameters utilized in detailed models are estimated, often inaccurately. To help replace rate estimates with more accurate calculations, we have developed AutoTST, an automated Transition State Theory rate calculator. This work describes improvements to AutoTST, including: a systematic conformer search to find an ensemble of low energy conformers, vibrational analysis to validate transition state geometries, more accurate symmetry number calculations, and a hindered rotor treatment when deriving kinetics. These improvements resulted in location of transition state geometry for 93% of cases and generation of kinetic parameters for 74% of cases. Newly calculated parameters agree well with benchmark calculations and perform well when used to replace estimated parameters in a detailed kinetic model of butanol combustion. ## Introduction Detailed kinetic models allow researchers to understand the chemistry of complex phenomena in systems such as combustion and hetrogeneous catalysis, thus enabling them to make informed experimental design choices, and to design and optimize processes and devices. Microkinetic models often contain hundreds of intermediates and thousands of reactions, for which thermodynamic and kinetic parameters need to be specified . These parameters are ideally determined experimentally or calculated theoretically with high accuracy, but most are estimated . These estimations allow parameters to be determined quickly, but usually with less fidelity . Thermochemistry estimates are often derived from Benson's group additivity, where groups of atoms with known thermochemistries are summed . These estimates are reasonable for most situations, but have been difficult to extend to some cases such as polycyclic species , motivating automated quantum mechanical or semi-emperical calculations . When estimating the kinetics of a reaction, the Evans-Polanyi relationship can be used to estimate kinetics based on the enthalpy change of a specific reaction , if the rates of similar reactions are sufficiently well known. Alternatively, group contribution methods can provide kinetic estimations in a similar fashion to Benson's additivity methods . Both Evans-Polanyi relationships and group additivity methods are fast and easily automated, which is especially useful in the generation of microkinetic mechanisms. Unfortunately, estimations fall short when exploring novel systems when the estimation rules are poorly known due to a lack of training data. In these cases, rule-based methods use less appropriate rules, and group-based methods utilize less specific group values, leading to errors in the rate as large as several orders of magnitude. With growing computational power, calculating accurate kinetic parameters through transition state theory (TST) is no longer infeasible. However, TST calculations require a trained guess of transition state (TS) geometries, and often manual entry to arrive at trustworthy parameters. Given the number of reactions that are present in a detailed combustion model, these quantum calculations need to be automatized. Automatizing TST calculations has been the focus of many . This paper focuses on recent improvements to the AutoTST framework first developed by Bhoorasingh and co-workers . AutoTST is an automated algorithm to locate reactant, product, and transition state (TS) geometries using quantum chemical calculations, to arrive at reaction rate expressions. AutoTST was originally built as a module within the Reaction Mechanism Generator (RMG) software and could determine modified Arrhenius parameters from RMG reaction objects matched to one of three specific reaction families: unimolecular hydrogen migration (1 reactant to 1 product), radical addition to a multiple bond (2 reactants to 1 product), and bimolecular hydrogen abstraction (2 reactants to 2 products). Bhoorasingh noted that this was a sizable step in automated kinetic calculations, but the workflow needed some improvements. This work addresses these improvements to increase fidelity and speed of calculations, such as including a detailed conformer search, 1-D hindered rotor approximations, graph based symmetry number calculations, and parallelization of calculations using the Simple Linux Utility for Resource Management (SLURM) job scheduler. We observe these changes by recalculating reaction rates present in the Lawerence Livermore National Lab's (LLNL) butanol model by Sarathy and co-workers , comparing to benchmark calculations, and assessing the impact on iginition delay time predictions of the whole model. ## Algorithm The first generation of AutoTST is described by Bhoorasingh and co-workers . This paper underscores important aspects of the original framework and highlight improvements. An overall updated workflow is described in Fig. 1a with five broad steps. First, initial geometry estimates are created (section 2.3) for both the reactants (using the distance-geometry and force-field methods in RDKit ) and the TS (using the original AutoTST algorithm to modify the bounds matrix ). Then an ensemble of conformers are generated for each structure (section 2.4). The conformers are then processed in parallel to optimize the geometries (section 2.5) and validate them (section 2.6) to ensure the desired molecule or TS has been found. Finally, temperature-dependent kinetics are calculated (section 2.7), including a correction to estimate the effect of hindered internal rotors. Figures 1b, 1c, and 1d are sub-workflows for the species, TS, and kinetics, that are performed during the overall workflow. ## Inputs AutoTST originally required users to provide reactions matched to one of three supported reaction families, or templates, present in RMG , and the electronic structure calculation settings resulted in long and complex input scripts. The user interface has now been updated, making inputs more straightforward. Users no longer have to provide matched, templated RMG reactions -the workflow will now automatically match the reaction of interest to one of the supported reaction families and identify the reacting atoms. Current supported reaction families are hydrogen abstraction, radical addition to multiple bonds, and intramolecular hydrogen migration (Fig. 2). A user will supply a reaction of interest as an instance of an RMG Reaction class, or in a simple text string formed from the reactants and products in SMILES format (e.g. ). If unable to match the input reaction to a supported reaction family, AutoTST will return an error. Reactions matched to reaction templates are then used to create three dimensional TS geometries, described in subsequent sections. Users are also required to specify electronic structure calculation settings, such as functional and basis set, and SLURM settings (e.g. username, partitions, accounts, excluded nodes) to enable AutoTST to perform calculations in parallel on computer clusters. ## Representation of Species The newest version of AutoTST uses the Atomic Simulation Environment (ASE) , RD-Kit , and RMG software packages to generate Species and Conformer objects that can be easily manipulated. A Species object requires a user to provide one or more SMILES strings. If multiple SMILES strings are provided, AutoTST will check that they are resonance isomers of each other; if only one is provided then additional resonance structures will be generated using RMG. Once the list has been vetted, the Species object can generate Conformer objects for each resonance structure. Species objects act as a hierarchical class to organize one or multiple Conformer objects that represent the 3D geometry of a species. generate_conformers may be called to perform a systematic conformational analysis to create an ensemble of low energy conformer objects using an ASE electronic structure calculator. The conformer generation workflow is described in detail in the following sections. Just as Species and Conformer objects are necessary for the representation of species and their conformers, AutoTST Reaction and TS objects are necessary to organize TSs and their conformers. Reaction objects represent the chemical reaction, while the TS objects act as individual conformers for the reaction's transition state. Discussion of how TS geometries are constructed is described in the following section. ## Creating Initial 3D Geometries Initial species and TS geometries are generated as they were in Bhoorsingh and co-workers' original workflow . For stable species, the embed feature in RDKit is used to generate reasonable geometry guesses; TS geometries require additional treatment. For TSs, reacting atoms are identified using the matched reaction family. In the case of hydrogen abstraction, these reacting atoms represent the abstracted hydrogen, the atom bound to the abstracting hydrogen, and the radical atom abstracting the hydrogen ( 2 H, 1 R, and 3 R• in Fig. 2, respectively). The TS complex is then passed to a hierarchical decision tree that provides guesses of key distances between reacting atoms based on the reaction family and functional groups near the reaction center. The decision tree is descended to find the node where the functional group and its proximity to the reaction center closely matches the TS of interest. This node then provides the distances between the reacting atoms and serves as our "key distances". A 3D geometry of the TS complex is created using the geometry embedding feature in RDKit to generate a distance matrix. The distance matrix is a square matrix that describes the maximum and minimum allowable distances between pairs of atoms. The entries in the distance matrix are edited such that distances between reacting atoms are specified using the "key distances" determined by the decision tree. A new constrained TS geometry is generated by RDKit using the edited distance matrix. ## Conformer Analysis The original version of AutoTST would generate species and TS geometries at random using RDKit and would not ensure that AutoTST found the lowest energy, or most probable, conformation for each geometry . AutoTST now performs a systematic conformer search on species and TSs by considering all dihedral angles not in a ring or containing a terminal methyl group, all invertible double bonds, and all chiral centers. Possible geometries are generated by creating all combinations of dihedral angles (range 0°to 300°with a 60°spacing), invertible double bonds (E vs Z configuration), and invertible chiral centers (R vs S configuration). These initial geometries are optimized using Hotbit , a density-functional tight-binding calculator, and the BFGS optimizer provided in ASE with a maximum of 1,000 optimization steps. Through this workflow, all optimizations are to energy minima where species optimizations are completely unconstrained but TS geometries undergo a constrained optimization with the distances between reacting atoms fixed. Optimized conformers are compared against the initial geometry to ensure that isomorphism is maintained. Isomorphic conformers are then compared to identify unique conformers by asserting the average root mean square deviation between all other conformers is greater than 0.5 . All unique conformers within a specified energy cutoff of the lowest energy conformer are further optimized using DFT. For this study, we used 10 kcal/mol as our energy cutoff. ## Optimize Geometries Both versions of AutoTST relax species to minima but handle transition state geometry optimizations differently . After determining the initial TS geometry from a group-estimation tree, the original workflow would perform three consecutive geometry optimizations on the TS. The first optimization froze distances between reacting atoms, while relaxing all others to an energy minimum. The second optimization froze distances between non-reacting atoms and relaxed all others to a saddle point. Lastly, the entire geometry is relaxed to a saddle point. To reduce computational costs, the new workflow skips the reaction center optimization. ## Validate Geometries The original AutoTST would validate TS geometries by performing intrinsic reaction coordinate (IRC) calculations and compare the output geometries against the input reactants and products. If the geometries matched, then the TS is validated. These calculations were the bottleneck of the previous workflow and needed to be improved. Inspired by the procedure of Van de Vijver in Genesys , rather than performing an IRC calculation on each TS identified, a vibrational vetting step is performed and followed by an IRC calculation if vetting is inconclusive. This vibrational vetting step entails reading in the log file of the saddle point and ensuring that there is: 1) only one negative frequency, 2) the distance between reacting atoms is less than two , and 3) the change in bond length by vibrational translation for reacting bonds is an order of magnitude greater than the change in bond lengths of non reacting bonds. If this vetting step is passed, no IRC calculations are performed because it is assumed the saddle point corresponds to the reaction of interest. If this vetting step is not passed, an IRC calculation is performed and used to validate the saddle point. ## Kinetics Estimation AutoTST uses the software package Arkane to estimate kinetic parameters . Arkane is bundled with RMG , and can be used for pressure-dependent rate calculations but here is used for canonical TST calculations Following traditional TST, it is assumed that the reactants and the TS are in a quasi-equilibrium state in a vacuum and the rate limiting step is the transition from the TS to the products. From these assumptions, a modified form of the Eyring equation (equation 1) is used to relate the thermodynamic properties of the TS and reactants to the elementary rate of reaction : where k B is the Boltzmann constant, T is the temperature, h is Plank's constant, R is the gas law constant, κ is the correction faction for quantum tunneling, and ∆G ‡ is the change in Gibb's energy between the TS and the reactants. Factors such as the change in Gibb's free energy and tunneling correction are conformer dependent and, as such, will be different in the new AutoTST workflow. ## Symmetry Number Calculations Symmetry numbers are a measure of the number of indistinguishable orientations a molecule or TS geometry can have . Symmetry numbers were previously calculated using the SYMMETRY package which uses a 3D based approach . However, symmetry numbers were often under estimated because the 3D comparison would often break symmetry for minor deviations. To combat this, we removed the SYMMETRY package from our workflow and now use the point group calculator included in Arkane. This allows Arkane to automatically calculate symmetry numbers when calculating kinetic and thermodynamic parameters, based on an analysis of the molecular graph or connectivity. ## Hindered Rotor Correction AutoTST previously used the rigid-rotor harmonic oscillator (RRHO) approximation when calculating kinetics, but this approximation can be inaccurate for geometries with internal rotors . For most cases, the one-dimensional hindered rotor (1DHR) approximation is a more accurate representation of the internal rotation of stable species and TSs. We have added a modified 1DHR workflow using the approximation provided by Cohen to account for internal rotation, without performing 1DHR scans as these are difficult to automate for saddle points. This correction is performed on kinetics after they have been calculated using Arkane . First, we identify all rotatable torsions that are not present in a cycle, and for each rotor, we count the minimum number of substituents on either end of the rotor. This number is used to estimate the barrier to internal rotation using estimates from Benson (Table 1). Number of substituents V , kcal/mol 0 0.0 1 1.1 2 2.2 3 3.5 Table 1: Estimated barrier heights to internal rotation based on number of substituents. The reduced internal moment of inertia and the internal symmetry of the torsion are calculated using methods available in RMG and these, with the barrier height, are used to calculate the approximate vibrational frequency using Eq. ( 2) and the free rotor partition function using Eq. ( 3): (2) In Eq. ( 2) and Eq. ( 3), σ int is the internal symmetry of the torsion, I is the reduced internal moment of inertia, V is the barrier height, R is the gas law constant, k B is the Boltzmann constant, T is the temperature, and h is Planck's constant. Values for vibrational frequencies, barrier heights, and free rotor partition functions of each rotor are used with Table 2 and Table 3 to interpolate the vibrational and the hindered rotor contribution to the rate constant, k vib and k h.i.r. respectively. For each rotor of each reactant and TS, the ratio of these numbers is calculated over temperatures ranging from 298-2500 K and is used to correct the rate constant calculated by the RRHO approximation: where k RRHO is the temperature dependent rate constant calculated using the RRHO approximation, k h.i.r,i is the 1DHR contribution to the rate constant for rotor i, k vib,i is the RRHO contribution to the rate constant for rotor i. For reactants, the product of these ratios over N reactants number of reactants is taken. Modified rates are fit to a three-parameter Arrhenius expression and returned to the user. ## Methods To test the efficacy of recent changes in AutoTST, Sarathy and coworker's model for the combustion of butanol was revisited . To assess our changes, we wanted to study three distinct categories: 1. Success rate: how many reactions were we able to obtain TSs and kinetics for? 2. Micro-effects: how do AutoTST calculated rates compare to a set of benchmarks? 3. Macro-effects: how do AutoTST calculated rates impact an observable like ignition delay? We attempted calculations on reactions present in the previous model used for the original AutoTST study, compared these calculations to benchmarks, and used these calculations to observe the change in ignition delay. These steps are described in detail in subsequent sections. ## Success Rate To assess the improvements made in AutoTST, we used the same set of 1117 reactions used in the original AutoTST paper plus three new hydrogen abstraction reactions that we were able to find using an updated version of RMG. These reactions come from Sarathy and coworker's model for the combustion of butanol and are summarized in ## Model Chemistry Reactions calculated from the original AutoTST paper were performed using the M06-2X functional with the MG3S basis set. For this study, calculations were performed using the using the M06-2X function with the cc-pVTZ basis set. Different model chemistry was used because Arkane no longer supports the MG3S basis set. ## Benchmark calculations As part of the validation methods from the original AutoTST paper, Bhoorsingh and co-workers performed a series of benchmark calculations on six reactions shown in Table 5 . These reactions represent two reactions from each of the three supported reactions families where the AutoTST rate expression disagreed with RMG predicted expressions by more than a factor of 100 at 1000 K and 1 bar. These benchmark calculations were performed using the TS calculated from the original Au-toTST workflow, but with 1-D hindered rotor calculations performed on each rotatable dihedral. If a lower energy conformer was identified during a scan, the lower energy conformer was optimized and 1-D hindered rotor scans were restarted on the new conformer. Single point energies were calculated using ORCA at the CCSD(T)-F12/RI method with the cc-VTZ-F12 and the cc-VTZ-F12-CABS basis sets. In addition, point groups for symmetry numbers were determined by hand. These additional treatments were used to calculate rate expressions with Arkane and are referred to as the "Benchmark" calculations. In this work, these benchmark calculations were compared to the original AutoTST, LLNL, and new AutoTST kinetic parameters and discrepancies were measured. ## Generation of alternative models To observe the impact of AutoTST calculations, we generated alternate models using our newly calculated AutoTST kinetics. Kinetics that were successfully calculated by the improved workflow were swapped into the LLNL butanol model one at a time to generate alternate models. E.g. if we calculated kinetics for reaction X, we generated an alternative model by swapping in kinetics calculated by the original workflow and the new workflow. These models were tested against four sets of ignition delay data using PyTeCK (described in the following section) to quantify error of our alternative models against experiments. ## PyTeCK PyTeCK, Python tool for Testing Chemical Kinetics , was used to quantify the error in theoretical models against experimental data. PyTeCK works by reading in experimental data in the human-and machine-readable ChemKED, Chemical Kinetic Experimental Data, format . PyTeCK will read the experimental conditions from a ChemKED file, perform the simulation using Cantera at the experimental condition, and return the error between the simulated value and the experimental value as described in equations 5 and 6. PyTeCK is currently limited to ignition delay experiments. In equations 5 and 6, E j is the average error for the jth data set with n data points, τ exp i is the experimental ignition delay measured at conditions i, τ sim i is the simulated ignition delay at conditions i, σ exp i is the uncertainty of the the experimental measurement for the ith data point, E tot is the total error over m data sets. PyTeCK was used to compare alternative models against four sets of experimental summarized in Table 6. Author Name Year Isomers Studied Temperature (K) Pressure (atm) Equivalence Ratios (φ) Moss et al. 2008 n-, 2-, iso-, tert-1200 -1800 0.99 -3.95 0.5 -1 Stranic et al. 2011 n-, 2-, iso-, tert-1050 -1600 1.5 -43 0.5 -1 Zhu et al. 2013 n-700 -1100 20 -40 0.5 -2 Bec et al. 2014 2-, iso-, tert-800 -1100 20 -30 0.5 -1 Table 6: A summary of the conditions for the ChemKED data utilized when measuring using PyTeCK. ## Workflow Efficacy One goal of this work was to improve the success rate of AutoTST, which is measured here in two ways: (1) observe the percentage of reactions where AutoTST found a TS and (2) observe the percentage of reactions where AutoTST arrived at a rate expression. These results are summarized in Table 7 The updated AutoTST workflow was able to find validated TS geometries for 1041 of the 1120 reactions tested. This high success rate was independent of reaction family and is most likely attributed to the systematic conformer search. By considering many potential low energy conformers, AutoTST had a greater chance of finding at least one validated saddle point, resulting in an increased success rate for a TS search. Of the 1120 test reactions, 79 were unable to arrive at a saddle point. When diagnosing the source of these errors, we found two main causes: convergence errors and validation errors. 24 of these failures were convergence errors, which occur when Gaussian optimization is unable to find at least one saddle point. These errors are caused when the optimization is unable to fix distances between reacting atoms, often because of the linear-like configuration of some TS geometries. This could be remedied by adding a fixed dummy atom as a reference point for constraints similar to the constrained optimizations that are performed in Cavolitti and Klippenstein's EStokTP . Alternatively, convergence errors occur when a Gaussian optimization does not meet a convergence criteria or the optimization runs out of iterations. Adding additional optimization steps or loosening convergence criteria could help reduce the number of convergence errors and are recommended for future work. The remaining 55 failures were validation errors that occurred when AutoTST was able to arrive at a saddle point but it did not corresponded to the reaction of interest. A different trend is noticeable when observing the percentage of reactions where AutoTST found a rate expression. For our re-run, calculations were attempted on 1120 reactions in the LLNL butanol model and rate expressions were found for 833 reactions or 74.4% -a small increase compared to the original workflow. In the original workflow, Bhoorasingh and co-workers found the success rate was independent of reaction family , but here, there is an inverse correlation between the success rate and the number of reactants and products. Hydrogen abstraction reactions (two reactants and two products) had a success rate of 70.9%, radical addition reactions (two reactants and one product) had a success rate of 83.2%, and hydrogen migration reactions (one reactant and one product) had a success rate of 92.3%. There were 208 failures when calculating kinetics from an apparently successful TS optimization. 197 of these failures were barrier height errors (186 from hydrogen abstraction and 11 from radical addition reactions). These failures occur when either the forward or reverse barriers in the Eckart model are negative, indicating that the energies of the reactants or products is greater than the TS. This can be because the TS connects to two van der Waals (vdW) wells rather than the bimolecular entrance or exit channels. These wells allow for intermolecular interactions to occur between reacting species that lower the potential energy of the complex, leading to a saddle point that can be submerged below the reactants or products. These errors were more common in abstraction reactions with oxygen-containing reactants (e.g abstraction by O --O, • OH, • OOH) which have strong vdW interactions. These errors might be addressed by treatment of vdW wells with a 3TS model solved with a Master Equation (ME), as is done in EStokTP . These barrier height errors account for almost all of the failures in the hydrogen abstraction reaction family. The remaining 11 errors are attributed log parsing errors that occur when AutoTST interfaces with Arkane. These errors can be remedied through development of Arkane, RMG, and AutoTST in tandem. ## Benchmark Calculations We revisited the six benchmark calculations that were performed in the original AutoTST study . This set of reactions consists of two representatives from each reaction family supported by AutoTST. For each of these reactions, the rate coefficients estimated by RMG and from the original AutoTST disagreed by an order of at least 100 when calculated at 1000 K. Arrhenius plots were generated using rate expressions from this work, the original AutoTST study, the original butanol model, and the benchmarks from the original AutoTST study in Fig. 3. From Fig. 3 we see good agreement between newly calculated rate expressions and the bench-marks. To quantify the agreement, a discrepancy measurement was defined as: where d i (T ) is the discrepancy between the benchmark rate expression, k 0 , and the rate expression for the ith source, k i , at temperature T . This discrepancy represents an order of magnitude difference between a rate expression and the benchmark at a particular temperature. The average difference between the two rate expressions is calculated over temperatures between 300 K and 2000 K with 100 K spacing. Table 9 shows the average and standard deviations in discrepancies between kinetics from the original workflow, the updated workflow, and the LLNL butanol model. A high standard deviation indicates a difference in apparent activation energy. 9: Discrepancies between kinetic parameters and benchmark calculations from . In five out of six benchmark reactions, the updated workflow had the lowest average discrepancy overall. Reaction 2 was the only exception where rate coefficients from the original workflow performed the best. In all cases, the updated workflow performed better than the LLNL butanol model. When taking into account the variability of the discrepancy measurement, it is difficult to conclude that any one set of kinetics performed the best. However, it is clear that the modifications have resulted in more accurate kinetics as a whole. ## Comparison against experimental data We generated 833 alternate models for the combustion of butanol by swapping in our newly calculated kinetic parameters one at a time into the LLNL kinetic model. The error of these alternate models and the unmodified model were measured with PyTeCK and the difference (∆E) is reported in Fig. 4. In Fig. 4, a majority of the data fall in a symmetric distribution centered about ∆E = 0, where 788 of the 833 models shown (or 94.6%) have small changes in error between -1 and 1. Reactions with the largest decrease or increase in error are summarized in Tables 10 and 11, respectively. These tables also summarize the absolute difference in log 10 k at 1000 K and 100 bar. In addition, reactions with the greatest disagreement in log 10 k are reported in Table 12. Table 11: The top five reactions where the error was increased the most and their kinetic sources in the LLNL butanol model. For the reaction rate substitutions that decreased prediction error the most ( Table 12: Five reactions where the rate constants between AutoTST and LLNL differed the most at 1000 K and their kinetic sources in the LLNL butanol model. in enthalpy for the reaction and a tabulated ring strain of the TS to estimate activation energy, and literature sources to estimate pre-exponential factors. The "Approximations" used for reactions 1187 and 1194 indicate that parameters came from educated guesses by the creators of the model. In these cases, the error was reduced by replacing kinetics from these rules and approximations with the AutoTST calculations. This highlights that some reactions are particularly sensitive in this model: For reactions 1187 and 1929, the error decreased noticeably because of a change in one reaction rate of less than one order of magnitude, so these reactions ought to be studied carefully. In addition, the remaining reactions in Table 10 may benefit from a more thorough study: these rate coefficients changed by less than two orders of magnitude but still caused a noticeable decrease in error. It is also understandable that in some cases AutoTST-calculated parameters increased error. LLNL kinetics from four out of the five reactions where AutoTST increased prediction error the most (Table 11) come from detailed computational studies of these reactions . All of these studies used DFT to find the geometries and vibrations, coupled cluster methods to determine accurate single point energies, and DFT to perform 1DHR scans, and for all cases the kinetics were compared to either benchmarks, experimental data, or previous calculations to assess that parameters were calculated accurately. It is prudent to say that kinetic parameters from a thorough computational study would best any automated rate calculation, so the increase in error when substituting AutoTST rates is not surprising. revisited, or at least updated before they are used as the basis for new modeling studies. In addition, the increases in error seen in Table 11 are drastically higher than the decreases in Table 10 but the change in calculated rate expression is smaller on average. This signals that in these cases (reactions in Table 11), ignition delay is more sensitive to these rate expressions. ## For reaction CH Finally, we investigated the sources of kinetics that disagree the most (Table 12) irrespective of their impact on prediction error ∆E. Reactions 1123, 1429, and 338 were approximations so it is reasonable that AutoTST would disagree by many orders of magnitude -probably indicating that the estimation was poor. However, we changed these rate expressions by over nine orders of magnitude and saw almost no change in error, meaning that the mechanism predictions are not sensitive to these reactions. Reaction 386 was was changed by almost 12 orders of magnitude, leading to a slight increase in prediction error. The details of the original rate calculation are not clear, so we are unable to comment on why the difference is so large. Lastly, AutoTST disagreed with reaction 278 by over 15 orders of magnitude. The LLNL butanol model uses the pressure-dependent rate expression from a computational study by Lee and Bozzelli , accounting for the multiple well reaction network. The AutoTST calculated rate agrees much more favorably (∆ log 10 k of 2.1 at 1000 K) with their high pressure limit rate expression. This serves as a reminder that AutoTST provides only high-pressure limit rates of elementary reactions, which should be used as input to a pressure-dependent Master Equation solver (such as Arkane, included with RMG) for fall-off, chemically activated, and multi-well systems. These results highlight that kinetics calculated from AutoTST perform as expected when utilized in a detailed kinetic model -they perform poorly in comparison to parameters from careful computational studies, but improve upon kinetics that come from estimates or approximations. They also improve upon calculations from the first version of AutoTST. ## Conclusions Through this work, improvements have been made to AutoTST which include: rewriting of the code to make user inputs simpler (as well as developing, debugging, and testing), inclusion of a systematic conformer search, modification of how symmetry numbers are calculated, and addition of a 1-D hindered rotor treatment. These changes were measured by attempting automated calculations on 1120 reactions present in the LLNL model for the combustion of butanol , the model investigated in the original AutoTST study . Of the 1120 reactions, 858 were bimolecular hydrogen abstraction reactions, 78 were unimolecular hydrogen migration reactions, 184 were radical addition reactions. The updated version of AutoTST was able to find TS geometries for 93%, or 1041 reactions: 797 (93%) for hydrogen abstraction, 73 agreed the most with the benchmarks) and for all six reactions, the updated version of AutoTST had a lower discrepancy than LLNL parameters. We also observed the effect of our calculated parameters on ignition delay. By utilizing PyTeCK, we saw the impact of our calculations when implemented in a detailed kinetic model. In 94% of cases our modified kinetic parameters had a negligible impact on ignition delay when applied individually. AutoTST was able to decrease the error for reactions that were sourced from either approximations or rate rules, but increased the error for reactions that came from computational studies, which is to be expected. AutoTST is designed to efficiently provide on-the-fly kinetic parameters that perform better than estimates, but will most likely not outperform thorough computational studies. When comparing LLNL and updated AutoTST kinetics there are some kinetic parameters that disagreed by over 10 orders of magnitude. In most of these cases the LLNL parameters originated from approximations so it is likely that the parameters from AutoTST are more accurate. However, they have little impact on ignition delays for this system. Although many improvements have been made to AutoTST, there are still more to be made. The systematic conformer search is adequate for small species and TSs but is too computationally expensive for larger complexes, so stochastic methods should be investigated. In addition, the hindered rotor treatment in AutoTST could be improved. Proper rotor scans should be performed to accurately account for internal rotor effects in both species and saddle points. High fidelity singlepoint energy calculations could increase accuracy. Finally, treatment of vdW wells and multiple TSs would improve the calculation of abstraction reactions. In addition to the developments of AutoTST, there are a number of future research directions that can make use of AutoTST or automated rate calculators. AutoTST can be used to help researchers identify and resolve errors in detailed kinetic models by (1) finding discrepancies in kinetic parameters and (2) utilizing updated parameters in a detailed kinetic model to bring it closer to the Chemical Truth. Finally, using AutoTST in tandem with an automated reaction mechanism generator like RMG to create kinetic models, where a majority of rates are currently estimated, would be fruitful.

chemsum

{"title": "Advances in automated transition state theory calculations: improvements on the AutoTST framework", "journal": "ChemRxiv"}

a_transient_directing_group_strategy_enables_enantioselective_reductive_heck_hydroarylation_of_alken

## Abstract: Metal-coordinating directing groups have seen extensive use in the field of transition-metalmediated functionalization of alkenes; however, their waste-generating installation and removal steps limit the efficiency and practicality of reactions that rely on their use. Inspired by developments in the field of C-H activation, herein we report a transient directing group approach to reductive Heck hydroarylation of alkenyl benzaldehyde substrates that proceeds under mild conditions. Highly stereoselective migratory insertion is facilitated by in situ formation of an imine from catalytic amounts of commercially available amino acid additive. Computational studies reveal an unusual mode of enantioinduction by the remote chiral center in the transient directing group.Main Text: Among the foremost challenges in modern synthetic chemistry is forming carbon-carbon bonds to establish stereocenters remote from existing functional groups. Catalytic, enantioselective alkene functionalization has emerged as an attractive strategy owing to the widespread accessibility and unique reactivity profile of alkenes. Within the contemporary alkene functionalization toolkit, palladium-catalyzed Heck-type arylations (1-4) constitute an especially powerful mode of reactivity as a result of their broad substrate scope and functional group compatibility. Nevertheless, significant limitations in Heck-type chemistry remain, including controlling regioselectivity with multi-substituted, electronically neutral alkenes; suppressing β-hydride elimination to enable interception with additional reaction partners, and achieving stereoinduction during intermolecular migratory insertion (5-7). The use of directing groups (DGs), functional motifs containing one or more binding sites capable of facilitating efficient intramolecular delivery of a reagent or catalyst, is a classical strategy in organic synthesis for controlling reactivity and selectivity (8). In recent years, removable DG strategies have been successfully applied to various alkene functionalization reactions (9), including classical Mizoroki-Heck alkene arylations (10,11). In the realm of difunctionalization and hydrofunctionalization of acyclic internal alkenes, removable bidentate DGs, such as 8-aminoquinoline-derived amides, have proven especially valuable due to their ability to suppress β-hydride elimination and allow for subsequent transmetalation/ligand exchange (12,13). While reactions involving removable DGs are intrinsically valuable, they are limited by the fact that the DG needs to be installed and cleaved, requiring a minimum of two concession steps. Additionally, reactions using DGs are difficult to render enantioselective owing to a lack of available coordination sites on the metal for a chiral ligand (14,15). Developing co-catalytic chiral transient directing groups (TDGs) to enable these types of transformations is an exciting prospect but is fraught with challenges. An appropriate TDG must interact in a highly selective, yet reversible, manner with a native functional group on the substrate, such as an alcohol, aldehyde, or amine (Fig. 1B). Additionally, this TDG needs to be chemo-orthogonal to the catalytic reaction of interest, such that both catalytic cycles are not inhibited or perturbed to instead form undesired side products. The viability of catalytic TDGs has previously been established in several mechanistically distinct transition-metal catalyzed reactions, including in the field of C-H activation (16)(17)(18)(19)(20)(21)(22)(23); however, there are (26) and Tan (27) independently reported transiently directed hydroformylations of homoallylic and bishomoallylic alcohols using phosphinite and amino phosphine TDGs, respectively. Tan subsequently rendered this approach enantioselective with allylic amine substrates (Fig. 1c, middle) (28). A related alkene functionalization strategy using catalytic aldehyde TDGs was reported by Beauchemin and coworkers to effect metal-free Cope-type hydroaminations (29). More recently, Yu and coworkers have successfully developed amino acid TDGs to promoted enantioselective benzylic C-H activation with palladium(II) (Fig. 1c, right) (30,31). These advances notwithstanding, application of a chiral TDG strategy in versatile palladium(0)catalyzed Heck-type coupling remains unknown. ## B Transient directing group (TDG) strategy for alkene functionalization (1 step) We envisioned that it would be possible to use a chiral amine TDG to facilitate stereocontrolled Heck-type migratory insertion with an alkenyl aldehyde substrate. In particular, we hypothesized that a reductive Heck system (32,33) would tolerate the presence of the free amine and water and could thus be adapted to operate synergistically with a TDG cycle; moreover, the mild reaction conditions typical of these reactions were expected to curb undesired side reactions. After migratory insertion, the TDG would suppress β-hydride elimination to enable interception with a third reaction partner, a hydride source. Successful implementation of this strategy would offer a significant advantage over aforementioned chiral auxiliary based strategies and would serve as a complementary method to elegant work on asymmetric redox-relay Heck reactions reported by Sigman that achieve alkene hydroarylation via a chain walking cascade that formally oxidizes a remote alcohol to a carbonyl (34)(35)(36)(37). To reduce this general idea to practice, we selected (Z)-2-(prop-1-en-1-yl)benzaldehyde as a model substrate based on the fact that alkyl-substituted internal styrenes are typically challenging substrates with low reactivity in intermolecular palladium-catalyzed reductive Heck coupling (33). After a brief unsuccessful survey of commercially available chiral amines, we turned to amino acids, taking inspiration from the pioneering work of Yu (30,31). We screened a variety of conditions and amino acids for competency in the reaction, with L-tert-leucine giving both high yields and enantioselectivity (see SI for details). Conveniently, the reaction conditions are mild and operationally simple. Setup can be done in open air using micropipettes to dispense most of the reagents, and solvent dryness did not have a major impact on the reaction outcome. Notably, the use of the analogous ketone substrate (1-(2-(prop-1-en-1yl)phenyl)ethan-1-one) produced the arylated product in trace yield, presumably because the ketimine is more difficult to form in situ. Having optimized the reaction conditions, we next investigated the aryl iodide scope using (E)-2-(prop-1en-1-yl)benzaldehyde (1) as the representative alkene substrate (Fig. 2). Both para-and meta-substituted electron-deficient and electron-rich aryl iodides were competent coupling partners, affording the desired products in moderate to good yields. Notably, free carboxylic acid (2c), unprotected alcohol (2d), and Bocprotected amine (2e) groups were all tolerated. In addition, a variety of heteroaryl iodides were reactive (2j-2m); however, some azaheterocycles were found to give poor yields (2n), presumably due to competitive binding to the metal center. Finally, aryl iodides with free amine substituents gave complex mixtures of products, likely due to off-target condensation with the benzaldehyde substrate. Next we probed the alkene scope with 3-iodoanisole as coupling partner (Fig. 3A). A clear dichotomy of reactivity can be observed when comparing the Z-and E-alkene substrates, with Z-alkenes giving higher yields at the expense of slightly diminished enantioselectivity and E-alkenes giving exceptional ee at the expense of yield. Sterically bulky substituents, such as the cyclopropyl groups in 4f and 4g, led to attenuated reactivity. ortho-Alkenyl benzaldehydes containing various functional groups at the ortho-, meta-, and para-positions relative to the aldehyde moiety delivered yields ranging from modest to excellent with high enantioselectivity. Noticeably lower enantioselectivity was observed in the case of bulky substitution at the ortho-position (4n). In order to demonstrate the simplicity and robustness of the protocol, we performed a representative reaction on gram-scale (4o). The reaction proceeded to full conversion (as in the smaller scale trial), and gratifyingly, the pure product could be obtained without chromatography by using an aqueous workup procedure with potassium metabisulfite (84% isolated yield) (38). The aldehyde moiety in the product can be readily decarbonylated in near quantitative yield (Fig. 3C), allowing it to function as a traceless directing group and enable access to diverse enantioenriched 1,2-diarylethane core structures. To understand the origins of the exquisite regio-and enantioselectivity in this system, we combined insights from experiment and theory, as outlined in Figure 4. First, to probe the role of formate, we performed a labeling experiment using sodium formate-d (Figure 4A). As expected, full deuterium incorporation at the benzylic position was observed in the product (7), confirming that formate serves as a hydride source following decarboxylation. To rule out an alternate pathway involving classical Mizoroki−Heck arylation followed by reduction of the intermediate diarylalkene, we subjected a representative stilbene substrate (8) to standard conditions with exclusion of aryl iodide, and in this case only unreacted starting material was obtained (Figure 4B). A control experiment without amino acid additive resulted in no reaction, demonstrating the critical role of this co-catalyst for reactivity as well as selectivity. DFT calculations showed that the alkene migratory insertion is irreversible and enantiodetermining because the subsequent decarboxylation requires a lower barrier (see SI for details). Therefore, we calculated the facial selectivity in the migratory insertion step using 1 as the model substrate (Figure 4C). The p-alkene complex (IM1) and the migratory insertion transition state (TS1) leading to the (S)-enantiomer of the product are both significantly more stable than corresponding structures in the pathway that forms the (R)-enantiomer (IM2 and TS2, respectively). The origin of this energy difference is not a direct steric repulsion between the t-Bu moiety of the directing group and either of the two newly formed stereocenters (34). Instead, it is attributed to the conformation of the fused benzene ring. In both IM1 and TS1, the benzene ring fused to the 6-membered palladacycle is puckered below the plane of the square planar Pd, while the t-Bu moiety of the directing group is puckered above the plane. This geometry allows the imine nitrogen to remain in the preferred planar geometry and the Pd center to be perfectly square planar. By contrast, in IM2 and TS2, the benzene ring and the t-Bu moiety are both puckered above the plane. This induces significant distortion around the sp 2 -hybridized imine nitrogen and distorts the Pd from the preferred square planar geometry. Although the chiral center on the TDG is remote from the alkene, the distortion effect leads to a 5.6 kcal/mol energy difference between TS1 and TS2. Thus, the formation of the (S)-product is predicted to be strongly favored, which is consistent with the absolute configuration of the X-ray structure as demonstrated in Figure 3B. Based upon previous results, literature precedent, and our computational investigation, we propose that the reaction proceeds via dual catalytic cycles as outlined in Figure 4D. Following the oxidative addition of the aryl iodide and coordination with the condensed imine, intermediate IM1 undergoes an irreversible migratory insertion step with enantioinduction resulting from the conformation of the fused benzene ring. Next, the stabilized alkylpalladium(II) intermediate is intercepted with formate, which decarboxylates to generate a Pd-H species. Following C-H reductive elimination, Pd(0) is regenerated to close the catalytic cycle, and dissociation affords the desired product. The data presented herein suggests that a chiral TDGbased strategy can by successfully applied in the asymmetric hydroarylation of alkenes remote to aldehyde functional groups. Additional studies will be required to further elucidate the mechanism of the transformation and establish whether this general strategy can be extended to alkene difunctionalization.

chemsum

{"title": "A Transient Directing Group Strategy Enables Enantioselective Reductive Heck Hydroarylation of Alkenes", "journal": "ChemRxiv"}

ultralong_organic_room-temperature_phosphorescence_of_electron-donating_and_commercially_available_h

## Abstract: Ultralong organic room-temperature phosphorescence (RTP) materials have attracted great attention recently due to its diverse application potentials. Several ultralong organic RTP materials mimicking the host-guest architecture of inorganic systems have been exploited successfully. However, complicated synthesis and high expenditure are still inevitable in these cases. Herein, we develop a series of novel host-guest organic phosphore systems, in which all chromophores are electron-rich, commercially available and halogen atom free. The maximum phosphorescence efficiency and the longest lifetime reach at 23.6% and 362 ms, respectively. Most importantly, experimental results and theoretical calculation indicate that the host molecules not only play a vital role in providing a rigid environment to suppress non-radiative decay of the guest, but also show a synergistic effect to the guest through Förster energy transfer (FERT). The commercial availability, facile preparation and unique properties also make these new host-guest materials an excellent candidate for anti-counterfeiting devices. Room-temperature phosphorescence (RTP) materials have arisen considerable attention because of their distinctive photophysical properties such as long emission lifetimes and the excited state energy, which can be widely applied in data encryption, anti-counterfeiting, background-free bioimaging, chemical sensors, and so on. Although nature is abundant in inorganic phosphorescence materials such as luminous pearls and glow-in the dark stones, most of them suffer from some intrinsic problems, including high cost, potential toxicity, and low biocompatibility etc. For example, a highly efficient RTP system of SrAl2O4 doped with europium and dysprosium was developed in the mid-1990s, and this inorganic system formed the basis of most commercial glow-in-the-dark paints because of its long emission (> 10 h) and high durability. However, this system requires not only rare-earth elements for long-lived emission but also very high fabrication temperatures of more than 1000 °C. Moreover, the manufacture of paints from the insoluble SrAl2O4 requires tedious procedures, including grinding of the compounds into micrometer-scale powders for dispersion into solvents or matrices, and light scattering by the powders prevents the formation of transparent paints. It is urgent to explore another counterpart with extraordinary features. Inspired by generally accepted principle in inorganic RTP systems that impurity is responsible and involved three steps: (1) excitation of the guest species (impurity), (2) trapping of the excited electrons by defects in host matrices, and (3) slow charge recombination of the trapped electrons by thermal energy, the researchers have successfully developed some ultralong organic phosphorescence materials imitating the host-guest architecture of minerals. However, complicated synthesis and high expenditure are still inevitable in these systems. Moreover, electron-withdrawing species is necessary. For example, recently, Adachi et al. obtained a purely organic afterglow lasting for more than 1 h via blending one electron-donating guest in an electronwithdrawing host, namely D-A structure, but the overall emission quantum yield was only 7% in air. Tang et al. developed a series of purely organic host−guest materials which are both electrondeficient, and halogen atoms, carbonyl groups and cyano units were incorporated. Therefore, a new facile and robust hostguest strategy utilizing only electron-rich materials is proposed. Our group accidently observed a notable RTP phenomenon in a normal experiment by mixing N,N,N',N'-tetraphenylbenzidine (TPB), triphenylamine (TPA) and triphenylphosphine (TPP) (Figure 1a). As we all know, TPA and TPP are commonly used starting materials in organic experiments, and TPB is a widely employed hole-transport material in organic light-emitting diodes (OLEDs). In other words, all these three molecules are commercially available with extremely low price. Moreover, both host (TPA or TPP) and guest (TPB) molecules are electron-rich and contain no halogen atoms. When TPB was doped into host molecules, an enhanced phosphorescence quantum yield (P) and extended emission lifetime were witnessed. The maximum P and the longest lifetime could reach at 23.6% and 362 ms, respectively. Furthermore, the experimental results and theoretical calculation revealed that the host molecules not only play a vital role in providing a rigid environment and suppressing non-radiative decay of the guest, but also show a synergistic effect to the guest in the photophysical process, where Förster resonance energy transfer (FRET) is a key issue to facilitate the phosphorescence of TPB. At the same time, these new hostguest luminescence systems can be facilely fabricated and potentially be an excellent candidate for anti-counterfeiting devices. Thus, the integrated merits including low cost using the most common compounds, absence of halogen atoms, facile preparation, excellent performances and clear mechanism insights of this new host-guest RTP system with all electron-rich species are expected to attract wide attentions and inspire further innovation among the researchers. To avoid the possible interference of impurities of the host and guest materials, strict purification procedures including column chromatography and recrystallization for three times were used, and their purity was unambiguously confirmed by HPLC measurement (Figure S1). Afterward, the luminescent properties of the individual component and doped materials were investigated at room temperature (Figure 1b), respectively. TPB in the crystalline state exhibits a violet blue fluorescence, emitting at 422 nm with a lifetime of 1.3 ns (Table 1 and Figure S2). At the same time, a very weak phosphorescence at 530 nm and a lifetime of only 9.7 s under ambient condition were recorded. The large twisted angle of the biphenyl core of TPB was verified to be the origin of RTP (Figure S3a). Meanwhile, strong intermolecular interactions lock the molecular conformation and restrict the molecular vibration, resulting in the reduction of nonradiative transition of triplet excitons to boost RTP under ambient atmosphere (Figure S3b). Unique luminescence property was also observed on TPP crystalline powders. As shown in Figures 1b and S4, an obvious fluorescence emission at 285 nm was recorded in the photoluminescence (PL) spectrum of TPP in tetrahydrofuran (THF) solution, while no corresponding emission was found in its solid state. Furthermore, the emission peaked at 439 nm in its crystalline state was supposed to be phosphorescence owing to the long lifetime of 1.4 s (Figure S5). Besides the lifetime, this emission peak was further proven by measuring its PL spectrum at 77 K (Figure S6), from which a similar emission profile was obtained. Different from TPP, TPA only emits fluorescence peaked at 393 nm, but no phosphorescence was recorded at room temperature. [e] Data was collected at 77 K. The crystals of TPP and TPA showed weak luminescence with overall PL quantum yields (PL) of 4.6% and 9.7%, respectively. When doping 1.0 mol% TPB into TPP or TPA, the formed TPB/TPP and TPB/TPA blends exhibited enhanced quantum yields and ultralong lifetime (Table 1). As can be seen from Figure 1b, these host-guest systems gave purple fluorescence with the wavelength ranging from 404 to 417 nm. Meanwhile, the maximum phosphorescence of TPB/TPP and TPB/TPA redshifted from 505 to 522 nm, with shoulder peaks at 458 and 540 nm, and 552 nm, respectively. In addition, TPB/TPP and TPB/TPA co-crystals were obtained through careful cultivation, whose X-ray diffraction (XRD) patterns exhibited strong and sharp peaks (Figure S7). Furthermore, the formation of co-crystalline structures was confirmed by differential scanning calorimetry (DSC) measurement. The only one melting point of TPB/TPP and TPB/TPA crystalline powders which was different from TPP and TPA themselves (Figure S8), indicating co-crystalline formation of the host and guest molecules. The co-crystals all exhibited high P with values of 23.6% and 19.4%, respectively. According to the phosphorescence decay curves, the RTP lifetime of cocrystals of TPB/TPP and TPB/TPA could reach at 198 to 362 ms (Figure S9), resulting in a long afterglow that lasts for several seconds under ambient conditions. The relatively stronger phosphorescence p of TPB/TPP than that of TPB/TPA (Table 1) might be due to the heavy atomic effect of phosphine in the former. In general, the host materials only provide a rigid environment to prevent the triplet excitations from quenching by interaction with oxygen. However, the lifetime of TPB/TPP and TPB/TPA cocrystals at room temperature just enhanced slightly under nitrogen over in air (Figure S10). Therefore, the great luminous ability of host-guest materials under ambient conditions indicates that the rigid host molecules also restrict the motion of the guest ones and decrease the non-radiation transitions, thus improving the light conversion efficiency. To have a deep insight into aforementioned phenomena, density functional theory calculations were performed to obtain the singlet and triplet energy levels of the host and guest molecules (Figure 2). The results manifested that the spin−orbit coupling constant (SOC, ξ(S1,T1)) of the TPB was only 0.05 cm −1 due to the lack of heavy atoms, which is consistent with the short lifetime and low quantum yield of the single component of TPB. However, the persistent RTP can be generated by doping TPB into either TPP or TPA host materials. It means that the host and guest molecules must act synergistically in the photo-excited electronic transitions. Thus, the energy levels of TPB, TPP and TPA were calculated. As displayed in Figure 2b, large band gap between the lowest singlet state (S1) and lowest triplet state (T1) of guest molecules does not facilitate the intersystem crossing (ISC), while the S1 of the guest molecule of TPB and the T1 of the host molecules of TPP and TPA were very close. By combing the big proportion of spectral overlap between the host emission and the guest absorption (Figure S6), we conclude that the FRET process exists in these host-guest doping systems, in which the TPP or TPA acted as an energy donor and TPB as an acceptor. It is well-known that FRET is a distance-sensitive energy-transfer process and is dependent on distance of the donor and acceptor (RDA 6 ). The efficiency of the FRET is usually measured by the energy transfer rate from a host to a guest (kET), the efficiency (ET) and the Förster radius (R0), at which the FRET is 50% efficient. The kET can be calculated using the following equation: where D is the decay lifetime of a host in the absence of a guest, R0 is the Förster distance, and RDA is the host-to-guest distance. The efficiency of energy transferET is given by the following formula: The calculated values of R0, kET and ET of our host-guest systems are listed in Table 2. It is evident that the ET of TPB/TPA system (91.50%) is much higher than that of TPB/TPP (74.33%). The great distinction of ET values reveals that TPB/TPA cocrystals experience an absolute FRET process so that no emission originated from TPA appeared in the phosphorescence spectrum. In contrast, TPB/TPP co-crystals underwent an incomplete FRET process following an efficient intersystem crossing from the singlet state of TPP to triplet state, thus causing the extra shoulder peak at 458 nm originated from the phosphorescence of TPP. Consequently, persistent RTP systems through host-guest strategy involving efficient FRET could well occur in different systems as plotted in Figure 2c and 2d. [a] The detailed computational processes are shown in the Supporting Information. After having a deep understanding of the mechanism, we demonstrated the potential application of these host-guest RTP systems. Since the guest can be mixed into the hosts at molecular level through facile evaporation process, these co-crystals are of great potential to serve as anti-counterfeiting materials. Considering the similar purple fluorescence color of TPB, TPB/TPP and TPB/TPA crystals, and doped materials all exhibited bright green phosphorescence with various lifetimes of afterglow (Figure 3a), herein, an anti-counterfeiting pattern "888" was fabricated using different luminophores (Figure 3b). Under a 365 nm UV-lamp irradiation, the pattern displayed a digital number "888", which became an obvious word "UOP" (which means Ultralong Organic Phosphorescence) visible to the naked eyes due to the synergistic emission of TPB/TPP and TPB/TPA co-crystals after ceasing UV light irradiation instantly. At the end, the digital number of "1 1 1" emitting from TPB/TPP appeared after 2 seconds (Figure 3c). Hence, triple data encryption processes have been realized by this simple preparation method, which would be much promising in practical application. In summary, we present a new strategy to realize purely organic RTP with both high efficiency and ultralong lifetime via mixing commercially available TPB guest into either TPP or TPA host. The experimental results and theoretical simulation indicate that: (1) the rigid construction of host molecules not only play a vital role in avoiding the quenching of the triplet excited states by oxygen, but also restrain the non-radiation transitions to advance the light conversion efficiency; (2) host and guest molecules work synergistically in the photo-excited electronic transition processes; (3) an efficient FRET process is activated in the co-crystals of the host and guest to facilitate the luminescence originated from TPB. This new strategy enjoys the advantages including low cost, absence of halogen atoms, facile preparation and excellent performances, which shows great potentials in practical applications. Therefore, this work broadens the way for the fabrication of purely organic RTP materials and offers a novel platform for the development of outstanding applications. ## Table of Contents New type of ultralong organic room-temperature phosphorescence (RTP) systems constructed from commercially available host and guest molecules are presented, which show the maximum phosphorescence efficiency and the longest lifetime of 23.6% and 362 ms, respectively. This new host-guest strategy involving efficient Förster resonance energy transfer (FRET) proves that host molecules not only play passive roles such as rigidifying the guest to suppress the non-radiative decay, but also undergo the excitation and emission processes with guest molecules.

chemsum

{"title": "Ultralong Organic Room-Temperature Phosphorescence of Electron-donating and Commercially Available Host and Guest Molecules through Efficient F\u00f6rster Resonance Energy Transfer", "journal": "ChemRxiv"}

first_total_synthesis_of_concavine

## Abstract: The synthesis of the unusual alkaloid concavine, isolated from Clitocybe concava (Basidiomycetae), has been accomplished. The synthetic route features regio-and stereoselective manipulation of polycyclic imide intermediates via enolate substitution and Grignard addition, along with a key bridge-forming step involving a new method for sulfenylative radical cyclisation. The NMR data for synthetic concavine demonstrate that the original data reported for the natural product refer to the derived acetic acid salt, probably formed as an artefact of isolation or purification. ## Introduction In 2005 the group of Nasini reported the isolation of an alkaloid possessing a novel ring system from cultures of Clitocybe concava (Basidiomycetae). 1 This new compound, called concavine, was assigned the structure 1 (Fig. 1) based on extensive NMR experiments. The compound is an example of an abnormal diterpene alkaloid that incorporates an additional two carbons in the form of an internalised, serine-derived, ethanolamine fragment. This type of biosynthetic genesis has been explored in more detail for the related atisine-type alkaloids. 2 Concavine features a tetracyclic core structure that incorporates a bicyclo [3.2.1]octane motif fused to a pyrrolo-oxazepane, and this unique polycyclic array, including the presence of fve stereogenic centres, make concavine an interesting synthetic prospect. 3 Our retrosynthetic analysis of concavine is summarized in Scheme 1. Concavine was envisaged to evolve from methylenation of the corresponding C-2 ketone. A late intermediate 2 (with undefned functionality Y) would serve as a precursor to this ketone, which would also enable diastereoselective installation of the C-1 prenyl group by enolate alkylation. The oxazepane ring in 2 would be formed from an intermediate 3, or equivalent, by closure of an ether linkage. Access to intermediate lactam 3 would require regio-and stereoselective manipulation of a precursor imide 4, so as to enable selective allylation of the more highly substituted imide C]O function. Stereocontrol in the introduction of an appropriate allyl substituent, most likely allyl (R ¼ H) or methallyl (R ¼ Me) would be possible via appropriate ordering of allylation and reduction events, the second of which would involve a reactive N-acyliminium intermediate. Formation of the bicyclo[3.2.1]octane subunit by a C-C bond forming event was envisaged from a precursor 5, equipped with an appropriate functionality Z. At the outset the identity of the functions Y and Z and the precise ordering of various events were not clear, although a number of options presented themselves. In any case, enolate substitution of an appropriate imide (6), derived from commercial anhydride 7, would provide imide 5, which would need to incorporate functionality (Z) to enable subsequent two-carbon bridge formation. It is worth noting that any route planned around an initial imide desymmetrization might offer opportunities for an asymmetric variant, for example by our well-proven chiral lithium amide base methodology. 4 However, at the outset we opted for an initial approach to the alkaloid in racemic form. ## Results and discussion This plan required us to solve a number of key selectivity issues, including the identities of various groups and functions that would ultimately allow access to concavine. The initial phase of our synthesis is shown in Scheme 2. Preparation of multigram quantities of a suitable starting imide 9 was trivial, starting with anhydride 7 and installing a suitably protected N-hydroxyethyl fragment. Our frst explorations of imide enolate substitution using 9 centred on the use of activated electrophiles, particularly various bromoacetates. However, the derived imide esters proved problematic substrates in subsequent allylation chemistry and we instead made use of a suitably protected iodoethanol fragment, 5 which provided imide 10 in respectable yield. Deprotection and iodide formation under standard conditions proceeded smoothly to give 11. The closure of the two carbon bridge onto the cyclohexene ring was required to proceed with concomitant functionalisation (group Y in our initial analysis). Initial attempts focused on an oxidative radical cyclisation that would provide an alcohol intermediate 4 (Y ¼ OH). We applied protocols described by the groups of Nakamura and Prandi, 6,7 which utilise combinations of Bu 3 SnH and oxygen, to iodide 11 but were not able to achieve productive cyclisation. Instead, we turned to a range of organometallic methods, probably involving cyclisation of radical intermediates, involving the use of either EtMgBr or SmI 2 , with electrophilic trapping. 8,9 Although in some cases we saw evidence of C-C bond formation we could not trap the cyclised organometallic with electrophiles such as PhSSPh. Instead, we developed a variant of the procedure described by Renaud for radical cyclization with trapping by benzenesulfonyl azide to give azides. 10 By employing a mixture of hexabutylditin and PhSSPh, and irradiating with a sun lamp, we were able to cleanly convert iodide 11 into the cyclised sulfde 12, isolated as a single diastereoisomer, in acceptable yield. Attempts to install oxygen in place of the sulfur substituent, by substituting TEMPO for PhSSPh in this type of process were not fruitful. 11 At this stage we required to establish the regio-and stereocontrolled additions to the imide function in which reaction of the more substituted position (C-5a, concavine numbering) would occur in preference to addition at C-10. This type of control would be expected based on models of imide reduction developed by Speckamp, which included consideration of Bürgi-Dunitz type trajectories for the nucleophilic attack. 12 Gratifyingly, addition of Grignard reagents, including allylmagnesium bromide and MeMgBr provided this mode of regioselectivity and it became evident that the facial bias of the tricyclic system present in 12 provided for very high levels of diastereocontrol, with nucleophiles attacking from the more exposed convex face with respect to the original bicyclic imide motif, Fig. 2. We believe that the conformation of the cyclohexane ring is boat-like (see crystal structure of related sulfone in Fig. 3) and that the overall topology and substitution pattern effectively block competing pathways as illustrated in Fig. 2. This mode of reaction was also observed in NaBH 4 reductionsee later. In the case of allylation, this overall pattern of selectivity allowed for efficient conversion of 12 into 13, with only minor amounts (ca. 7%) of an alternative regioisomeric adduct being observed. Subsequent controlled reduction was initially attempted using BF 3 -OEt 2 in combination with silanes such as Et 3 SiH or Ph 3 SiH. 13 However, these reagents resulted in facile elimination of water and formation of a dienamine product. After some experimentation we established that the use of NaBH 3 CN under Brønsted acidic conditions provided excellent results, 14 the stereocontrol following the pattern observed previously, so as to convert 13 into 14 with inversion of the orientation of the allyl substituent. With the carbocyclic skeleton of concavine complete the remaining tasks included: (i) formation of the oxazepane ring, (ii) conversion of the sulfde function into a ketone and, (iii) installation of the remaining prenyl substituent. The successful sequence that emerged, leading to the complete structure of concavine, is shown in Scheme 3. Preliminary experiments probed the viability of either an oxidative desulphonylation of a sulfone corresponding to sulfde 14 (or a later intermediate lacking the amide function), or a Pummerer-type process applied to an appropriate sulfoxide. Although an intermediate sulfone was prepared from 14, we were unable to effect oxidative desulfonylation by applying a number of published protocols. 15 We chose instead to progress sulfoxide intermediate 15, which was easily prepared as an inconsequential 1 : 1 mixture of diastereoisomers at the sulfur centre. Our initial projection for the next steps involved allyl chain modifcation and formation of the oxazepane prior to a Pummerer ketone synthesis. This sequence was disrupted by the observation that removal of the benzyl ether protection was problematic in the presence of a sulfoxide. To our surprise, we also observed that this system provided vinyl thioethers, and not ketones, on exposure to Pummerer conditions. 16 These observations led us to an unplanned order of steps, starting with the application of classical Wacker conditions to 15 to give the desired ketone 16 in good yield. 17 Treatment of this compound with the aforementioned Pummerer conditions gave vinyl thioether 17 (and not a problematic diketone), and subsequent reaction with MeMgBr then gave tertiary alcohol 18. Progression of this intermediate proved highly problematic, since removal of the benzyl ether protection failed under conditions used on simpler models (e.g. TMSI or H 2 , Pd(OH) 2 ), and the vinyl thioether proved resistant to conventional hydrolysis conditions (TiCl 4 , CH 2 Cl 2aq ; HClO 4 ; Fe(NO 3 ) 3 -bentonite K-10). After considerable experimentation, and to our delight, we found that exposure of sulfde 18 to BBr 3 effected both the desired deprotection and the vinyl sulfde hydrolysis in a convenient onepot process, providing ketone 19 in 70% yield. With this key hurdle overcome, the key ring closure was found to be possible using very strongly acidic conditions, to provide tetracyclic oxazepane 20. In all of the stereoselective steps explored on this framework we had observed high levels of selectivity corresponding to reagent attack on the exposed convex face with respect to the initial bicyclic imide framework. This mode of reactivity was observed again in the regio-and stereoselective prenylation of ketolactam 20 using LHMDS and prenyl bromide. Thus, overall topological control from the azabicyclo[4.3.0] motif dominates and provides the desired product 21, despite the fact that this corresponds to endo-alkylation with respect to the bicyclo[3.2.1] ketone fragment. 22 Evidence for the stereochemical outcome included an nOe correlation between the remaining methine at C-1 and one of the hydrogens in the one-carbon bridge at C-4. The remaining steps involving ketone methylenation to provide 22 and lactam reduction were high yielding and provided a fnal product anticipated to correspond with the natural product concavine (1). However, comparison of the 1 H and 13 C NMR data of our synthetic sample of 1 with the published data (in two solvents) revealed some signifcant discrepancies. In the 1 H NMR spectra the differences were clustered around the pyrrolidine nitrogen, for example the NCH methine at C-5a appeared at 0.42 (acetone-d 6 ) to 0.56 (CDCl 3 )ppm upfeld from the reported shift values. Shift deviations were also observed for the C-5 CH 2 of up to 0.55 ppm in the 1 H NMR spectra and up to 2.60 ppm in the 13 C spectra. The identity of our synthetic series, including the relative stereochemistry, appeared secure, based on extensive spectroscopic characterisation. In addition, we had supplies of sulfones from the abortive oxidative desulfonylation approach mentioned above, and one such intermediate 23, prepared in a six-step sequence from sulfde 14 provided crystals suitable for X-ray crystallographic analysis, Fig. 3. 23 The structure is completely in accord with expectations, including the completed oxazepane ring and the correct relative confguration at C-5a. Interestingly, the sulfone substituent is equatorially orientated on a cyclohexane that adopts a boat conformation. Although the extensive NMR data reported by Nasini and coworkers appeared consistent with our own, the chemical shift differences shown by our synthetic material, and centred around the pyrrolidine ring, were too substantial to ignore. Cognisant of the possible pK a sensitivity of a basic natural product, we were aware of reports that HCl salt formation had resolved downfeld shifts in other alkaloids, for example in the Nishida synthesis of Nakadomarin A, the spectra of the synthetic material were reconciled with those described for the natural product only after formation of an alkaloid-2HCl salt. 24 Unfortunately, in our case, formation of HCl salts of 1, resulted in excessive downfeld shifting of a number of signals with the result that the overall correlation of both 1 H and 13 C NMR spectra was worse than for our neutral concavine sample. With the apparent problem seemingly centred on the pyrrolidine region, we speculated that the natural product could be the C-5a epimer. We were able to make this compound through minor modifcations to our route, starting with the sulfenyl imide 12, Scheme 4. By simply switching the ordering of allylation and reduction steps applied to 12, relative to Scheme 2, we were able to prepare a mixture of 24 and 25, each with the desired epimeric (compared to 14) confguration at C-5a. Compound 25 was transformed uneventfully through an unoptimised 7 step sequence parallel to that employed for concavine. To our chagrin, this synthetic effort generated an epimeric alkaloid epi-1 with NMR spectral data that deviated even more signifcantly from those originally publishedin excess of a 15 ppm shift for C-15 in the 13 C NMR for example. Fortunately, at this stage, a member of the Italian group responsible for isolating concavine was able to locate an original sample and forward it to us. 25 Preliminary examination of this sample by NMR showed close correspondence with the published data, although the sample showed signs of contamination or decomposition. Remarkably, passing this sample through a short silica column gave a clean alkaloid the data for which were a complete match to our synthetic material! Further examination of the spectra from the initially received sample from Milan provided an explanation for this phenomenon, since there was evidence of acetate (AcO ) signals in the spectra, including a methyl signal at d ¼ 1.98 ppm in the 1 H spectrum and at d ¼ 22.8 in the 13 C spectrum. Titration of our synthetic sample with HOAc then gave 1 H and 13 C spectra in CDCl 3 , which clearly demonstrated that the compound is identical to the reported structure. Thus, the reported data appear to be for a concavine-HOAc salt, presumably generated as an artefact of purifcation and some signals in the spectra were overlooked (perhaps discounted as solvent). Interestingly, the original isolation paper describes purifcation of concavine with rather polar eluants containing 5-10% MeOH. In our hands the use of EtOAc-petroleum ether mixtures was adequate to effect alkaloid purifcation, and it appears that any salts (HCl or HOAc) are decomposed to give the neutral amine when purifcation is carried out this way. In conclusion, we have described the frst synthesis of the unusual alkaloid concavine (1), using a stereocontrolled route that is flexible enough to accommodate the synthesis of an epimer (epi-1). The route features a new sulfenative radical cyclisation protocol, an unexpected vinyl sulfde synthesis and a fortuitous hydrolysis of said vinyl sulfde, concomitant with an ether cleavage, using BBr 3 . Importantly, we have clarifed that the reported data for concavine actually relate to a derived acetic acid salt.

chemsum

{"title": "First total synthesis of concavine", "journal": "Royal Society of Chemistry (RSC)"}

heavy-atom-free_room-temperature_phosphorescent_organic_light-emitting_diodes_enabled_by_excited_sta

## Abstract: Room temperature phosphorescence materials offer great opportunities for applications in optoelectronics, due to their unique photophysical characteristics. However, purely organic emitters that can realize distinct electrophosphorescence are rarely exploited. Herein a new approach for designing heavy-atom-free organic room temperature phosphorescence emitters for organic light-emitting diodes is presented. The subtle tuning of the energy diagrams of singlet and triplet excited states by appropriate choice of host matrix allows tailored emission properties and switching of emission channels between thermally activated delayed fluorescence and room temperature phosphorescence. Moreover, an efficient and heavy-atom-free room temperature phosphorescence organic light-emitting diodes using the developed emitter is realized. ## Introduction Electroluminescence of organic molecules is one of the groundbreaking discoveries of the last century. 1 Owing to it and the significant advancement in organic electronics, we are surrounded by organic light-emitting diode (OLED) products. 2,3 In addition to lightness, flexibility, and low production cost, the major advantage of using organic emitters is that their properties are tailored by modifying their molecular structures. From the viewpoint of emission mechanisms, there are two types of emitters: fluorescence and phosphorescence emitters. Fluorescence emitters are known as the 1 st generation emitters for OLEDs, but they fatally have a problem of a discouraging upper limit (25%) of internal quantum efficiency (IQE), because of the spin-statistics issue. The 2 nd generation emitters, phosphorescence emitters, can show much higher IQE up to 100%, and for this reason, they are commonly used in OLEDs displays. 2,3,6,7 However, currently used phosphorescence emitters are organometallic complexes that contain expensive and rare heavy metals. There are two approaches to achieve 100% IQE in OLED devices with metal-free emitters: the utilization of purely organic 1) thermally activated delayed fluorescence (TADF) emitters 14 or 2) room temperature phosphorescence (RTP) emitters. TADF emitters, which are dubbed as 3 rd generation emitters, can harvest 100% of electrically-generated excitons by converting triplet states (T1) into energetically-close singlet excited states (S1) with thermal energies and thereby can emit photons through irradiative relaxation to the ground state (S0). An issue of TADF emitters is a broad Gaussian-type emission spectrum, which might be suitable for lighting applications but not for the biggest market, displays. 15,16 Although purely organic and Heavy-Atom-Free (HAF) RTP emitters would be a promising alternative as OLEDs emitters, this class of emitters has been rarely exploited. For example, Chaudhuri et al. reported that RTP can contribute to the emission from an OLED fabricated with small organic π-conjugated compounds with light atoms (C, H, and N), 13 but external quantum efficiencies (EQEs) of the devices were technically too low (<10 -4 %). From a mechanistic point of view, TADF and RTP processes appear to be opposing ones (from T1 to S1 followed by S1 to S0, and from T1 to S0 transition, respectively). But, theoretically, it should be possible to connect them, as both processes are required to have strong spin-orbit coupling (SOC) to increases efficiency. To endow emissive organic compounds with RTP characteristics, molecular rotations and vibrations should be locked so that non-irradiative deactivation pathways are shutout. 17 On the other hand, to realize TADF characteristics, vibrations between electron-donor (D) and the acceptor (A) moieties are necessary. 17,18 As we presented before, it is possible to slow down the rotation between donor and acceptor moiety in order to observe the RTP process by attaching bulky, steric hindering groups. 19 The limiting value here is the temperature and decay lifetime. Phosphorescence is efficient at low temperature. On the other hand, by increasing temperature, we observe reverse intersystem crossing (RISC) and thereby TADF occurs, but the phosphorescence is not observed. In order to hinder the possibility of the TADF process, the singlet-triplet gap (ΔEST) needs to be theoretically higher than 0.4 eV. Nevertheless, in spite of these conflicting design criteria, it is possible to switch between RTP and TADF processes by choosing appropriate hosts and control the ΔEST. Generally, TADF occurs through the RISC from an excited local triplet state ( 3 LE) to the excited charge transfer singlet state ( 1 CT), with the 1 CT energy highly sensitive toward host polarity (Figure 1a). Therefore, if the ΔEST lies on the edge of the TADF process, it would be feasible to boost the RTP emission in a certain temperature range (Figure 1b). Nevertheless, an overlooked issue regarding purely organic RTP compounds is that the irradiative deactivation pathway from the T1 state is commonly a very slow process (typical lifetime >1 ms). This disproves the possibility of applying such RTP-active organic compounds for OLED application, as the efficiency and stability of those OLEDs are poor, mainly due to exciton-polaron quenching and triplet-triplet annihilation (TTA). To overcome these issues, we decided to utilize a faster phosphorescence pathway from the T2 state which are accessible through thermally activated reverse internal conversion (RIC) process from longer-lived T1 state, where the S1 ( 1 CT1) is higher-lying than T2 and T1 states (Figure 1c). According to literature, the CT character of T2 state and a large SOC between T2 and S0 states should result in the much faster irradiative transition from T2 than that from T1 state to suppress the triplet quenching. 24 Herein, we present a novel TADF/RTP hybrid HAF emitter SiAz (Figure 1c) comprising of naturally-abundant elements only (C, H, N, and Si). In addition to the material design, our theoretical investigation gives proof that the excited states energy alignment (S1, T1, and T2) of the emitter is engineered by host materials by affecting CT energy levels with polarity effect and inverting the energy levels of 3 CT and 3 LE states, leading to switching TADF and RTP outcomes. Most importantly, the developed emitter can be used to a HAF organic electrophosphorescent device with the highest EQE reported so far. ## Results and Discussions Design and Synthesis of SiAz. To realize the HAF-RTP emitter that satisfies the abovementioned conflicting requisites, we designed a donor-acceptordonor (D-A-D) molecule SiAz (Figure 1c) comprising of naturally-abundant elements only (C, H, N, and Si). The design principles for SiAz involve the installation of i) weaker electron-donors (i.e., dihydrophenazasiline) than phenoxazine 25 and phenothiazine 23 to increase the ΔEST by destabilizing the 1 CT state so that RISC becomes less effective and to make T2 state below S1 ( 1 CT) and ii) rigid structure to suppress the non-radiative pathways from T1. The preliminary computational conformer search revealed that SiAz can take 35 meta-stable conformers (Figure S18), which can take equatorial-equatorial (eqeq), equatorial-axial (eq-ax), or axial-axial (ax-ax) orientations. Further optimization of representative structures using density functional theory (DFT) calculations (ωB97X-D/cc-pVDZ) revealed that eq-eq is always the dominant conformer in solutions (around 8.5 and 14.7 kcal mol -1 more stable than eq-ax and axax conformers, Figure 2). Excited state calculations revealed that there are substantial variations in energy alignment of S1, T1, and T2 depending on the materials environments (Figure 2, for the details, see the Table S5-S10 in the Supporting Information). As expected from our previous works, 23,25 the S1 state has hybrid CT character in many solvents (methylcyclohexane: MCH, toluene, and THF) (see the NTO pair of Figure 2). The T1 and T2 states in MCH and toluene are 3 LEA and 3 CT states, while the energy alignment is switched in a more polar solvent (THF) (Figure 2b and c). It should be noted that in less polar solvents such as MCH and toluene not only the energy alignment but also the magnitude relation of SOC matrix elements for T2-S0 (ca. 5.8 cm -1 ), T1-S1 (ca. 2.0-2.2 cm -1 ), and T1-S0 (ca. 0.9 cm -1 ) transitions are ideal for designed RTP scenario (Figure 1c and Figure 2a). Furthermore, what was interesting is the significant stabilization of 1 CT energy in a polar environment (THF) would yield small ΔEST enough for thermally activating RISC process for TADF. The set of predictions encouraged us to start this project to fine-tune the excited states energy alignment by molecular design and host materials. The designed material SiAz was successfully synthesized through a Pd-catalyzed double amination of dibrominated dibenzophenazine 23 with a silicon-bridged diphenylamine (dihydrophenazasiline) in high yield (Equation S1). Thermogravimetric analysis revealed that SiAz is stable toward heating (5 wt% loss temperature = 496 °C), indicating the feasibility of fabrication of OLEDs devices by sublimation. ## Steady-state photophysics of SiAz in solutions and solid matrices. Initially, we investigated the ultraviolet-visible (UV-Vis) absorption and photoluminescence (PL) properties of SiAz. UV-Vis spectra (Figure S1a) in both non-polar and polar solvents displayed transitions below 480 nm, similar to those seen in other D-A-D compounds containing dibenzo[a,j]phenazine acceptor. 19,23,25 Concomitantly, steady-state emission spectra showed typical CT behaviour, where emissions bathochromically shifted as the polarity of solvents increases (Figure S1b). A degassed toluene solution of SiAz showed almost no difference in PL intensity from the aerated solution, indicating that there are little or none emissive pathways after the excursion tothe triplet state in solution (Figure 3a). Different results, however, were obtained in solid matrices, by removing oxygen from the sample of 1% of SiAz in Zeonex ® . The significant rise of the emission was observed, together with the new band appeari ing in the higher wavelength from the RTP process (Figure 3b). This concludes we have both emissions from singlet and triplet states in this system. 9 As for the sample in DPEPO matrix, the increase of the emission is observed after degassing but the shape of the emission spectrum remains, which would suggest the triplet-driven emission is observed in this system and based on the literature this is a delayed fluorescence (DF) process S3. (Figure 3c). 16,19,20,21,23 The highest rise of the emission is observed in the TCTA matrix where there is 2.21 times rise of the emission with the appearance of a new band of longer wavelength, suggesting the RTP process (Figure 3d). 9 The sample of SiAz in CBP also shows a small increase in the emission after oxygen removal, but the increase was limited to 1.1% of the original intensity (Figure 3e). Comparison of the emission spectra of SiAz in DPEPO, TCTA and CBP matrix before and after degassing revealed that only the SiAz in TCTA displays the distinct rise of a new band which is associated with RTP emission (Figure 3d). Otherwise, we are supposed to simply observe the rise of a similar CT emission band with those observed in CBP and DPEPO that would correspond to DF process rather than appearing of a new band (Figure 3d). The photoluminescence quantum yield of the degassed toluene solution was 22% (Figure S2b), further suggesting that non-radiative deactivation pathways from the excited states play a major role in the decay kinetics of SiAz in solution. Intriguingly, aggregation-induced enhanced emission (AIEE) behaviour of SiAz was observed, when the ratio of water/THF blends increased (Figure S3). ## Time-resolved photophysics of SiAz in host matrices. SiAz affords complicated time-resolved photophysics, 28 depending on the solid environments. The excited states energies extracted from the time-resolved photophysical spectroscopic measurements are summarized in Table S3. In order to elucidate the excited state dynamics in the solid-state, a film made by dispersing SiAz in Zeonex ® matrix was spectroscopically analysed at temperatures ranging from 80 to 340 K in a time region between 1 ns and 79 ms (Figure 3f). SiAz decays in two distinct time regions: one in the ns and the other in the ms region. Therefore, the decay profiles were almost devoid of emissive outputs in the µs temporal time region, where delayed emission processes such as TADF, 8 TTA, 29 and triplet state emission are commonly observed. The emission detected in the ns region decays with a single exponential with an emission lifetime of ca. 4 ns at all temperatures. The emission spectra acquired in the ns time region (0-36 ns) at 300 K showed Gaussian shapes with the identical onsets (438 nm, 2.83 eV) and peaks (481 nm, 2.58 eV), which are attributable to the emission from the 1 CT (Figure 3g). At first sight, SiAz appears to decay in the ms time region in a similar fashion at all temperatures (Figure 3f). However, careful inspection of the spectra in the ms delay time revealed that a complicated competition between singlet and triplet excited states exists (Figure 3g). Firstly, in the ms time region, at temperatures below 240 K, SiAz showed a structured spectrum (onset = 516 nm, 2.40 eV, blue line, Figure 3g), which is bathochromically shifted from the 1 CT emission observed in the prompt ns decay region (for the steady-state emission spectra at varied temperatures, see the Figure S6). The emission observed at low temperatures were identified as the phosphorescence irradiated from the locally excited triplet state of the acceptor ( 3 LEA), which was confirmed from the individual phosphorescence spectra of the acceptor. 25 The ΔEST ( 1 CT-3 LE gap) calculated from the onset energies was larger than 370 meV, which corresponds to an energy gap proven to stymie the RISC process in many emissive materials. Additionally, the almost identical energy level of the 3 LE1 of SiAz with that of the acceptor (2.40 eV) 25 suggested that the D-A torsional angles are very close to 90 o (i.e., eq-orientation of donors against the acceptor), where the wavefunctions of the D and A moieties would be decoupled, and thereby manifesting the locally excited energetics of the acceptor. 16 At temperatures above 240 K, the millisecond delay time revealed dual emission comprising of 1 CT and triplet emissions (Figure 3g). The 1 CT emission was identified to be TADF by analysing the dependence of emission intensity on laser pulse fluency at 340 K (Figure S5). Since the TADF is only accessed at temperatures above 240 K, the competition between the radiative decay from the 3 LE1 and the RISC from the 3 LE1 to the 1 CT was affirmed by the large energy barrier between the two excited states (Figure 3g). The RISC rate at 300 K esimated from eq 1 is ca. 10 1 -10 2 s -1 , which ranges in similar decay rates with those of typical phosphorescence. 30 This justifies the presence of dual singlet and triplet emissions at 300 K, allowing for a strong white emission at room temperature at 50 ms (Figure 3g and Figure S6). While the observation of dual TADF/RTP is quite rare, it is not unprecedented in photophysical studies. 31,32 Furthermore, Huang et al. demonstrated that the design principles for TADF and RTP are not mutually exclusive. 31 Both mechanisms require the suppression of the decay from S1 to make kISC, kRISC (required for efficient TADF), and kphos (required only for RTP) competitive. They subsequently highlighted two important design principles: 1) angular molecular geometry within D-A-D scaffolds shifts S1 and T1 in closer proximity and facilitates ISC; 2) the presence of heteroatoms (even ones with a low atomic weight such as nitrogen) 33 enhances triplet formation. Similar considerations on molecular design were also addressed by Chen et. al. when designing phosphorescence compounds with fast ISC rates. 34 In our case, SiAz satisfies Huang's design principles. Based on the ΔEST (0.42 eV) obtained from theoretical calculations in MCH, which can mimic the non-polar polymeric matrix environment (Figure 2 and Tables S5-10), and the behaviour of the emission, the emission observed in the Zeonex ® matrix should be generated from the majority of eq-eq-type conformers. In Zeonex ® matrix, experimental ΔEST is on the edge value (0.43 eV, Table S3) for efficient RISC, which is in good agreement with calculated one ( calc ΔEST = 0.42 eV), thereby reducing the RISC rate from 3 LE1 to 1 CT to show dual emission of TADF and RTP from 3 LE1 (Figure 3h). Given that cal ΔETT (0.30 eV) is slightly smaller than calc ΔEST (0.42 eV) and SOC matrix element between T2 and S0 is larger than that between T1 and S0 states, RTP from the T2 state might be involved in the emission (Figure 3h). This could be also experimentally reasonable, as three emission bands were observed in the steady-state spectra in Zeonex ® (Figure S6). The white emission observed at room temperature in Zeonex ® was not, however, observed in a polar matrix bis [2-(diphenylphosphino)phenyl] ether oxide (DPEPO) (Figure 3i and j). In other words, RTP was not observed in this host. This can be explained by several accumulative factors. Firstly, the polarity of DPEPO, which contains phosphine oxide moieties, is much higher than that of the polymeric cyclic olefin matrix Zeonex ® . Therefore, DPEPO stabilizes CT states over LE states, coaxing the 1 CT to be in closer resonance with the 3 LE, which is evident from a bathochromic shift of the 1 CT emission (Figure 3j). From the theoretical analysis, when compared to the case in MCH (approximate environment of Zeonex ® host), THF (mimic of DPEPO host) allows for a decrease in the energy of 3 CT2 to be lower than the 3 LE1 state, resulting in 3 CT1 and 3 LE2 (Figure 2 and Tables S5-10) close enough to form a resonance structure to feed an efficient RISC process. The triplet onset energies of the 3 LE1 in Zeonex ® (2.40 eV) and DPEPO matrices (2.39 eV) are very close (Table S3), implying that this 3 LE is less affected by host polarity than the 1 CT and 3 CT and also suggesting the same majority of eq-eq conformers irradiate the emission in the matrix. 20 Both the prompt and delayed fluorescence decays displayed CT characteristics, which are bathochromically shifted from the emission in the Zeonex ® matrix, with complicated kinetics (Figure 3i). This is highlighted upon inspection of the emission spectra associated with each delay region. As for SiAz in DPEPO matrix, a bathochromic shift in emission ranging from 460.9 to 482.5 nm (2.69-2.57 eV) at the time between 1 to 54 ns (Figure S7) was observed. This prompt emission from the singlet states does not decay with simple exponential kinetics, most likely due to the conformational heterogeneity of SiAz molecules in the solid rigid host. The delayed emission (>100 ns) shows similar spectral heterogeneity with CT states closest to the triplet state decaying first ( 1 CT = 497.4 nm at 2.8 µs time delay), and molecules in conformations that do not promote such strong CT characteristics decay at later delay times ( 1 CT = 476.7 nm at 17 ms time delay). This diminished energy gap between 3 LE1 and 1 CT, therefore, promotes the RISC process through faster decay (shorter lifetime) from the 1 CT state after accessing the triplet state. Therefore, both prompt and delayed time decays are not mono-exponential. In DPEPO, the observed significant red-shift of prompt emission from those in Zeonex ® and TCTA is an indication of 1 CT emission, which is somewhat higher (2.37 eV) than theoretical value (Figure 2c). Taking into consideration that the SOC matrix elements for the 3 LE and 1 CT states are always higher than those between 3 CT and 1 CT (Figure 2) and that phosphorescence spectra of SiAz in DPEPO matrix is similar to those in TCTA and Zeonex matrices (Figure S9), the RISC should occur from the T1 state with 3 LE character, which then exclusively gives rise to TADF (Figure 3k). The decay kinetics from SiAz appear even more complicated when impregnated into tris(4-carbazoyl-9ylphenyl)amine (TCTA) matrix. TCTA has lower polarity than DPEPO, which corresponds to a smaller bathochromic shift of the 1 CT1 band (Figure 3m). From inspection of the decay kinetics at different temperatures (Figure 3l), we found that SiAz decays in three distinct time regions; one in the ns, µs and the ms time region. While within the ns time regions, SiAz displayed depopulation from the 1 CT (2.88 eV, onset energy), with the first component being as prompt emission. The states in the µs and ms time region, however, are more ambiguous. The emission observed at 2.41 eV (onset energy) in the ms delay times at low temperatures below 240 K was identified as phosphorescence from 3 LE1 state. At higher temperatures, up to 300 K, this phosphorescence was still prevalent (displaying RTP in an OLED host matrix) and yet observed with the addition of a second hypsochromic peak which does not match the onset energy of the peak observed at prompt time scales (Figure 3m). To investigate this unique phenomenon, we conducted an additional photophysical analysis of SiAz in TCTA host, which revealed that long-lived emissive species (up to 50 ms) are not a mixture of TADF and RTP but pure RTP from the T1 and T2 states (Figure 3n). From the analysis of the emission spectra of SiAz in TCTA at varied temperatures, only one new band peaked at 506 nm was appearing with a rise of the temperature (Figure 5). Since the two emission bands are quite close to each other, all excited energy would be transferred to the singlet state ( 1 CT) if this new band were TADF. But such a phenomenon was not observed here (Figure 3n), suggesting RTP processes. Also, from the theoretical analysis (toluene), one will notice that the singlet energy state ( 1 CT) is much higher than the triplet energy state (ΔEST = 0.35 eV), and T1 and T2 energy gap (ΔETT = 0.22 eV) is smaller than ΔEST. In addition, SOC matrix elements between T2-S0 (5.85 cm -1 ) is much larger than that between T1-S1 (2.21 cm -1 ) (Figure 2). Moreover, the T2 state has CT and a slightly LE (hybrid CT) character (Figure 2), which would support the possibility to reduce the phosphorescence lifetime and observe much faster decay from 3 CT2 than from the phosphorescence from 3 LE1 state. The experimental energy difference between this state and T1 states ( exp ΔETT = 0.20 eV, Table S3) is smaller than that obtained from theoretical results ( calc ΔETT = 0.22 eV, Figure 2). This discrepancy would be rationalized by i) a larger dielectric constant of TCTA than toluene and ii) lower T2 energy at the T2 geometry than that at the T1 geometry (vide infra). In TCTA, similar to the RISC process in the TADF mechanism, SOC mixes the different spin states and dissolves the forbidden singlet-triplet transitions. The mixed CT character of T2 state allows for the much faster photo-irradiation process than from T1 (Figure 3o), which is supported by the large difference in the calculated SOC matrix elements in toluene (Figure 2b). Hence, the real excited "triplet" state is not a pure triplet state but has some singlet character, and therefore transition to the S0 is more probable even at room temperature. Therefore, fast radiative decay process from the T2 state is observed at such long times, it would be mainly because the T2 state is fed from long-lived T1 state through thermally activated reverse internal conversion (RIC, Figure 3o). Through the mechanism, so as long as the T1 state allows for RIC process, the short-lived T2 emission will be observed (Figure 3o). ## OLED fabrication and characterization. To investigate whether this material works in OLED devices, the electrochemical behaviour of SiAz was analyzed. The ionization potential (IP) (5.85 eV) and electron affinity (EA) (3.49 eV) were determined with cyclic voltammetry (Figure S14), and they were found at similar potentials to the reported D-A-D emitters containing a dibenzo[a,j]phenazine acceptor. 23,25 The behaviour of SiAz in 3 hosts, CBP [4,4′-bis(N-carbazolyl)-1,1′-biphenyl], DPEPO, and TCTA, was then determined in OLED devices. The 3 OLED devices were fabricated and analysed in the following configurations: Device 1 -ITO/NPB (40 nm)/TSBPA (10 nm)/10% SiAz in CBP (30 nm)/TPBi [2,2',2"-(1,3,5-benzinetriyl)-tris(1-phenyl-1-H-benzimidazole)] (60 nm)/LiF (1 nm)/Al (100 nm); Device 2 -ITO/HAT-CN (10 nm)/NPB (30 nm)/mCP (10 nm)/10% SiAz in DPEPO (20 nm)/TPBi (60 nm)/LiF (1 nm)/Al (100 nm); Device 3 -ITO/NPB (40 nm)/TAPC [4,4′-cyclohexylidenebis(N,N-bis(4methylphenyl)benzenamine)] (10 nm)/10% SiAz in TCTA (30 nm)/TPBi (60 nm)/LiF (1 nm)/Al (100 nm) (Figure 4). The characteristics of the OLED devices revealed a significant increase of OLED efficiency depending on the host used. The device based on CBP was found be the most efficient (5.27% EQE), but it showed substantial roll-off and luminance up to (17,240 cd/m 2 ) in comparison to the devices fabricated with other hosts, subsequently making DPEPO and TCTA better hosts. The DPEPO host also shifted the EL of SiAz to the lower energies and gave a boost in the form of TADF enhancement up to 5.32% EQE and 17,105 cd/m 2 luminance (Figure 4b). As for the TCTA-based device, the current response and the efficiency behaved similarly to other devices, resulting in 4.06% EQE and 22,561 cd/m 2 luminance (Figure 4). However, unique to all other host materials was the observation of a change in emission with increasing voltage. Devices based on the TCTA host showed two emissions, at lower potential <7 V and at higher potential >9 V and a mixture in between these voltages (Figure 4d). This voltage-dependent emission is reversible, and therefore, we concluded that this behaviour is not caused by the degradation of the device. The emission in a lower voltage gave 4.06% EQE and up to 2,000 cd/m 2 luminance, whereas the brighter one in a higher voltage exhibited much lower efficiency (ca. 2% EQE) with much higher luminance (above 20,000 cd/m 2 ). This clearly shows that there are 2 different processes involved in the same device. The phosphorescence from the mixed 3 LE1 and 3 CT2 states matches the EL from the device at a low voltage, while the EL observed at a higher potential matches the majority of emission from 3 CT2 states as the population from 3 CT2 increased (Figure 4d). When all devices are compared, there is a correlation between ΔEST and device efficiency. The devices with the smaller gap (ca. 0.3 eV) showed efficiency around 5% (CBP, DPEPO, and TCTAbased devices), where the gap around 0.4 eV produced devices with only 2% efficiency (mCP). ## Conclusion In conclusion, we have successfully developed a unique purely organic emitter SiAz that could revolutionize the approach in designing novel OLED materials. The introduction of less-electron-donating, rigid, and sterically hindered donors into the acceptor suppressed the rotation and vibration around the D-A units, which turned-on RTP emission. Depending on the hosts in which emitter was impregnated, TADF, RTP, or both emission channels were boosted. This characteristic offers unique opportunities for tailoring emission properties of OLED devices by choosing hosts and applied voltages. Moreover, white emission within a single organic emitter was achieved by joining RTP and TADF emission in Zeonex ® matrix. Most importantly, not only were we able to prove the significant involvement of RTP in OLEDs emission from photophysical analyses but also we realized the most efficient heavy atom-free RTP-based OLED device (4.06%) to date.

chemsum

{"title": "Heavy-Atom-Free Room-Temperature Phosphorescent Organic Light-Emitting Diodes Enabled by Excited States Engineering", "journal": "ChemRxiv"}

synthesis,_structural_and_morphological_characterizations_of_nano-ru-based_perovskites/rgo_composite

## Abstract: Highly-dispersed Ru-based perovskites supported on reduced graphene oxide (A-RG) nanocomposites are prepared using different A-metal salts (sr(No 3 ) 2 , Ba(No 3 ) 2 and Ca(No 3 ) 2 ). the procedure is based on a redox reaction between the metal precursors and graphene oxide (GO) using two different routes of reaction initiation: through thermal heating or by microwave-assisted heating. the resulting nanocomposites do not require further calcination, making this method less energy-demanding. In addition, no additional chemical reagents are required for either the Go reduction or the metal precursor oxidation, leading to an overall simple and direct synthesis method. the structure and morphology of the as-prepared A-RG (non-calcined) nanocomposites are characterized using various structural analyses including XRD, Xps, seM/eDX and HR-teM. Changing metal A in the perovskite as well as the "activation method" resulted in significant structural and morphological changes of the formed composites. srRuo 3 and BaRuo 3 in combination with Ruo 2 are obtained using a conventional combustion method, while srRuo 3 (~1 nm size) in combination with Ru nanoparticles are successfully prepared using microwave irradiation. For the first time, a microwave-assisted synthesis method (without calcination) was used to form crystalline nano-CaRuo 3 .Owing to their high catalytic activity, high thermal stability and high specific capacity, Ru-based materials proved useful in electrocatalysis 1,2 and energy storage applications (e.g., in Li-ion batteries 3 and supercapacitors [4][5][6][7][8] ). Various methods are used for the preparation of Ru and/or RuO 2 -based materials such as hydrothermal 9 , sol-gel 10 , chemical reduction or oxidation 11,12 and precipitation methods 13 . However, this class of materials is rather sensitive to the preparation parameters, and consequently the resulting structure, morphology and catalytic properties varies depending on the adopted preparation approach.One drawback of using Ru-based materials in different applications is related to their low porosity and high preparation cost. To reduce the costs for commercial applications, usually highly-dispersed Ru-based particles are deposited on less expensive supports such as high surface area carbon materials 14 . This helps decreasing the overall costs of the Ru-based electrode materials and increasing the specific surface area and porosity. Several reports discussed the preparation of Ru or RuO 2 over carbon materials. For instance, Lou et al. synthesized Ru nanoparticles over activated carbon, derived from Moringa Oleifera fruit shells for supercapacitor applications 15 . Their method is based on the thermal reduction of Ru 3+ ions at high temperature using ZnCl 2 as an activating agent for carbon formation. Also for supercapacitors, He et al. 14 prepared hydrous RuO x over activated carbon black by a chemical impregnation technique. They found that the specific capacitance is greatly affected by the mass loading of RuO x as well as the specific surface area. Moreover, Zheng et al. 16 reported that the performance of RuO 2 as supercapacitor is highly altered by the calcination temperature. Carbon nanotubes were also used as a support for RuO 2 nanoparticles prepared by the reaction of Ru(VI) and Ru(VII) 17 . Recently, graphene has been widely used as a support for RuO 2 4-8,14 due to its high surface area and outstanding thermal and electronic characteristics 14,18 . experimental preparation of graphene oxide. Graphene oxide is prepared by a modification of Hummer's method, following the same procedure reported by Kovtyukhova et al. 44 . Briefly, 5 g of high purity graphite is cured with H 2 SO 4 , P 2 O 5 and K 2 S 2 O 8 to prepare the pre-oxidized graphite. This is followed by stirring 5 g of dried pre-oxidized graphite with 115 mL of concentrated H 2 SO 4 in an ice bath for 10 minutes. After that, 15 g of KMnO 4 is gradually added and stirred for two hours. The mixture acquired a bright yellow color after dilution with water and treatment with H 2 O 2 . The bright yellow suspension is filtered and washed with 1:10 (v/v) HCl-solution, and finally dried overnight in an oven at 80 °C. preparation of ARuo 3 /RGo nanocomposites by combustion method. Ru-based materials supported on the reduced graphene oxide (A-RG) are prepared by mixing 0.2 g of graphene oxide (GO) with 0.33 mmol of RuCl 3 and 0.33 mmol of metal salt (Sr(NO 3 ) 2 , Ba(NO 3 ) 2 or CaCO 3 treated with concentrated HNO 3 in 20 mL distilled water to form Sr-RG-C, Ba-RG-C or Ca-RG-C, respectively. The resulting suspensions are ultra-sonicated for two hours until a homogeneous mixture of metal precursor is attained. An ammonia solution is used to adjust the pH of the suspensions to 8.0 ± 0.05. The mixture is then heated on a conventional hotplate at 200 °C in air; the mixture dehydrated and transformed into a viscous mass that is followed by its self-ignition. The resulting composites are left on the hotplate for a total of two hours, which is the total reaction time. After the ignition process (firing takes place), black powders are obtained, which is an evidence of the successful reduction of RGO. For comparison, Ru/RGO was prepared by the same method but in absence of Sr, Ba or Ca salts. preparation of ARuo 3 /RGo nanocomposites microwave method. The same solutions are prepared (as those created with combustion methods) by mixing 0.2 g of GO with 0.33 mmol of RuCl 3 and 0.33 mmol of metal salt (Sr(NO 3 ) 2 , Ba (NO 3 ) 2 or CaCO 3 treated with conc. HNO 3 in 20 mL distilled water to form Sr-RG-M, Ba-RG-M or Ca-RG-M, respectively. The mixtures are then placed in a conventional microwave (720 Watts) using 30 s-cycles (the microwave irradiation is switched on and off for 20 s and 10 s, respectively) until the ignition process takes place; the total reaction time is 30 minutes. During the irradiation, the suspension becomes viscous with time and dehydrated. The resulting powder is ignited, and a strong firing takes place, which generates a black powder that indicates the successful reduction to RGO. structural, spectral and surface analyses. All prepared materials are characterized using X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), field-emission scanning electron microscopy (FESEM) with energy dispersive analysis by X-ray (EDX) and high-resolution transmission electron microscopy (HR-TEM). XRDs are recorded with Panlytical X'Pert using Cu-K α radiation (λ = 1.540 Ǻ). The surface morphology is analyzed by HR-TEM (Tecnai G20, FEI, Netherland, 200 kV, LaB6 Gun) and FESEM with EDX (JEOL JSM-6360LA and Philips XL30). TEM measurements were performed on Tecnai G20, FEI, instrument, Netherland, 200 kV, LaB6 Gun. The particles size was calculated using Image J software. At least 500 particles were evaluated and collected from several shots for the same sample. The percentage of the particles was plotted against their size to generate particle size distribution curve. ## Results and Discussion The reduced graphene oxide (RGO) supported Ru-based perovskite composites are successfully prepared in the absence of any reducing or stabilizing agents under controlled pH synthesis conditions using different metal salts (Sr(NO 3 ) 2 , Ba(NO 3 ) 2 or CaCO 3 ) and two initiation methods. Elaborate characterization of the resulting materials including chemical composition and morphological characterization is accomplished using a combination of analysis techniques such as XRD, XPS, SEM-EDX and TEM. sr-RG nanocomposites. Figure (1a) shows the XRD of Sr-RG-C prepared by the combustion method. The diffraction pattern reveals the presence of crystalline RuO 2 clusters as indicated by the appearance of typical diffraction peaks at 2θ = 28°, 35.2°, 39.96°, 45.09°, 59.4° and 59.6°, corresponding to crystal faces of {110}, {101}, {200}, {210}, {002} and {301} orientations, respectively. This result is in good agreement with the XRD reference card of tetragonal RuO 2 (reference card number: 04-015-7002) 45 . Some remaining residuals of soluble SrCl 2 and insoluble SrSO 4 are also observed, which could not be removed under the present preparation conditions. The source of Cl − and SO 4 2− ions are Ru precursors and GO (SO 4 2− remains as impurity in GO during its preparation), respectively. It should be noted that the diffraction pattern of the as-prepared composite does not show any peaks related to the "originally used" graphene oxide (GO) (see Fig. (S1)). This indicates that the preparative approach is successful, leading to a complete transformation of GO to RGO. The XRD chart did not show any diffraction peak related to the predicted SrRuO 3 perovskite structure. There are three possible scenarios: i) SrRuO 3 is not formed; ii) instead, an amorphous SrRuO 3 structure is formed; or iii) a highly-dispersed SrRuO 3 in the nanometer or sub-nano size range is formed that is difficult to be distinguished in the XRD data. Microwave irradiation is based on an efficient superheating of the material so that the product may acquire different properties and structure 46 . Unlike the case of Sr-RG-C, XRD of Sr-RG-M (prepared by the microwave irradiation) shows the formation of Ru nanoparticles instead of RuO 2 (Fig. (1b)). The following diffractions are observed in the XRD chart of Sr-RG-M at 2θ = 38.34°, 42°, 43.96°, 58°, 69°, corresponding to XRD reference card (04-001-2957) of hexagonal Ru 47 . Some remaining traces of SrCl 2 and SrSO 4 are also observed, but with lower intensity compared to Sr-RG-C (prepared by the combustion method). This indicates that microwave irradiation minimizes the formation of impurities within the sample. XRD is considered a bulk analysis tool with high depth profiling and is convenient to provide some information on the structure of the bulk and surface of a sample. However, taking into account the XRD lower detection limit it is challenging to extract quantitative information for subnano-meter size particles from any spectra. For our prepared samples, it is thus difficult to resolve the signal from the baseline noise. Thus, we believe that XRD can be reliable to elucidate the structure of Sr-RG-C, but additional surface analysis tools should be used for any further assessment. XPS experiments are used to gain more information about the chemical composition, electronic state of different elements and their relative ratios in the prepared composites. Figure (1c-f) show the XPS spectra of Sr-RG-C and Sr-RG-M, allowing comparing the structures obtained by the different preparation routes described above. The C-1s spectra of both samples indicate the successful reduction of GO into RGO. The ratios between C-C/ C=C (284.4 ± 1.0 eV) and oxygenated peaks (C=O; 287.8 ± 0.1 eV) are calculated as 5.9 and 1.2 for Sr-RG-C and Sr-RG-M, respectively. This indicates that higher reduction efficiency is realized when using the combustion method. The peak at 279.9 ± 0.1 eV in both samples is assigned for Sr 3p 1/2 48 . The Ru 3d peak in Sr-RG-C sample (prepared by combustion) is deconvoluted into four peak components at 280.9 ± 0.1 eV and 285.0 ± 0.1 eV that correspond to Ru 3d 5/2 and Ru 3d 3/5 of RuO 2, and at 282.6 ± 0.1 eV and 286.56 ± 0.1 eV that correspond to Ru 3d 5/2 and Ru 3d 3/2 in SrRuO 3 , respectively. On the other hand, the Ru 3d peak in Sr-RG-M sample (prepared by microwave treatment) is deconvoluted into 5 peaks at 280.5 and 284.2 eV that correspond to Ru 3d 5/2 and Ru 3d 3/2 of metallic Ru, and at 281.9, 283.1 and 286.3 eV that correspond to Ru 3d 5/2 , its satellite and Ru 3d 3/2 of SrRuO 3 . The O 1s spectra of Sr-RG-C (Fig. 1e) is deconvoluted into three peak components, one at 528.6 eV is assigned to perovskitic oxygen 49 , the second at 529.7 eV is assigned to Ru-O* in RuO 2 3 , and the peak at 531.4 ± 0.1 eV is assigned to graphenic C-O*/C=O* 50 . In case of Sr-RG-M, the O 1s (Fig. 1f) peak could be deconvoluted to three peak components, one at 528.9 ± 0.1 eV assigned for the perovskitic oxygen, the second at 530.4 ± 0.1 eV corresponding to graphenic C-O* and/or adsorbed -OH 49 , the third component at 532.3 ± 0.1 eV assigned for graphenic C=O*. The XPS data of Sr 3d in Sr-RG-C are shown in Supplementary Fig. (S2a). The main 3d peak is deconvoluted into four components: two peaks at 132.0 ± 0.1 eV and 133.8 ± 0.1 eV corresponding to Sr 3d 5/2 and Sr 3d 3/2 of Sr 2+ in the perovskite structure with a binding energy separation of 1.8 eV, which is in agreement with the doublet splitting of Sr 3d. The other two peaks at 133.4 ± 0.1 eV and 135.1 ± 0.1 eV corresponding to Sr 3d 5/2 and Sr 3d 3/2 of SrCl 2 . For the microwave prepared sample (Sr-RG-M), the Sr 3d spectrum is deconvoluted into four peaks at 132.8 and 134.6 ± 0.1 eV corresponding to Sr 3d 5/2 and Sr 3d 3/2 of Sr 2+ in the perovskite structure and at 134.6 and 135.5 ± 0.1 eV corresponding to Sr 3d 5/2 and Sr 3d 3/2 of SrSO 4 . A quantitative analysis of the atomic surface percentages/ratios of different elements in the prepared composites is also done based on the atomic sensitivity corrected (normalized) intensities for different lines (the results are summarized in Table 1) 51 . The ratio of Sr 3d:Ru 3d:O1s deconvoluted peaks are assigned for SrRuO 3 to ascertain the ABO 3 structure. Thus, the calculated Sr 3d: Ru 3d: O1s p ratios in Sr-RG-C and Sr-RG-M are 1.2:1:3.1 and 1:1.3:3.1 that are consistent with ABO 3 structure. The appearance of XPS components of SrRuO 3 with high intensities indicates that SrRuO 3 exists on the surface of both Sr-RG-C and Sr-RG-M composites. The aforementioned findings indicate that SrRuO 3 is prepared successfully using both methods. It is important to consider the results from the high-resolution electron microscopy to examine the dispersion of the perovskites on the surface. HR-TEM images of Sr-RG composites are depicted in Fig. (2a & c) showing typical wrinkled and highly-folded graphene sheets with homogeneously distributed nanoparticles grown over their www.nature.com/scientificreports www.nature.com/scientificreports/ surfaces. The particle size distribution shows that most of the particles in Sr-RG-C are in the range of 1-3 nm (see Fig. 2a,b). In case of Sr-RG-M, a significant fraction of particles is in the sub-nanometer size range of ≈0.9 nm in diameter (see Fig. 2c,d). The fraction of particles larger than 3.2 nm in both samples is relatively low which sufficiently explains the absence of SrRuO 3 peaks in the XRD data and proves the successful preparation of sub-nano sized well-crystalline SrRuO 3 in combination with RuO 2 and Ru nanoparticles in case of Sr-RG-C and Sr-RG-M, respectively. The direct heating in air (as in case of combustion method) facilitates the oxidation of the fraction of metallic Ru in the sample to RuO 2 . The atomic ratio of SrRuO 3 to RuO 2 in Sr-RG-C as deduced from XPS measurements is nearly 2:3, while the ratio of SrRuO 3 to Ru nanoparticles in Sr-RG-M is about 4:3. The morphology of the formed material is strongly affected by the preparation method. Therefore, we investigated the effect of the preparation method on the morphology of the nanocomposites. Figure (2e-h) show the SEM images and the corresponding EDX analyses of Sr-RG-C and Sr-RG-M. Sr-RG-C in Fig. 2e shows fewer wrinkles and a highly porous structure, while Sr-RG-M in Fig. (2g) shows a sponge-like structure and the particles are distributed at the edges of graphene sheets in both cases. EDX analysis of Sr-RG-C shows that the atomic percentage of Ru is about 16.5 compared to 13.5 in the case of Sr-RG-M. A 7% sulfur content (as SrSO 4 ) is observed in the microwave sample that is consistent with our obtained XRD results. Very fine particles are observed at the edges of the graphene sheets, being also consistent with the obtained TEM images. Ba-RG nanocomposites. The XRD spectrum of Ba-RG-C (Fig. 3a) shows peaks with low intensities at 2θ = 28°, being characteristic for RuO 2 . In addition, the diffraction peaks at 25.8°, 31.5°, 42.5°, 55° and 65° are characteristic for the {411}, {510}, {631}, {755} and {1020} facets of cubic BaRuO 3 (reference card number 00-037-0846) 52 . As will be discussed later, the broad peaks recorded for Ba-RG-C indicate the formation of small-sized crystalline BaRuO 3 . On the other hand, Ba-RG-M shows an amorphous structure without any distinct peaks that would be characteristic for a crystalline perovskite structure (see Fig. 3b). However, in order to really elucidate their structures again additional analysis techniques are necessary. to Ba 3d 5/2 and Ba 3d 3/2 , respectively. Ba 3d in the case of Ba-RG-M is shifted to higher binding energies (about 0.5 eV). Consequently, we can conclude that crystalline BaRuO 3 is formed by the combustion method, while the well-crystalline perovskite did not form in the case of Ba-RG-M and its amorphous structure is most likely due to the formation of an intermediate state of BaRuO 3 . Table 1 summarizes the calculated atomic percentage of each element in the different Ba-RG samples as deduced from our XPS analyses. In order to confirm the structure of BaRuO 3 , we calculated the ratio Ba 3d 5/2 : Ru 3d 5/2 : O 1s (perovskitic) for Ba-RG-C and Ba-RG-M. The obtained values are 1.1:1:3.2 and 1:1.3:3.2 for Ba-RG-C and Ba-RG-M, respectively, being in agreement with ABO 3 perovskitic structures. HR-TEM images can provide extra information about the morphology of a sample. The corresponding images for both samples together with an analysis of the particle size distributions are shown in Fig. (4a-d). It can be noticed that Ba-RG-C shows less crimped graphene sheets and rather densely packed particles with less interspace separation in the distribution pattern. The particle size distribution shows that most of the particles are in the sub-to nanometer-range. This explains the large broadening of the XRD peaks of Ba-RG-C. On the other hand, HR-TEM of Ba-RG-M shows that the particles are larger and elongated, where the majority of particles are in the range of 5 to 25 nm with an average distribution of 10 nm. SAED-pattern of Ba-RG-C (see the inset of Fig. (S3c)) shows diffractions with d-spacing of: 1.58, 1.74, 1.86, 2.38 and 2.95 corresponding to {910}, {653}, {642}, {622}, {510} orientations, respectively, in the cubic BaRuO 3 reference card. In contrast, no diffractions were observed for Ba-RG-M. This finding indicates an amorphous structure of BaRuO 3 obtained by the microwave synthesis, while crystalline sub-nanometer sized BaRuO 3 is formed via the combustion-based synthesis (i.e., Ba-RG-C). The surface morphologies of our Ba-RG samples were also investigated by SEM (with EDX) as shown in Fig. (4e-h). Ba-RG-C (Fig. 4e) appears to have more wrinkles and lower porosity compared to Sr-RG-C, while the SEM of Ba-RG-M (Fig. 4g) reveals a fog-like shape with larger pore sizes compared to Ba-RG-C. Nitrogen is observed in the EDX of Ba-RG-C and Ba-RG-M that may be due to the formation of N-doped graphene. Ru atomic percentage in Ba-RG-C and Ba-RG-M is 16.5 and 11.8, respectively. Generally, the morphology of composites prepared by the microwave method is more ordered and particle shapes are more identified compared to those prepared by the conventional combustion method. However, the conventional combustion method produces particles with relatively smaller sizes and higher Ru contents. Ca-RG nanocomposites. The XRD pattern of Ca-RG-C illustrated in Fig. (5a) reveals the dominant formation of Ru nanoparticles as the main phase together with a smaller fraction of a CaRuO 3 phase with relative intensity of 6.7%. Surprisingly, the XRD of Ca-RG-M in Fig. (5b) reveals the formation of CaRuO 3 as the main phase with "some" Ru nanoparticles as a secondary phase. The remaining insoluble CaCO 3 appears in the XRD pattern of both samples. 54 . In this case, the diffractions of CaRuO 3 appear with distinct intensities and can be identified from the XRD data. This indicates that the microwave irradiation combined with a highly exothermic redox reaction between Ca and Ru precursors and GO lead to the formation of highly-crystalline CaRuO 3 without the need of any further calcination step. This finding is supported by our XPS analysis as shown in Fig. (5c-f). The C 1s spectra in Ca-RG-C is also deconvoluted into four peaks at 284.4, 286.8, 287.8 and 288.5 eV, corresponding to graphenic C-C/C=C, C-O, The observed Ru 3d spectrum of Ca-RG-C does not show any component related to metallic Ru, although the XRD pattern shows metallic Ru to be the main phase in Ca-RG-C. Considering the sensitivity of Ru to air exposure, the absence of metallic Ru could be attributed to either the oxidation of the outermost metallic Ru shell or to the presence of CaRuO 3 on the surface of the sample, suggesting the formation of Ru/CaRuO 3 core/ shell-nanoparticles. The absence of the lattice O 1s of RuO 2 in the O 1s spectra of Ca-RG-C makes the second suggestion more likely. Here, Ru 3d in Ca-RG-C is deconvoluted into four peaks at 281.7, 283.4, 285.9 and 287.3 eV, corresponding to Ru 3d 5/2 and Ru 3d 3/2 of Ru 4+ and their satellites, respectively, in the perovskite structure 55 . On the other hand, the Ru 3d spectrum in Ca-RG-M is deconvoluted into four peaks at 280.5 and 284.7 eV corresponding to Ru 3d 5/2 and Ru 3d 3/2 of metallic Ru and at 283 and 287 eV corresponding to Ru 3d 5/2 and Ru 3d 3/2 of the perovskitic Ru 4+ . The O 1s spectrum is also deconvoluted into four peaks in both samples with binding energies at 530.7, 532.1, 533.2 and 534.2 eV for Ca-RG-C and at 531.0, 532.0, 532.8 and 534.3 eV for Ca-RG-M. These correspond to perovskitic lattice oxygen, graphenic C-O*, graphenic C=O* and oxygen in carbonate species. The higher binding energy for lattice oxygen in the case of CaRuO 3 has also been reported in literature already 55 . On the other hand, Ca 2p is deconvoluted into four peaks at 348. 1. The calculated Ca 2p 3/2 : Ru 3d 5/2 : O 1s (perovskitic) ratios are 1:1:3.2 and 1:1:2.9 for Ca-RG-C and Ca-RG-M, respectively. The TEM image of Ca-RG-C (Fig. (6a)) shows graphene sheets with poorly distributed particles compared to Ca-RG-M, for which more compact structures are observable. Moreover, the microwave method results in graphene sheets with a high degree of folding, while the particles size distributions (Fig. 6b,d) show that both Ca-RG-C and Ca-RG-M have broad size distributions and larger particles (around 10 nm). The broad particle size distribution and the presence of relatively larger particles explain the appearance of the diffractions of CaRuO 3 in the XRD chart of Ca-RG composites compared to the other composites that have smaller sized particles with narrow size distributions. Our SEM images of Ca-RG-C and Ca-RG-M (Fig. 6e,g www.nature.com/scientificreports www.nature.com/scientificreports/ 40° and 54° that are consistent with tetragonal RuO 2 45 . The poor crystallinity may be an indication for unreacted Ru 3+ ions. Upon dispersion of RG-C in DMF the solution assumes a yellowish color, indicating the presence of unreacted RuCl 3 . The C 1s, O 1s and Ru 3d spectra of RG-C are shown in Fig. (S4b). The Ru 3d level is deconvoluted into four peaks: two peaks at 280.7 and 284.9 eV that correspond to Ru 3d 5/2 and Ru 3d 3/2 of RuO 2 and additional two peaks with higher areas at 282.0 and 286.0 eV that correspond to Ru 3d 5/2 and Ru 3d 3/2 of RuCl 3 . This finding also proves that most Ru 3+ ions are not completely converted into RuO 2 ; they rather remain in the form of RuCl 3 . The XPS spectrum of O 1s (Fig. (S4c)) is deconvoluted into three peaks at 529.3, 530.7 and 532.7 eV corresponding to oxygen bonded to transition metal (RuO 2 ), OH/C-O and O-C=O, respectively. The C 1s spectra are deconvoluted into three peaks at 284.4, 285.9 and 288.2 eV, corresponding to C-C/C=C, C-O and C=O, respectively. The total intensity of the C-C/C=C peak is higher than that of the oxygenated peaks, indicating the reduction of GO into RGO. It is worth noting that in our previous work 58 Ru-based RGO nanocomposites have been successfully prepared by the microwave method without using reducing or hazardous materials. We have found that a mixture of Ru and RuO 2 nanoparticles is successfully loaded on RGO by adjusting the pH of the precursors to 8.0, while pure Ru nanoparticles are loaded on RGO sheets if the pH is adjusted to 4.0 58 . Raman spectroscopy was used as additional evidence for the successful reduction of GO into RGO. Here, two bands appear for the GO sample: one at 1600 cm −1 (G-band) corresponding to the first scattering of the E 2g phonon of sp 2 carbon and a second band at 1334.5 cm −1 (D-band) that arises from a breathing mode of K-point phonons of A 1g symmetry with an I D /I G ratio 0.9. Another indication of the successful reduction of GO into RGO lies in the increased I D /I G ratio. The shifts in wave number of both G-and D-bands are also indications for RGO formation. The Raman spectra of RG-C and A-RG nanocomposites prepared by the current method are shown in Fig. (7). An increase in the I D /I G ratio is observed for all composites compared to GO. ## Conclusion Nano-sized Ru-based perovskites/RGO nanocomposites were prepared using a "green", one-pot low-temperature method. The preparation method is based on the redox reaction between a salt A (A = Ca, Ba or Sr), RuCl 3 as well as GO, initiated by either conventional combustion or microwave irradiation. The structural analyses of the resulting composites revealed that sub-nano sized SrRuO 3 was formed in combination with RuO 2 or Ru nanoparticles when using the conventional combustion or microwave irradiation, respectively. For the first time, highly crystalline nano-sized CaRuO 3 is formed as the main phase when using the microwave method. For BaRuO 3 particles with a diameter of ~1 nm was successfully prepared together with RuO 2 using the conventional combustion route, while amorphous BaRuO 3 was prepared by microwave irradiation. No further calcination step was needed to prepare such composites. Generally, the microwave irradiation route leads to ordered and well-distinct nanoparticles compared to those prepared by the conventional combustion method, which are usually smaller in size with highly porous structures. Consequently, microwave irradiation seems to promote the preparation of Ru nanoparticles, while the direct hotplate heating (conventional combustion method) promotes RuO 2 formation.

chemsum

{"title": "Synthesis, structural and morphological characterizations of nano-Ru-based perovskites/RGO composites", "journal": "Scientific Reports - Nature"}

building_blocks_for_recognition-encoded_oligoesters_that_form_h-bonded_duplexes

## Abstract: Competition from intramolecular folding is a major challenge in the design of synthetic oligomers that form intermolecular duplexes in a sequence-selective manner. One strategy is to use very rigid backbones that prevent folding, but this design can prejudice duplex formation if the geometry is not exactly right. The alternative approach found in nucleic acids is to use bases (or recognition units) that have different dimensions. A long-short base-pairing scheme makes folding geometrically difficult and is compatible with the flexible backbones that are required to guarantee duplex formation. A monomer building block equipped with a long hydrogen bond donor (phenol, D) recognition unit and a monomer building block equipped with a short hydrogen bond acceptor (phosphine oxide, A) recognition unit were prepared with differentially protected alcohol and carboxylic acid groups. These compounds were used to synthesise the homo and hetero-sequence 2-mers AA, DD and AD. 19 F and 31 P NMR experiments were used to characterize the assembly properties of these compounds in toluene solution. AA and DD form a stable doubly-hydrogen-bonded duplex with an effective molarity of 20 mM for formation of the second intramolecular hydrogen bond. AD forms a duplex of similar stability. There is no evidence of intramolecular folding in the monomeric state of this compound, which shows that the long-short basepairing scheme is effective. The ester coupling chemistry used here is an attractive method for the synthesis of long oligomers, and the properties of the 2-mers indicate that this molecular architecture should give longer mixed sequence oligomers that show high fidelity sequence-selective duplex formation. ## Background Two sequence-complementary strands of nucleic acid will form a stable duplex due to hydrogen bonding interactions between the bases. This supramolecular structure was immediately recognised to provide a plausible mechanism for information transfer between a template strand and a copy in the key biological processes of replication, translation and transcription, where the sequence of the copy is organised by the same basepairing interactions that lead to duplex formation. 1,2 These copying processes are currently unique to nucleic acids and represent the molecular basis for the evolution of life on this planet. Synthetic systems that form duplexes in the same way are therefore likely to provide a platform for template-directed synthesis of mixed sequence oligomers, and ultimately to the application of directed evolution for the discovery of new functional non-biological molecules. It is clear that duplex formation is not restricted to the precise molecular structure found in DNA and RNA. A range of nucleic acid analogues have been prepared in which the phosphate diester, the bases, 7, and the sugar have been replaced, and all of these oligomers form stable duplexes. Synthetic oligomers that bear no relation to nucleic acids have also been shown to form duplexes through various non-covalent interactions: metal-ligand coordination, 23,24 salt bridges, 25,26 aromatic interactions, 27 and hydrogen bonding. By using two different complementary recognition sites as the equivalent of the nucleic acid bases, it is also possible to encode sequence information into synthetic oligomers, and sequence-selective duplex formation has been demonstrated for short sequences. 26,31 We have been using a single hydrogen bond between a hydrogen bond donor (e.g. phenol, D) and a hydrogen bond acceptor (e.g. phosphine oxide, A) as the base-pairing interaction for duplex formation. This two letter alphabet allows information to be encoded in an oligomer as the sequence of A and D recognition sites. Provided the backbone does not contain any polar functional groups that could compete with the base-pairing interactions, the use of a single hydrogen bond as the base-pair removes any possibility of mismatches, because A cannot interact with A and D cannot interact with D. A number of different backbone architectures have been characterized, and the nature of the backbone was found to play a crucial role in the assembly properties of these oligomers. The different possible self-assembly channels are illustrated in Fig. 1. The key requirement for duplex formation is that the equilibrium constant for propagation of the intramolecular hydrogen bonds that zip up the duplex, K EM p , is greater than one (K is the association constant for formation of an intermolecular hydrogen bond, and EM p is the effective molarity for propagation of intramolecular hydrogen bonds in the duplex). 32,33 One of the competing assembly channels is formation of multiple intermolecular interactions that lead to higher order networks, but this process can be avoided by operating at a concentration, c, which is lower than the value of EM i , the effective molarity for formation of the frst intramolecular hydrogen bond that initiates duplex formation. The other major competing assembly channel is due to the formation an intramolecular hydrogen bond within an oligomer, which leads to folding. The probability of this process is determined by the equilibrium constant K EM f , where EM f is the effective molarity for folding. The values of the three effective molarity parameters depend on the conformational properties of the backbone. For the very flexible backbone shown in Fig. 2(a), the values of EM i and EM p are 10 mM to 30 mM, and the duplex channel dominates for length complementary homo-oligomers. 34 For the very rigid backbone shown in Fig. 2(b), similar results were obtained with EM i and EM p values of 40 mM to 70 mM. 35 Geometry is critical for more rigid backbones. The backbone shown in Fig. 2(b) has a well-defned geometry, which places the recognition groups in the correct orientation for duplex formation. However, for backbones of intermediate rigidity, where the conformational properties are more difficult to predict, mixed results were obtained. The backbone shown in Fig. 2(c) formed duplexes with EM i ¼ EM p ¼ 10 mM, 36 but the backbones shown in Fig. 2(d) and (e) did not lead to extended duplexes. For these two systems, EM i was similar to the values found for the other backbones (10 mM to 20 mM), but the geometry was not compatible with duplex propagation, and EM p was too low to measure. 37 The results obtained for homo-oligomers suggest that highly flexible backbones should provide a reliable platform for the design of duplex-forming oligomers. Conformational flexibility ensures that the backbone will always be able to adapt to a geometry compatible with base-pair formation in an extended duplex. More rigid backbones are difficult to design with the degree of accuracy required to guarantee the geometric complementarity needed for formation of an extended duplex. 37 The values of effective molarity measured for the very flexible backbone and the very rigid backbone shown in Fig. 2 are similar, so it appears that effective molarities associated with duplex formation are not adversely affected by conformational flexibility. Very flexible backbones are easily accessed, so this approach would make backbone design straightforward. However, the effective molarity for intramolecular folding, EM f , also depends on the conformational properties of the backbone. As shown in Fig. 3(a), a long flexible backbone promotes 1,2-folding between A and D recognition units that are adjacent in sequence. The value of EM f for this system is about 10 mM, which is comparable to the values of effective molarity for zipping up the duplex, so the folding channel will dominate for mixed sequence oligomers of this architecture. 39 Of course, longer mixed sequence oligomers will always be able to fold, no matter what backbone is used, and indeed sequence- 38 encoded folding of single-stranded RNA is key to the biological properties. 2 Folded nucleic acid structures involve looped out bases, so if a single-stranded nucleic acid is annealed with a sequence-complementary strand, duplex formation will dominate, because additional base-pairing interactions are made in the duplex. However, Fig. 1 shows that if 1,2-folding is possible, the number of base-pairs formed in the folding and duplex channels can be identical, so the folding channel will dominate. Minimising 1,2-folding is therefore critical to the design of recognition-encoded oligomers that form sequenceselective duplexes with high fdelity. One strategy for avoiding 1,2-folding is to reduce the value of EM f by increasing the rigidity of the backbone. As shown in Fig. 3(b), the very rigid backbone that we studied previously does not fold, so duplex formation is the dominant assembly channel for mixed sequence oligomers of this architecture. However, it would be preferable to work with more flexible backbones to guarantee duplex formation, as explained above. Here, we explore an alternative strategy for preventing 1,2folding in oligomers with a very flexible backbone. If two short bases are attached to a long flexible backbone, 1,2-folding is favoured (Fig. 4(a)). Fig. 4(b) illustrates how folding can be prevented by attaching the two short bases to a rigid backbone. Fig. 4(c) shows how changing the dimensions of the bases can be used to prevent folding. By making one of the bases longer than the other, the probability of fnding a backbone conformation compatible with folding is signifcantly reduced, and the duplex assembly channel should dominate. Fig. 4(d) shows the corresponding molecular design that we validate in this paper. It is worth noting that this short-long base-pairing scheme has similar geometrical properties to the purinepyrimidine base-pairing system found in nucleic acids. The backbone proposed in Fig. 4(d) uses ester linkages as the coupling chemistry for the synthesis of oligomers. Esters are sufficiently weak hydrogen bond acceptors (b z 5.5) not to compete signifcantly with the phosphine oxide recognition units (b z 10.5). 40 Ester coupling is sufficiently high-yielding to be used for the synthesis of polymers, and iterative coupling could be automated in a peptide synthesiser. Orthogonal protecting groups have been developed for the preparation of oligoesters with sequences of different building blocks. Here, we describe synthesis of the required monomer building blocks, demonstrate their use in the synthesis of different 2-mer sequences, and show that the long-short base-pairing scheme successfully prevents 1,2-folding for this oligomer architecture. ## Synthesis A divergent approach to the synthesis of the monomer building blocks was employed, in which a common aromatic bromide intermediate was coupled with the hydrogen bond donor and acceptor recognition units, as shown in Scheme 1. Commercially available 2-bromoethanol 5 was protected as the silyl ether 6, which was then used for alkylation of 4-bromoaniline to yield 7. Aniline 7 was alkylated with benzyl bromoacetate to give the key intermediate 8. Commercially available phenol 1 was converted to the boronic ester 2, which was coupled with 8 under Suzuki-Miyaura conditions to give the hydrogen bond donor monomer 9 (D). Treatment of commercially available diethyl phosphite 3 with iso-butylmagnesium chloride gave 4, which was coupled with 8 using palladium(0) and XantPhos to yield the hydrogen bond acceptor monomer 10 (A). For the ester coupling reactions, the potentially reactive phenol moiety in 9 was frst protected as the acetyl ester 11 (Scheme 2). The benzyl and TBDPS protecting groups in 10 and 11 were removed orthogonally to give the four precursors 12-15 required for ester coupling reactions. Treatment with hydrogen gas over palladium on charcoal gave the monoprotected carboxylic acids 12 and 14. Alternatively, reaction with n-tetrabutylammonium fluoride buffered with acetic acid gave the monoprotected alcohols 13 and 15. These monoprotected hydroxyacid monomers were used to synthesise three different 2-mer sequences by EDC coupling with a catalytic amount of N,N-dimethylaminopyridine (Scheme 3). Coupling 14 with 15 gave AA directly. AD and DD were obtained with the phenol groups protected as acetate esters, but these groups were removed quantitatively by stirring in a solution of ammonium acetate in water and methanol. ## NMR binding studies Duplex formation and folding were investigated using 19 association constant for formation of the A$D complex, which makes a single intermolecular hydrogen bond, was measured by titrating A into D. A large upfeld change in the 19 F NMR chemical shift of D was observed, and the data ft well to a 1 : 1 binding isotherm to give an association constant of K A D ¼ 3.8 10 3 M 1 (Table 1). The association constant for the AA$DD complex was similarly measured by titrating AA into DD, and the association constant for dimerization of AD was determined by a 19 F NMR dilution experiment in toluene-d 8 at 298 K. The association constants for the AA$DD and AD$AD complexes are both two orders of magnitude higher than that for A$D, which indicates that there are two cooperative hydrogen bonding interactions in the complexes formed by the sequence complementary 2-mers (Table 1). The limiting 19 F and 31 P NMR chemical shifts of the free species (d free ) and fully bound complexes (d bound ) were determined by extrapolation of the binding isotherms (Table 1). The values are similar for all three complexes. The large upfeld limiting complexation-induced changes in 19 F NMR chemical shift (0.4 ppm) indicate that all of the phenol groups form hydrogen bonds in all of the complexes. The large downfeld limiting complexation-induced changes in 31 P NMR chemical shift (5-7 ppm) indicate that all of the phosphine oxide groups form hydrogen bonds in all of the complexes. Comparison of the values of the free 19 F and 31 P NMR chemical shifts of AA, DD, and AD show that there is no signifcant intramolecular hydrogen bonding in AD, i.e. there is no folding in the monomeric state. The results indicate that both the AA$DD and AD$AD duplexes are fully assembled through the intended base-pairing interactions at mM concentrations in toluene solution at room temperature as shown in Fig. 5. A schematic representation of the equilibria involved in duplex assembly is shown in Fig. 6. For AA$DD, formation of the frst intermolecular hydrogen bond gives an open complex, and formation of the second intramolecular hydrogen bond gives the closed duplex. Assuming that all of the hydrogen bonds in the systems described here are of similar strength, it is possible to describe the association constant for formation of the closed c-AA$DD duplex in terms of the association constant for formation of a single intermolecular hydrogen bond K A$D and the effective molarity for the intramolecular interaction EM i . The backbone in these systems has a direction, because the hydroxyl and acid ends are different, so parallel and antiparallel orientations of the duplex are possible. As the end groups are spatially separated from the recognition sites, we assume that the two possible c-AA$DD have similar stability. Therefore, the open complex o-AA$DD has four equally populated states and the closed duplex c-AA$DD has two. It is possible to express the association constants for duplex formation in terms of K A$D and EM i : Hence the effective molarity for duplex formation can be determined as: The association constants in Table 1 were used to calculate EM i for this system as 19 AE 3 mM, which is consistent with values of supramolecular effective molarities we have measured for other hydrogen bonded duplexes. 31, 38,39 The equilibrium constant for closing the duplex is given by 1 2 K A$D EM i and is 40 for this system, which implies that the duplex is fully closed and only 2% of the species populate the partially-bound open state o-AD$AD. For the closed hetero-2-mer duplex c-AD$AD, there is no degeneracy associated with the backbone directionality, because the anti-parallel orientation is determined by the sequence. However, there is the possibility of intramolecular 1,2-folding in the monomeric state, which is governed by the corresponding effective molarity EM f . Hence, the observed dimerisation constant K AD$AD depends on the concentrations of the folded (AD folded ) and open (AD open ) species that are populated in the monomeric state: Assuming that the effective molarity for duplex formation, EM i , is the same for AA$DD and AD$AD, it is possible to combine eqn ( 2) and (4) to determine (K A$D EM f + 1), which is the factor that describes the fraction of monomeric AD that exists in the folded state: Substituting the values from Table 1 into eqn (5) gives a value of 1.0 for (K A$D EM f + 1), which is consistent with the NMR chemical shift data. These results indicate that virtually all monomeric AD exists in the open state and the 1,2-folding does not compete with duplex formation in this system. If the two arrangements of the c-AA$DD were not degenerate, the statistical factor in eqn (1) would be equal to one, giving (K A$D EM f + 1) z 1.4. This value would require that 30% of monomeric AD exists in the folded state, which is not consistent with the NMR chemical shift data, suggesting that assumption that the parallel and antiparallel backbone arrangements are equally populated in the c-AA$DD duplex is reasonable. ## Molecular mechanics calculations The competition between duplex formation and intramolecular 1,2-folding in AD were further investigated using molecular mechanics calculations. The OPLS3 force feld with implicit chloroform solvation model was employed, as implemented in the MacroModel software (the experiments were carried out in toluene, but chloroform is the only non-polar implicit solvent model implemented). 52 A conformational search was performed on the AD monomer and the lowest energy structure, shown in Fig. 7(a), is a folded species. The calculation is clearly inconsistent with the experimental results, reinforcing our previous fndings that computational methods do not provide a reliable method for predicting the self-assembly properties of synthetic molecules of this complexity. 39 To investigate whether this folded structure is strongly preferred over duplex formation by the force-feld, two molecules of AD were constrained to have one intermolecular hydrogen bond, and a conformational search gave the closed c-AD$AD duplex shown in Fig. 7(b) as the lowest energy structure. No open o-AD$AD structures were found within 5 kJ mol 1 of the minimum. The calculated energy of the duplex is 87 kJ mol 1 lower than the energy of two folded monomers, which suggests that there is considerable strain associated with folding in this system. ## Double hydrogen bonding Oxygen hydrogen bond acceptors can interact with more than one hydrogen bond donor, which can degrade the fdelity of sequence-selective duplex formation. 31 In order to investigate whether the base-pair recognition system used here would suffer from this problem, A was titrated into DD. The changes in 19 F NMR chemical shift of the DD did not ft to a 1 : 1 isotherm (see ESI †), so a 1 : 2 binding model was investigated: The two donor binding sites were assumed to be independent and identical, hence 2 could be fxed in the least squares regression analysis. The association constant for the DD$A was determined to be K 1 ¼ (15 000 AE 2000) M 1 , which is four times greater than the single hydrogen bond association constant K A$D and suggests additional stabilisation due to a hydrogen bond between the second phenol and the phosphine oxide. We can represent the equilibria leading to the doubly bonded complex as in Fig. 8. Noting that both 1 : 1 complexes give rise to the observed association constant, K 1 can be expressed as: The association constant for the formation of the second hydrogen bond is, therefore: Using eqn ( 9) and the measured value for K 1 , the association constant for the interaction of the second phenol donor with the acceptor is K 0 (1.0 AE 0.2), which means that the doublebonded complex represents 50% of the 1 : 1 complex. The ratio of K A$D EM i and K 0 describes the competition between a correctly recognised duplex and a doubly hydrogen-bonded mismatched complex. This ratio is 80 for this system, therefore sequence selectivity should be achieved for longer information oligoesters with fdelity of 99%. For comparison, the previously reported sequence-containing information oligomer shows K 0 ¼ 1.6 and K A$D EM i ¼ 9.9, hence exhibits sequence fdelity of 86%. 31,39 While the value of K 0 for the system described here is comparable with that reported earlier, the exceptionally strong hydrogen-bonding interaction between the recognition units should lead to superior performance the formation of closed duplexes with high sequence fdelity. ## Conclusions In conclusion, candidates for new information molecules were synthesised and their behaviour in toluene was studied through 19 F and 31 P NMR spectroscopy. The monomeric building blocks are readily accessible and 2-mers were easily synthesised through efficient ester coupling reactions, with scope for the synthesis of longer oligomers using the same methodology. A long-short base-pairing scheme akin to purines and pyrimidines in natural nucleic acids was employed in order to reduce intramolecular folding and a flexible backbone was used to ensure the geometric complementarity required for duplex formation. Homo-and hetero-2-mers were observed to form stable duplexes in toluene at 298 K with effective molarities for duplex formation of 20 mM and without any substantial 1,2folding. The observed trends were consistent with those previously reported using 31 P NMR, thus providing a convenient handle for studying supramolecular association. Formation of double hydrogen bonds to the oxygen-based acceptor was found to be much less favoured than the desired base-pairing interactions. This system appears to be ideally suited to the synthesis of longer oligomers which are expected to show the possibility of high-fdelity sequence-specifc information recognition via hydrogen bonding in organic solvents.

chemsum

{"title": "Building blocks for recognition-encoded oligoesters that form H-bonded duplexes", "journal": "Royal Society of Chemistry (RSC)"}

photoinduced_metal-free_borylation_of_aryl_halides_catalysed_by_an_<i>in_situ</i>_formed_donor–accep

## Abstract: Organoboron compounds are very important building blocks which can be applied in medicinal, biological and industrial fields. However, direct borylation in a metal free manner has been very rarely reported. Herein, we described the successful direct borylation of haloarenes under mild, operationally simple, catalyst-free conditions, promoted by irradiation with visible light. Mechanistic experiments and computational investigations indicate the formation of an excited donor-acceptor complex with a À3.12 V reduction potential, which is a highly active reductant and can facilitate single-electron-transfer (SET) with aryl halides to produce aryl radical intermediates. A two-step one-pot method was developed for site selective aromatic C-H bond borylation. The protocol's good functional group tolerance enables the functionalization of a variety of biologically relevant compounds, representing a new application of aryl radicals merged with photochemistry. ## Introduction Boronic esters are commonly encountered in pharmaceuticals, functional materials and agrochemicals. 1 Aryl halides and related arenes are the most commonly employed precursors to produce them. 2 Conventional routes for their synthesis include metalation (generally by Li or Mg reagents) 3 followed by nucleophilic substitution and transition-metal catalysis, 4 offering a broad substrate scope and good functional group compatibility if their disadvantages such as high cost, low reactivity and operational inconvenience could be ignored. Visible light catalysis 5 has become a well-accepted and powerful method for the direct borylation of drug like molecules. Seminal studies by Larionov, 6 Aggarwal, 7 Wu, 8 Schelter 9 and others 10 have established the viability and synthetic utility of this approach, circumventing the need for well recognized photocatalysts and metal additives. Studer 11 developed radical borylation of aryl iodides with bis(catecholato)diboron (B 2 cat 2 ) as the boron source under mild conditions. Recently, Jiao 12 reported the borylation of aryl halides using 4-phenylpyridine as an efficient catalyst and sodium methanolate as a strong base with bis(pinacolato)diboron (B 2 pin 2 ). Different from Jiao's work, Li, 13 Larionov 14 and Lin 15 independently described simple and efficient methods to convert aryl triflates and aryl halides into aryl radicals under UV irradiation or reductive electrophotocatalysis. There is a constant quest for more efficient clean borylation strategies (e.g., metal-free visible light catalysis using an in situ formed photocatalyst) from simple and readily available starting materials. In recent years, in situ formed donor-acceptor complexes have been successfully used in the single-electron-transfer (SET) process to produce radical species by hybridization and modulation of the relevant energy levels. 16,17 As a result, visible light excitation can occur favoured by a red shift of the absorption band, although both donor and acceptor components are individually insensitive to visible light irradiation. This novel protocol can provide a new approach for aryl radical intermediates through C-X bond activation in a metal-free manner (Scheme 1). In 2020, we reported a photoinduced dehalogenation of aryl halides. 18 Spectroscopic, computational and chemical studies indicate that the formation of a twisted intramolecular chargetransfer (TICT) species enables the population of higher-energy doublet states. König 19 developed a versatile photocatalytic strategy for the ipso-borylation of substituted arenes by using thiolate as a catalyst, forming a donor-acceptor complex between thiolate/B 2 pin 2 and boryl-anion-activated substrates. Inspired by our previous work and related reports, 20 we questioned whether borylation of aryl halides could be achieved with the photocatalytic intermediate formed in situ employing simple available substrates (e.g., N-containing heterocycles/ organic base/boron source). Herein, we described the successful direct borylation of haloarenes under mild, operationally simple, catalyst-free conditions, promoted by irradiation with visible light. ## Results and discussion We made our initial discovery of the borylation reaction employing 4-bromoanisole as the model substrate (Table 1). First, under visible light photoirradiation, when the 2,2 0 -bipyridine (nitrogen containing heterocycle: NCH)/triethylamine (NEt 3 ; fnal reductant)/B 2 pin 2 (boron source) system was evaluated in acetonitrile, the desired product can be detected in 27% yield (entry 1). In order to generate the optimal ateintermediate to facilitate the single-electron-transfer (SET) process, a series of NCHs were evaluated (entries 2-5). Only trace desired product was detected using the 2-phenyl pyridine/ NEt 3 /B 2 pin 2 system (entry 2) suggesting that the ateintermediate formed with 2-phenyl pyridine couldn't deliver a single electron to the aromatic C-Br bond. To our delight, the yield was improved remarkably (49% yield) employing isoquinoline as the ate-intermediate precursor (entry 3). Neither acridine (entry 4) nor 4-trifluoromethyl pyridine (entry 5) could beneft the borylation reaction, indicating that steric resistance and electron-defcient factors of NCHs can decrease the reductive ability of the ate-intermediate. A brief investigation of the solvent effect demonstrates that DMSO (entry 6), DMF (entry 7) and DCE (entry 8) are not as efficient as acetonitrile. Gratifyingly, the desired product could be isolated with 81% yield via increasing the loading of B 2 pin 2 /NEt 3 and prolonging the photoirradiation process (entry 9). Not surprisingly, control experiments without NCH (entry 10) and in the dark (entry 11) couldn't afford the desired product effectively. With the optimized reaction conditions in hand, we frst examined the substrate scope of aryl bromides (Table 2). Scheme 1 Synthetic approaches to aryl boronic esters. ## Table 1 Optimization of the reaction conditions a Entry Solvent B 2 pin 2 (X eq.) Various para-substituted aryl bromides, including electron donating group substituted (3a-3f), nonsubstituted (3g), unprotected -OH (3h)/-NH 2 (3i), electron withdrawing group substituted (3j-3o) and even simple alkene moieties (3o), could be well tolerated in this system, delivering the corresponding arylboronates in good to excellent yields. However, bromostyrene was not tolerated. A dual borylation product (3q) could be synthesized using dihaloarene or haloaryl boronic ester. Meta-(3r-3t) and ortho-(3u-3w) substituted arylboronates can both be obtained in good yields (e.g., 3w, 1-allyl-2bromobenzene, 63% yield). Moreover, 2,5-(3x)/3,4-(3y-3ac)/3,5disubstituted (3ad-3ae) aryl bromides produced borylation products in good yields (3aa and 3ac, 69% and 64%, respectively). Heteroaromatic bromides, such as benzothiophene (3ag) and indole (3ah) derivatives, can be employed to access the desired borylation products with good efficiencies other than brominated benzofuran (3af, acceptable 23% yield). This methodology also worked well with a series of diboron reagents to produce arylboronates 4a-6a with useful efficiencies (e.g., 4a, 69% yield). In contrast, B 2 (OH) 4 was not a good borylation candidate with a poor solubility in acetonitrile. Notably, except for aryl halides, 3a could also be prepared through more challenging Ar-O (7a)/Ar-N (8a) bond cleavage under standard conditions. Surprisingly, reductive deoxygenation of fluorenone was accomplished with high efficiency (9a, 65%). Next, we turned our attention to exploring the substrate scope of the borylation of unactivated aryl chlorides, since cleavage of the C-Cl bond in a metal-free manner remains quite challenging [e.g., for PhCl, E red ¼ 3.04 V vs. SCE]. 21 Our investigations revealed that both electron-rich and electron-defcient chlorobenzene derivatives could be employed to provide the corresponding arylboronates in good yields with a little more loading of NCH and longer irradiation time (Table 3; 68% NMR yield for 20 mol% NCH loading and 36 h irradiation vs. 82% NMR yield for 100 mol% NCH loading and 72 h irradiation). It's worth noting that the reaction was allowed for production of fluorinated boronic ester (11h). In addition, unprotected para, meta, and ortho -OH/-NH 2 substituted aryl chlorides (10e-10f; 10j; 10l) were well tolerated, illustrating the immediate utility of this approach in preparing value-added boronic ester in a protecting group-free manner. Similar to aryl bromide substrates, more electronrich and sterically hindered aryl chlorides (10m-10p) could also be transformed into desired products with good yields using this new protocol (Table 4). Site selective aromatic C-H functionalization provides rapid access to useful structural and functional molecular complexity. 22 Based on the optimized reaction conditions, we developed a two-step one-pot method for simple aromatic C-H borylation. In the frst step, almost equivalent regioselective bromination is accomplished using N-bromosuccinimide (NBS) as the bromination reagent, controlled by the electrophilicity and steric hindrance effect of the aromatic ring. 23 Subsequently, a series of aryl bromides without further purifcation were transformed into the corresponding boronic esters with high regioselectivities and efficiencies (e.g., 13a, 73% yield; rr up to 40 : 1). Aromatic silyl-ethers (13c-d), unprotected phenol (13e), aniline (13f) and polysubstituted benzenes (13g-h) were well tolerated, delivering site selective products with moderate to good yields (Table 5). The broad functional group tolerance and two-step one-pot method for site selective aromatic C-H borylation encouraged us to test this mild strategy for late stage functionalization of complex molecules, such as natural products and active pharmaceutical ingredients (APIs). Boronates derived from methyl dehydroabietate ( 14), amino acid tyrosine (15), and a variety of APIs (16-18) containing polar groups, sterically hindered substituents and polycyclic frameworks were still readily produced under the optimized conditions, highlighting the utility of this methodology in the late-stage functionalization of biologically relevant compounds. To obtain insights into the reaction mechanism, an N,Ndiboronate complex 24 (dimer A) was prepared through reductive coupling of isoquinoline and B 2 pin 2 (Scheme 2a). Under photoirradiation, no reaction of 4-bromoanisole and dimer A was Table 3 Substrate scope of aryl chlorides a a Reactions were run on a 0.2 mmol scale; isolated yield. b 20 mol% NCH (isoquinoline) was used and irradiation for 36 h. c NMR yield. Table 4 Two step, one-pot method for Ar C-H bond borylation a a Reactions were run on a 0.2 mmol scale; isolated yield. detected without addition of NEt 3 (see the ESI †). In contrast, the potential excited donor-acceptor complex was able to slowly reduce 4-bromoanisole into anisole in the presence of NEt 3 ; when additional B 2 pin 2 (2.0 eq.) was applied as the trapping reagent without loading of NEt 3 , no dehalogenation product except boronic ester product was detected; furthermore, the reduction of 4-bromoanisole took place more efficiently in the presence of NEt 3 and B 2 pin 2 , delivering the corresponding boronic ester as the major product (Scheme 2b). These observations indicated that the reduction ability of the coordination complex of acceptor dimer A and donor NEt 3 has indeed been enhanced after photoexcitation. To obtain a clearer picture of the complex of acceptor dimer A and donor NEt 3 , density functional theory (DFT) calculations were carried out. First, the computational UV-Vis spectrum was employed to compare with the experimental UV-Vis spectrum (Scheme 2c). Without NEt 3 , there exist two smooth peaks in the computational UV-Vis spectrum of dimer A and the peak at 227 nm is higher than that at 301 nm, which is in agreement with the experiments (Scheme 2c(I) and d(III)). When one NEt 3 molecule binds to dimer A in silico, the curve between the two peaks becomes sharp, which is also consistent with the experimental UV-Vis spectrum (Scheme 2c(II) and d(IV)). However, when two NEt 3 bind to dimer A, the computational UV-Vis spectrum reveals that the peak around 220 nm becomes lower than the peak around 300 nm, which is quite different from experiments (Scheme 2d(V)). Therefore, the computational results suggest that only one NEt 3 binds to dimer A under the optimized reaction conditions. In addition, the binding energy between dimer A and one NEt 3 is 11.6 kcal mol 1 by calculation. Considering that the ratio of dimer A (20 mol% isoquinoline / 10 mol% dimer A) to NEt 3 (5.0 eq.) is 1 : 50 under the optimized reaction conditions, dimer A and complex B can equilibrate (Scheme 2d). After B is determined to be the complex combining acceptor dimer A and donor NEt 3 , time dependent TD-DFT calculations were conducted for the excited-state of B. The excited-state reduction potential of B* is 3.12 V, indicating that B* is a highly active reductant and can facilitate SET with aryl halides to produce aryl radical intermediates. With this mechanistic insight, a plausible mechanism for the conversion of aryl halides to their corresponding boronic esters is proposed in Scheme 3. Reductive coupling of isoquinoline and B 2 pin 2 generates dimer A, which can coordinate with NEt 3 in situ to afford intermediate B. Donor-acceptor complex B is then excited by 390 nm light, populating a highly reducing excited state B*, which can undergo a single electron transfer process with an electronically matched aryl halide to generate a radical cation species C, aryl radical D, and regenerate isoquinoline/B 2 pin 2 . Radical cation C could be transformed into enamine E followed by hydrogen atom transfer and deprotonation processes. Finally, radical D is trapped by B 2 pin 2 immediately, yielding the desired boronic ester. ## Conclusions In summary, a facile transition-metal-free conversion of aryl halides to the corresponding boronic esters through a radical process has been developed. A series of aryl boronic esters were synthesized with good functional group tolerance in moderate to excellent yields under mild conditions. Mechanistic experiments and computational investigations indicate the formation of an excited donor-acceptor complex, serving as the super single electron reductant. Furthermore, the key role of the donoracceptor complex formed in situ was confrmed by control experiments and DFT calculations. The protocol's functional group tolerance and two-step one-pot method for site selective aromatic C-H bond borylation enabled the functionalization of a variety of biologically relevant compounds, representing a new application of aryl radical merged with photochemistry.

chemsum

{"title": "Photoinduced metal-free borylation of aryl halides catalysed by an <i>in situ</i> formed donor\u2013acceptor complex", "journal": "Royal Society of Chemistry (RSC)"}

cooperative_stabilisation_of_14-3-3σ_protein–protein_interactions_<i>via</i>_covalent_protein_modifi

## Abstract: 14-3-3 proteins are an important family of hub proteins that play important roles in many cellular processes via a large network of interactions with partner proteins. Many of these protein-protein interactions (PPI) are implicated in human diseases such as cancer and neurodegeneration. The stabilisation of selected 14-3-3 PPIs using drug-like 'molecular glues' is a novel therapeutic strategy with high potential.However, the examples reported to date have a number of drawbacks in terms of selectivity and potency. Here, we report that WR-1065, the active species of the approved drug amifostine, covalently modifies 14-3-3s at an isoform-unique cysteine residue, Cys38. This modification leads to isoformspecific stabilisation of two 14-3-3s PPIs in a manner that is cooperative with a well characterised molecular glue, fusicoccin A. Our findings reveal a novel stabilisation mechanism for 14-3-3s, an isoform with particular involvement in cancer pathways. This mechanism can be exploited to harness the enhanced potency conveyed by covalent drug molecules and dual ligand cooperativity. This is demonstrated in two cancer cell lines whereby the cooperative behaviour of fusicoccin A and WR-1065 leads to enhanced efficacy for inducing cell death and attenuating cell growth. ## Introduction The 14-3-3 family of hub proteins (seven isoforms: b, g, 3, h, s, s, z) play diverse and important roles in maintaining normal cell function through interaction with over 200 partner proteins. 1,2 These protein-protein interactions (PPIs) are typically dependent on the phosphorylation of specifc recognition motifs within disordered domains of the partner protein. Through these PPIs, 14-3-3 proteins modulate the subcellular localisation, protein folding, enzymatic activity or interaction profle of their partners. 3 Many 14-3-3 PPIs are implicated in human diseases, for example those in which there is an involvement of Raf kinases (cancer), 4 Cdc25 (cancer), 5 CFTR (cystic fbrosis), 6 tau (Alzheimer's disease), 7 LRRK2 (Parkinson's disease), 8 ERa (breast cancer) 9 and p53 (cancer). 10 Thus, 14-3-3 PPIs have attracted increasing interest as potential drug targets. 1 In particular, the stabilisation of certain 14-3-3 PPIs has enormous potential because of the desirable therapeutic benefts in anticancer therapy. Stabilisation of PPIs also presents an opportunity to selectively target unique protein-protein interfaces because shape complementarity is required with both protein partners. Selective inhibition in this context poses the much greater challenge. Examples of small molecule PPI stabilisers remain rare, 11 and the cooperative nature of ternary complex formation presents a challenge in terms of ligand discovery and optimisation. 12 A limited number of 14-3-3 PPI stabilisers have been reported, the majority of which target, or are predicted to target, pockets at the protein-protein interface within the amphipathic binding groove of 14-3-3. 1 Examples include the fusicoccane family of diterpene natural products (e.g. 1, fusicoccin A, Fig. 1A), 13 epibestatin, 14 pyrrolidone, 14 pyrazole, 15 and supramolecular ligands. 16 These compounds have a number of drawbacks in terms of synthetic tractability, drug-likeness, potency and/or selectivity. Furthermore, they do not distinguish between the seven human 14-3-3 isoforms that each have distinct roles. For example, gathering structure-activity relationship data around the fusicoccane family requires complex semi-synthesis, and the most widely available member, 1, targets all 14-3-3 isoforms and at least three different binding interfaces (ERa, CFTR and p53). Therefore, it is important to identify and rationally develop novel 14-3-3 stabilisation mechanisms that may circumvent such drawbacks. Amifostine (2) (Fig. 1A) is a clinically approved prodrug administered to alleviate side effects associated with chemotherapy and radiotherapy. 17 In vivo, 2 is dephosphorylated to release the active species, aminothiol WR-1065 (3, Fig. 1A). 3 exerts its effects through a number of possible mechanisms, most predominant of which is the scavenging of harmful reactive oxygen species. 17 Notably, 3 is known to modulate the activity of the transcription factors NFkB, AP-1 and p53 via covalent disulphide bond formation in cellular environments. 18 There is also compelling evidence that 3 enhances wild-type p53 activity 19 and rescues the activity of p53 mutants by modulating protein conformation. 20 Furthermore, 2 promotes the interaction of p53 with 14-3-3s presumably via 3 as the active species. 21 This PPI preserves p53 protein levels in the cell and enhances its transcriptional activity. 21,22 p53 interacts with 14-3-3s via one or more phosphorylated motifs in its disordered C-terminal domain (e.g. pT387, Fig. 1B). 22 To date, only the p53 DNA-binding domain has been associated with covalent modifcation by drug molecules. 23 Therefore, current evidence does not point to any clear rationale for how 3 could directly influence the p53 C-terminal domain interaction with 14-3-3s PPI. Here we report that 3 stabilises two distinct 14-3-3s PPIs (with p53 and ERa) via covalent protein modifcation of 14-3-3s at Cys38, a solvent exposed residue unique to this human 14-3-3 isoform (Fig. 1B). It exerts its effect via a mechanism that also enhances the effect of the known stabiliser 1 in a cooperative manner. The stabilisation effect is not specifc to a single 14-3-3s PPI, but it is specifc to the sigma isoform in vitro. Our fndings demonstrate a new mechanism for 14-3-3s stabilisation that can harness the enhanced potency of covalent drugs, and combination treatment, to achieve a desirable therapeutic response. The enhanced efficacy of combination treatment of 1 and 3 against cancer is demonstrated in two cancer cell models. Combination treatment more effectively induced a p53-specifc senolytic effect on senescent EJp53 cells and was more effective at attenuating the growth of estrogen-induced MCF-7 cell growth. Our results also highlight a new mechanism that can ultimately be used to achieve 14-3-3s isoform specifcity, and is likely to contribute to the in vivo pharmacodynamics of 2. ## Results and discussion WR-1065 covalently modifes 14-3-3s Cys38 Because 3 had previously been shown to elicit a biological response through covalent modifcation of target proteins, we frst used mass spectrometry to establish if it also reacted with Cys38 of 14-3-3s. In DMSO solution, 3 oxidised to form a homodimeric disulphide. This species reacted quantitatively with 14-3-3s using an equimolar ratio of reactants (both 100 mM) under non-reducing conditions, as shown by mass spectrometry analysis (Fig. 2A and S1 †). Despite the in vitro requirement for non-reducing conditions, 3 is known to covalently modify proteins in the cellular environment, and this might be particularly prevalent in cells under oxidative stress. 18 Disulphide exchange was unaffected by prior complexation of 14-3-3s with a phospho-peptide mimicking the binding motif of p53. The corresponding 14-3-3s mutant Cys38Ala was not modifedsupporting the hypothesis of Cys38 covalent modi-fcation (Fig. S2 †). This was expected because the only other Cys 3). (B) Crystal structure showing a phospho-peptide representing the C-terminus of p53 (pT387, stick representation, C atoms in green) in complex with 14-3-3s (grey). The position of Cys38 is indicated in yellow. PDB: 5MOC. 10 The binding site of 1 is indicated by the structure of 1 shown in faded colours (overlaid from PDB: 5MXO). 10 Fig. 2 Amino disulphides covalently modify 14-3-3s. (A) Deconvoluted mass spectra for 14-3-3s before and after modification by the homodisulphide of 3. (B) Structures of disulphide-containing molecules screened for binding to 14-3-3s (% ligation is given in parentheses). residue in the 14-3-3 sequence, Cys96, is not solvent exposed (based on the available structural data). Cys38 was previously shown to undergo disulphide exchange in a tethering screen. 24 However, no effect on partner protein binding has previously been attributed to modifcation of the 14-3-3s-unique Cys38 residue. Covalent imine tethering to Lys122 of 14-3-3 proteins has yielded stabilisers that are selective for specifc 14-3-3 binding partners, but Lys122 is conserved across the 14-3-3 family and thus its modifcation would not constitute an isoform specifc approach. 25,26 To further profle the reactivity of Cys38, 14-3-3s was incubated with the small sub-set of disulphide containing compounds 4-7, again using 100 mM of both reactants, and analysed by mass spectrometry (Fig. 2B and S3A-D †). Cystamine ( 4), an endogenous disulphide-containing molecule that is structurally related to 3, reacted quantitatively with the protein, whilst the non-natural di-N-propyl disulphide ( 5) and dithiopropionic acid (6) did not show any reactivity. This suggests that covalent protein modifcation at Cys38 is reliant on a basic amine in proximity to the electrophilic disulphide bond. Glutathione disulphide (7), a cell endogenous species, showed some reactivity, resulting in 8% ligation to the protein under these conditions. This is interesting because it suggests that Cys38 could play a broader chemical role in the cellular environment, and contribute to the unique activity profle of 14-3-3s. WR-1065 stabilises 14-3-3s PPIs Two 14-3-3 partner protein binding motifs were compared in order to establish if 3 selectively influenced their binding to 14-3-3s. The motifs for p53 (ref. 10) and for ERa 9 were selected because of their different 14-3-3 binding profles and our interest in them as potential oncology drug targets. Of the three possible binding motifs in p53, that around pT387 has been studied in most detail at the molecular level. It adopts a unique binding conformation with a turn local fold (Fig. 1B). 10 This interaction is moderately stabilised by 1 (4). 10 By contrast, ERa has a characteristic 'mode 3' C-terminal binding conformation that binds with 48-fold greater affinity and is preferentially stabilised by 1 (34). 9 Fluorescence polarisation (FP). FP assays were used to determine if covalent modifcation of Cys38 by 3 influenced the binding affinity of partner protein binding motifs to 14-3-3s. 14-3-3s was titrated to a fxed concentration of fluorescently labelled phospho-peptide designed to mimic the respective 14-3-3s binding motif. In the absence of 3, 14-3-3s bound to p53 and ERa phospho-peptides with apparent dissociation constant (K d ) values of 8.1 AE 1.2 mM and 0.17 AE 0.02 mM respectively, in agreement with previous reports. 9,10 In the presence of a molar excess of 3, the affinity of 14-3-3s binding to p53 was only enhanced by 1.4-fold (K d / 5.6 AE 0.8 mM, Fig. 3A). However, there was a signifcant effect on the 14-3-3s interaction with ERa whereby the affinity was enhanced by 2.8-fold (K d / 59 AE 5.3 nM, Fig. 3B). Next, we investigated if 3 influenced the stabilising effect of 1 on these PPIs. In the presence of fxed concentrations of 1, the apparent K d for 14-3-3s binding to ERa and p53 was lowered to 6.7 AE 0.6 nM (25) and 2.1 AE 0.2 mM (4) respectively (Fig. 3A and B). Both observations were again consistent with literature reports. 9,10 Intriguingly, when 3 was used in combination with 1, the stabilising effect on the 14-3-3s interaction with ERa was further enhanced by 3-fold: apparent K d 6.7 AE 0.6 nM / 2.3 AE 0.1 nM (Fig. 3B). However, the affinity of 14-3-3s binding to p53 was again only slightly enhanced: apparent K d 2.1 AE 0.2 mM / 1.6 AE 0.1 mM (1.3, Fig. 3A). This data shows that 1 and 3 operate in a cooperative manner and that the effect is more pronounced for the 14-3-3s-ERa PPI. To confrm that 3 was exerting its effect via covalent modi-fcation of Cys38, the same experiments were performed using a 14-3-3s Cys38Ala mutant construct (Fig. 3C and D). The apparent K d values for 14-3-3s Cys38Ala binding to both peptides alone, and in the presence of 1, were broadly comparable with the wild-type construct. However, in all cases the effect of 3 was abrogated, thus supporting a covalent mode of action. The same observation was made using 14-3-3z, an isoform that does not have an analogous cysteine residue (Fig. S5 †). This is signifcant because it suggests that covalent modifcation of Cys38 provides a strategy for selective targeting of the 14-3-3s isoform. Dose-response experiments further confrmed the specifcity of the cooperative effect between 1 and 3 for the 14-3-3s ERa PPI. Titration of 1 to fxed concentrations of 14-3-3s and the respective fluorescently labelled ERa phospho-peptide gave an EC 50 of 0.88 AE 0.1 mM (Fig. 3F). Upon the addition of 3 at an equimolar concentration to 14-3-3s, this EC 50 value was lowered relative to the control experiment (0.88 AE 0.1 mM / 0.56 AE 0.1 (Fig. 3F). In addition, the dynamic range of the assay was reduced, thus confrming the independent and cooperative stabilising effects of 3 on this PPI. In the case of the 14-3-3s-p53 PPI a reduced dynamic range was also observed in the presence of 3 (see Fig. S6 †). However, 1 is not a potent stabiliser of this PPI and there was no detectable change in EC 50 in the presence of 3. Isothermal titration calorimetry (ITC). We used ITC to further investigate how 3 influenced the binding of the phosphopeptide motifs to 14-3-3s. Unlabelled phospho-peptides were titrated to a fxed concentration of 14-3-3s in the presence of 1, 3, or a combination of both compounds. The data obtained was compared to that for the interactions in the absence of the ligands. At an equimolar concentration relative to 14-3-3s, 3 enhanced the affinity of both interactions in a manner that was consistent with the FP data. The K d for the 14-3-3s interaction with p53 was reduced only modestly (1.2 times) by addition of 3: K d 35.8 AE 5.6 mM / 29.1 AE 3.5 mM (Fig. 4A). This effect appears to be driven by a much increased enthalpic contribution to DG compared to the binary interaction that is predominantly driven by an increase in entropy, presumably due to a hydrophobic effect. In line with the FP data, 3 had a more pronounced effect on the interaction with ERa, enhancing the affinity by a factor of two: K d 2.4 AE 0.2 mM / 1.2 AE 0.1 mM (Fig. 4B). In this case the effect appears to be driven by an increased entropic contribution to DG. These differing thermodynamic profles indicate that 3 interacts differently with the respective binary complexes. In agreement with the FP data, 1 also stabilised the 14-3-3 complex with the two peptides to different extents, the stabilisation being further enhanced upon the addition of 3. In the case of p53, a molar excess of 1 relative to 14-3-3s increased affinity by 1.7, an effect that was further enhanced by 1.4 in the presence of 3 (Fig. 4A): K d 35.8 AE 5.6 mM (DMSO) / 20.7 AE 1.0 mM (1) / 14.9 AE 1.4 mM (1 + 3). For ERa, addition of an equimolar concentration of 1 relative to 14-3-3s enhanced the affinity by 3.4. In this experiment, 3 enhanced this effect to the same degree as with p53, 1.4 (Fig. 4B): K d 2.4 AE 0.2 mM (DMSO) / 0.7 AE 0.05 mM (1) / 0.5 AE 0.08 mM (1 + 3). The stabilisation offered by the addition of 3 to the 14-3-3s : ERa : 1 ternary complex appears to be driven by a favourable enthalpic contribution without additional entropic costs, a hallmark of ligands that adopt several conformations. 27 In other words, based on the ITC data, 3 may not fully order itself upon associating with the 14-3-3s : ERa : 1 ternary complex, likely binding in a number of different and interconverting poses, each contributing favourable interactions to the enthalpy of binding. Overall, the ITC data provides further evidence for cooperative stabilisation of the 14-3-3s PPIs by 1 and 3. ## Structural basis for stabilisation of the 14-3-3s-ERa PPI In order to establish a structural rationale for the cooperative stabilising effect of 3 in combination with 1, we determined the crystal structure of 14-3-3s in complex with an ERa phosphopeptide (in the presence of which the effect was most pronounced) and ligands 1 and 3. Crystals of an ERa-8mer phospho-peptide mimicking the 14-3-3 binding motif ( 588 AEGFPApTV 595 ) 9 in complex with 14-3-3s were obtained at 4 C within 3-5 days. Two 14-3-3s : ERa crystals were separately soaked at 4 C with: (i) 3 (40 mM, 2 h) and; (ii) an equimolar mixture of 1 and 3 (both 10 mM, overnight). The crystals were flash cooled in liquid nitrogen and exposed to X-rays. The 14-3-3s : ERa crystal soaked with 3 gave a 1. 19 resolution dataset and that soaked with 1 and 3 gave a 1.48 dataset (see Table S1 † for diffraction data statistics). Initial phases for the structure factor amplitudes of the 14-3-3s : ERa crystals soaked with 3 or 1 and 3 were computed by rigid-body refnement starting from the PDB ID 4JC3 and 4JDD models, respectively (these entries are crystal structures of human 14-3-3s : ERa phospho-peptide and of a human 14-3-3s : ERa : 1 ternary complex, respectively; see ESI † for further details on structure determination and refnement). 9 As expected, in both structures the 14-3-3s protein monomer accommodated one ERa phospho-peptide. In the crystal soaked with 3 alone no ordered density for the ligand was visible. In the crystal soaked with both 1 and 3, one molecule of 1 was observed to nestle in proximity to the C-terminus of ERa, in a manner consistent with that previously observed (PDB ID 4JDD). 9 Polar contacts between Asp215 of 14-3-3s and 1 lead to a shift of helix 9 (aa 210-230) in a clamping motion towards the ligand (Fig. 5A). The data also revealed at least two conformations for the side chain of Cys38 (see the 2Fo-Fc electron density map, contoured at 1s in Fig. 5B). Crucially, a positive peak in the Fo-Fc difference electron density map was found 2.0 away from the S g of the main conformer of the Cys38 side chain (see the Fo-Fc residual electron density map contoured at 2s in Fig. 5C). This peak is compatible with a disulphide bond between Cys38 and the sulphur atom of 3, confrming covalent modifcation of 14-3-3s. Disappointingly, the rest of the ligand was largely absent from the maps, indicating disorder and partial occupancy of multiple poses. To provide insight into how the predominant binding poses of 3 relate to that of 1 and ERa, the Cys38 side chain was frst modelled in three conformations, ftting positive residual electron density. Two of these (conformations A1 and A2) point into the 14-3-3s binding groove towards 1 and bear a disulphide bond to a distinct pose of 3. For both conformations, one joint occupancy was refned for the Cys38 side chain and the corresponding pose of the covalently bound ligand 3: occ A1 ¼ 0.28; occ A2 ¼ 0.27 (Fig. 5D). The third Cys38 conformation (occ B ¼ 0.45) points away from the binding groove and is free of ligand. Refnement statistics at the end of several cycles of iterative model building in Coot and refnement in autoBUSTER are gathered in Table S2. † The A1 and A2 poses of 3 point towards the glycan ring of 1 and Glu39 of 14-3-3s, the latter of which delineates the edge of the binding groove (Fig. 5D). Electrostatic interaction(s) between the positively charged amine(s) in 3 and the negatively charged acidic side chain of Glu39 may play a role in orientating 3 into this favourable position for disulphide bond formation with Cys38. This hypothesis is consistent with the mass spectrometry data that showed the requirement of a basic amine residue for covalent binding of disulphide compounds to 14-3-3s (Fig. 2). For both poses A1 and A2, the secondary amine of 3 appears to form polar contacts with an ordered water molecule that also Fig. 5 Crystal structure of human 14-3-3s protein and ERa phospho-peptide alone (PDB ID: 7NIZ) and in complex with ligands 1 and 3 (PDB ID: 7NFW). (A) Overview of 14-3-3s monomer complexed with ERa before soaking (grey) and after soaking (cyan) with 1 (yellow C atoms, red O atoms). (B and C) Electron density map 2Fo-Fc ((C); contoured as a grey mesh at 1s) and residual electron density map Fo-Fc ((D); contoured as a green mesh at 2s) around the side chain of Cys38 of 14-3-3s (cyan). (D-F) Cys38 side chain, modified with 3 (grey), modelled in conformation A1 (panels (D and E), occA1 ¼ 0.28) and A2 (panels (D and F), occA2 ¼ 0.27) forming H-bond interactions via an ordered water molecule, and direct long range interactions with 1 (yellow dashed lines). The third Cys38 conformation B (occB ¼ 0.45) points away from 14-3-3s binding groove and is not shown. coordinates to the oxygen of the allyl ether glycan substituent of 1 (Fig. 5E and F). This hydrogen bonding network is a highly plausible explanation for the biophysical data that suggests interaction between the two ligands. It is also possible that the terminal amine of 3 participates in a longer range attractive interaction with the same allyl ether substituent of 1 (e.g. ioninduced dipole interaction). ## WR-1065 and fusicoccin A kill senescent EJp53 cells in a cooperative and p53-specic manner To validate the cooperative effect of 1 and 3 on p53 in a biological context, the compounds were tested against the EJp53 bladder carcinoma cell line. 28 EJp53 has a tetracycline (TET)-off system for regulating the expression of p53. 29 When TET is removed from the culture media (EJp53), wild-type p53 is expressed and the cells enter a senescent state (Fig. S9 †). When TET is retained in the culture media (EJp53+), p53 is not expressed and the cells proliferate. Thus, specifc activity against EJp53 cells would be a strong indication of p53 activation that is consistent with stabilisation of the 14-3-3s-p53 PPI. Both senescent and proliferating cells were separately treated with 1, 3, and with combinations of the two molecules, and metabolic activity was quantifed after 72 hours using a CellTiter Glo (CTG) assay (Fig. 6A-C). In isolation, 1 and 3 both gave rise to a modest, but insignifcant, decrease in cell viability of EJp53 senescent cells compared to the proliferating EJp53+ cells (Fig. S10 †). However, when a combination of 1 (40 mM) and increasing concentrations of 3 were used, a signifcant dosedependent cytotoxic effect was observed in EJp53 cells relative to the effect on EJp53+ cells (Fig. 6A). The IC 50 of the combination treatment in EJp53 cells was calculated as 15.1 AE 2.1 mM compared to that in EJp53+ cells of 77.6 AE 38.2 mM. At the highest concentration point used (40 mM of 1 and 3) cell viability was reduced to 28% relative to a DMSO control (Fig. 6C). In contrast, the viability of the EJp53+ proliferating cells was reduced to a much lesser extent (64%, Fig. 6B). The same effect was observed when the concentration of 3 was fxed at 40 mM, and the concentration of 1 was varied (Fig. S10 †). These data are a strong indication that 1 and 3 have a cooperative cytotoxic effect against senescent EJp53 cells in a manner that is dependent on p53 expression. To further confrm that the effect of the two ligands was specifc to p53, the same experiments were performed in the EJp21 cell line. 30 EJp21 cells are p53-null but can undergo senescence by activation of p21 via a TET-off system (Fig. S9 †). Neither ligand by itself, or any combination of the two, had any cytotoxic effect on proliferating or senescent cells (Fig. S11 †). Together, these observations demonstrate the cellular relevance of the cooperative behaviour of 1 and 3. In combination, the compounds effectively reduce the viability of senescent EJp53 cells in a p53-specifc manner, which is a strong indication of an enhancement of p53 activity. Thus, a cell cycle arrest response is turned into cell death, consistent with previous reports showing that senescence can be converted into apoptosis by increasing p53 activity. 28 These data cannot explicitly attribute this to stabilisation of the 14-3-3s-p53 PPI, but the results are remarkably consistent with the effect that is to be expected based on the in vitro biophysical data. WR-1065 and fusicoccin A cooperatively attenuate estradiolinduced MCF-7 cell proliferation Next, the cooperative effect of 1 and 3 against the MCF-7 breast cancer cell line was investigated. MCF-7 cell proliferation is regulated by ERa dimerisation which leads to its enhanced interaction with chromatin and the expression of pro-survival target genes. ERa dimerisation is promoted by binding of 17bestradiol (E2), which increases MCF-7 cell proliferation. It was previously shown that this effect is attenuated by 1, which prevents ERa dimerisation by stabilising its interaction with 14-3-3s. 9 A CTG assay was used to establish if 1 and 3 exhibited a specifc effect against E2-induced MCF-7 cell growth. Cells were pre-treated separately with 1 or 3, or a combination of the two, for one hour, before the addition of E2 or DMSO as a control. Metabolic activity was quantifed after 72 hours. Treatment with 1 nM E2 resulted in a 25% increase in cell viability compared to non-treated cells. This effect was attenuated by the addition of 1 and 3 (Fig. S12 †), but again a combination of the two resulted in the most signifcant effect (Fig. 6D), although IC 50 values could not be calculated. At the highest concentration point used (40 mM of 1 and 3), the combination treatment resulted in MCF-7 cell viability being reduced to 78% compared to 1 (92%) and 3 (104%) alone (Fig. 6F). Importantly, this effect was only observed in MCF-7 cells treated with E2 (Fig. 6E and S13 †). Therefore, this is a strong indication that the cytotoxic activity of 1 and 3 is closely linked to antagonism of ERa dimerisation. As with EJp53, these data cannot explicitly prove that this results from stabilisation of the ERa-14-3-3s PPI in MCF-7 cells, but they do provide compelling evidence in support of that hypothesis and further exemplifcation of cooperativity between 1 and 3. ## Conclusions Here, we report that covalent modifcation of 14-3-3s Cys38 by the disulphide of WR-1065 (3) leads to isoform-specifc stabilisation of two 14-3-3s PPIs in vitro. Achieving isoform specifcity is a major challenge that must be overcome for the maturation of 14-3-3 drug discovery programmes. Our fndings demonstrate a novel strategy that can be further exploited in the development of 14-3-3s specifc stabilisers, and also for harnessing the inherent potency of covalent modulators. The data also indicate that in vitro 3 preferentially stabilises the 14-3-3s interaction with ERa, which binds 14-3-3s via a C-terminal 'mode 3' phosphorylated binding motif, compared to p53 which adopts a unique turn conformation. In the absence of structures of ternary complexes containing 3 alone, the structural basis for this is not yet clear. However, the data does highlight the potential for achieving PPI selectivity in addition to isoform selectivity. The cooperative behaviour of 1 and 3 reveals a fascinating stabilisation mechanism that is distinct from the only other example of cooperative dual ligand stabilisation of 14-3-3 PPIs. 16 In that report, a supramolecular arginine mimetic was shown to act cooperatively with 1 to stabilise the 14-3-3z interaction with ERa. 16 Although highly signifcant, that example was limited in terms of the physiochemical properties of the supramolecular ligand and scope for achieving isoform specifcity. Here, the cellular relevance of the cooperative effect between 1 and 3 has been demonstrated in two cancer cell lines. Furthermore, our 14-3-3s : ERa : 1 : 3 crystal structure provides a platform for structure-based design of more potent, drug-like and synthetically tractable analogues of 3. It also highlights the potential impact of covalent hybrids of 3 and 1, or other known 14-3-3 stabilisers that occupy a similar pocket. We envisage that the covalent and cooperative stabilisation mechanism revealed by this study will underpin the development of selective and efficacious 14-3-3s stabilisers that can be exploited in a range of therapeutic areas such as oncology and neurodegeneration. ## Experimental Full experimental details can be found in the ESI. †

chemsum

{"title": "Cooperative stabilisation of 14-3-3\u03c3 protein\u2013protein interactions <i>via</i> covalent protein modification", "journal": "Royal Society of Chemistry (RSC)"}

kinetics_of_the_ancestral_carbon_metabolism_pathways_in_deep-branching_bacteria_and_archaea

## Abstract: The origin of life is believed to be chemoautotrophic, deriving all biomass components from carbon dioxide, and all energy from inorganic redox couples in the environment. The reductive tricarboxylic acid cycle (rTCA) and the Wood-Ljungdahl pathway (WL) have been recognized as the most ancient carbon fixation pathways. The rTCA of the chemolithotrophic Thermosulfidibacter takaii, which was recently demonstrated to take place via an unexpected reverse reaction of citrate synthase, was reproduced using a kinetic network model, and a competition between reductive and oxidative fluxes on rTCA due to an acetyl coenzyme A (ACOA) influx upon acetate uptake was revealed. Avoiding ACOA direct influx into rTCA from WL is, therefore, raised as a kinetically necessary condition to maintain a complete rTCA. This hypothesis was confirmed for deep-branching bacteria and archaea, and explains the kinetic factors governing elementary processes in carbon metabolism evolution from the last universal common ancestor. T he concept of a last universal common ancestor (LUCA) has been attracting much interest among researchers on the early evolution and origin of life. Based on an extensive investigation of protein-coding genes in prokaryotic genomes, LUCA has been inferred to be an anaerobic chemoautotroph that uses the Wood-Ljungdahl (WL) pathway and exists in a hydrothermal setting 1 . Therefore, most of the organic carbon on early Earth was obtained by chemoautotrophic carbon fixation, thus sustaining life. Currently, there are seven known different autotrophic pathways responsible for carbon fixation, including the newly demonstrated reductive glycine (rGly) pathway 2 . These pathways are recognized to have partial overlaps, but the reasons for the redundancy and the forces of selection in their evolution remain unclear. The WL pathway, also called the reductive acetyl-CoA (coenzyme A) pathway, and the reductive tricarboxylic acid (rTCA) cycle are among the most ancient pathways responsible for the biosynthesis of the universal precursors of anabolism . Even if we focus only on these ancient carbon metabolism pathways, diverse observations on the phylogenetic distribution of the network topology have not yet been given a universal and phylum-compatible interpretation. The rTCA cycle is a reversal of the oxidative tricarboxylic acid (oTCA) cycle, which provides energy, reducing agents, and precursors for certain amino acids. However, citrate synthesis from acetyl-CoA (ACOA) and oxaloacetate, catalyzed by citrate synthase (CS), is regarded as one of the irreversible reactions in the oTCA cycle 7,8 . Therefore, in organisms using the rTCA cycle, CS has been interpreted to be substituted either by an adenosine triphosphate (ATP)-dependent citrate lyase or other homologous enzymes that catalyze the same reaction in two steps . In contrast, however, two research groups demonstrated that CS activity drove the rTCA cycle in the thermophilic bacteria Thermosulfidibacter takaii and Desulfurella acetivorans grown chemolithoautotrophically . In the present study, kinetic simulations of the autotrophic rTCA cycle for these bacteria are presented and the mechanism by which the CS reaction is reversed, which was thought to function only in the oxidative direction, is revealed in dynamical systems. In addition, bifurcation of the TCA cycle into reductive and oxidative reactions that has been demonstrated in the presence of succinate and acetate 13,14 is also simulated. Interestingly, it was observed in dynamical systems that a conflict between the reductive and oxidative fluxes in the TCA cycle was caused by an influx of ACOA upon uptake of acetate. This indicates that the coexistence of a WL pathway and a rTCA cycle, which yields ACOA influx into the rTCA cycle, results in a partial reduction of fluxes in the TCA cycle. Consequently, the less active enzymes of the rTCA cycle due to this coexistence would be lost to evolution or enzymes lacked in the rTCA cycle would be difficult to be newly incorporated into the less active part of that. In the pioneering work by Braakman and Smith 5 , a single connected redundant network consisting of a complete WL pathway and a full rTCA cycle was proposed as a more robust and plausible topology that LUCA or (proto) biological systems leading up to the LUCA possibly possess. However, our kinetic results do not support the coexistence of these two pathways in one organism, unless other predominant selective pressure that favors their combination exists as well. Remarkably, it was confirmed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) carbon metabolism database 15 that our kinetic hypothesis, that a complete rTCA cycle does not coexist with a WL pathway, was held for deeply branching archaea and bacteria . Based on the kinetic hypothesis and the fluxes simulated for these organisms, a fundamental model of carbon metabolism evolution for ancestral bacteria and archaea starting from LUCA is presented. ## Results Chemolithoautotrophic TCA fluxes are simulated with a reversible reaction of CS. In the genomes of T. takaii and D. acetivorans, typical key enzymes involved in the rTCA cycle of autotrophs were identified, namely, the ferredoxin-dependent enzymes 2-oxoglutarate:ferredoxin oxidoreductase and pyruvate:ferredoxin oxidoreductase. On the other hand, genes for ATP citrate lyase and its two-step variant citryl-CoA synthetase/ citryl-CoA lyase, which are regarded to be necessary for organisms with autotrophic rTCA cycle, were missing 13,14 : these organisms instead possessed genes of CS that were thought to be active only in oTCA cycle. In the present study, the kinetic reaction models for enzymes identified for carbon metabolism in T. takaii were developed (Eqs. B1-B18 in the Supplementary Information (SI)), and the kinetic network model in which these enzymatic reactions were incorporated was presented. To investigate the direction of carbon metabolic fluxes on the network, the steady-state fluxes were determined by the kinetic network model (Eqs. 1-16 in the SI) with the fixed concentrations of chemical species listed in the SI Table A1. The five universal precursors of anabolism, ACOA, pyruvate (PYR), phosphoenolpyruvate (PEP), oxaloacetate (OAA), and 2-oxoglutarate (AKG: α-ketoglutaric acid), were assumed to be consumed by biomass synthesis in the kinetic network model (Eq. B18 in the SI). The simulation utilizing the kinetic network model demonstrated the feasibility of the rTCA cycle due to a reversal of the CS reaction resulting in chemolithoautotrophic growth (Table S1a in the SI). Furthermore, the directions of obtained fluxes (Fig. 1a) were consistent with those experimentally assigned by Nunoura et al. for T. takaii grown chemolithoautotrophically, except for the directions between malate (MAL) and PYR and between OAA and PEP 13 . In addition, it was confirmed that the kinetic network simulations involving the gene knockouts of malic enzyme between MAL and PYR and/or of phosphoenolpyruvate carboxykinase between OAA and PEP reproduced a complete rTCA flux (Table S1b in the SI). These observations indicate that the directions of these fluxes are not essential to drive the rTCA cycle. Here the simulation with gene knockouts of enzyme means that it performed without the enzyme. Furthermore, the metabolite concentrations calculated in the kinetic network model were quantitatively consistent with those experimentally determined for D. acetivorans grown chemolithoautotrophically 14 (Table S2 in the SI). It was also confirmed that the autotrophic growth simulated here was robust across an extensively varied ratio of the concentrations of reduced ferredoxin (Fdx red ) and oxidized ferredoxin (Fdx ox ) (Fig. S2 in the SI). Recently, it was experimentally observed that high partial pressure from CO 2 drove autotrophic rTCA cycle with the reversal of the CS reaction. Our kinetic network model reproduced the CO 2 dependence of autotrophic growth rate observed for D. acetivorans 19 (Fig. S3 in the SI). The apparent Gibbs reaction energy (Δ r G tot i ; defined in Eq. A9) for the CS reaction was calculated to be −53.4 kJ/mol, and the total Δ r G tot i of the oTCA cycle was −108.8 kJ/mol (Table S3 in the SI). The reversal of the oTCA cycle with high exergonic Δ r G tot i requires an abundant supply of reducing agents. Reducing reactions in anaerobic organisms are more efficiently driven by reduced ferredoxins than reduced nicotinamide adenine dinucleotide (NADH) or reduced nicotinamide adenine dinucleotide phosphate (NADPH) in terms of Δ r G tot i (Table S3 in the SI). In addition, the ferredoxin-dependent enzymes working in the rTCA cycle involve two reduced/oxidized ferredoxins in the reductive reactions (Table S3 in the SI); thus, the concentration ratio between reduced and oxidized ferredoxins needed to overcome the energetically unfavorable reaction is lower than that with NADH/NADPH. It is generally recognized that sufficient depletion of both ACOA and OAA is necessary to overcome the huge energetic barrier of the CS reaction and reverse it. This suggests that organisms capable of doing so must efficiently convert ACOA into PYR or cellular biomolecules and OAA into MAL 12 . To examine these hypotheses, we determined how the gene knockout of pyruvate synthase (PS), which converts ACOA to PYR, and pyruvate carboxylase (PC), which converts PYR to OAA, affects the rTCA cycle. It is assumed that the model of D. acetivorans gains a PYR influx into the rTCA cycle via serine (SER) metabolism, whereas that of T. takaii does not because of its lack of the necessary enzyme (Fig. S5e, f in the SI). As summarized in SI Table S4, both models with the gene knockout of PC lost the rTCA flux. However, the model of D. acetivorans with the gene knockout of PS maintained/reduced the rTCA flux depending on the PYR influx, whereas that of T. takaii significantly reduced it. These observations indicate that the depression of product concentrations in the reversed CS reaction is not sufficient to maintain the rTCA cycle and a reinvestment of the outflux equivalent of ACOA into the rTCA, which is caused by the OAA influx via PC, is additionally required. This kinetic requirement is schematically illustrated in Fig. S1a. The necessity of PC as a kinetic condition for maintaining rTCA flux revealed here plays a crucial role in interpreting the phenotype evolution of carbon metabolism for deep-branching organisms, as discussed later. However, it is not obvious whether the differences in the rTCA fluxes between T. takaii and D. acetivorans are actually due to the gene knockout of PS because the effect of PYR influx via SER metabolism on D. acetivorans is uncertain. Nonetheless, the investigation of the role of PC and PS by utilizing the T. takaiiand D. acetivorans-type models was useful since it shed light on the importance of PC in the rTCA cycle. The rTCA fluxes are bifurcated and partially impaired in the presence of succinate or acetate. The kinetic network model with the autotrophic growth condition, as used in Fig. 1a, reproduced the directions of experimentally determined carbon fluxes for T. takaii grown chemolithomixotrophically in the presence of succinate (SUC) (Fig. 1b) and acetate (Fig. 1c) 13 . Upon an influx of SUC, the bifurcating oxidative flux from SUC toward MAL and the reductive flux for the remaining part of the TCA cycle were observed (Fig. 1b). The directions of the obtained fluxes were consistent with the experimental observations for T. takaii grown chemolithomixotrophically with SUC, except for the directions between MAL and OAA and between OAA and PYR 13 . Next, upon an influx of ACOA, the bifurcating oxidative flux from both OAA and ACOA toward AKG and the reductive flux for the remaining part of the TCA cycle were simultaneously observed (Fig. 1c). The directions of the obtained fluxes were consistent with the experimental observations for T. takaii grown chemolithomixotrophically with acetate, which is converted into ACOA, except for the directions between SCOA and AKG and between OAA and PEP 13 . The reductive flux was increased by the ACOA influx because of an increase in the flux through the ACOA-PYR-OAA pathway, whereas the absolute value of oxidative flux from ACOA to AKG likewise caused by the ACOA influx was 5 × 10 7 times smaller than that of the reductive flux, upon comparison with the reductive flux (Fig. 1c). This is because the oxidative flux by the ACOA influx was weakened by the reductive flux. Furthermore, a smaller ACOA influx did not result in a small oxidative flux but instead impaired the reductive flux from AKG toward both OAA and ACOA (Fig. 1d). This is because the smaller oxidative flux caused by the smaller ACOA influx was slightly overcome by the reductive flux. The absolute value of the impaired reductive flux was ~1 × 10 5 times smaller than that of the reductive flux on the left side of rTCA cycle due to the competition with the oxidative flux component. This result was consistent with the experimental observation for D. acetivorans grown chemolithomixotrophically in the presence of acetate, where the majority of acetate uptake was mainly transformed into PYR and incorporated into the rTCA cycle via the abovementioned pathway from PYR to OAA 14 (Fig. 1d, also see Fig. S1a in the SI). Such a conflict between the rTCA cycle and an ACOA influx shown in Fig. 1d is held even if the ACOA influx is much smaller than the normal reductive flux from OAA to AKG (see Fig. S4 in the SI). Taken together, the ACOA influx causes a partial competition between the reductive and oxidative flux on the rTCA cycle and results in the impairment of reductive flux between AKG and OAA (ACOA). This observation further raises one kinetic hypothesis: the coexistence of either the WL or another pathway yielding ACOA influx with the rTCA cycle is kinetically inefficient because of the competition between the rTCA flux and the oxidative flux caused by the ACOA influx. Therefore, unless the ACOA influx disappears, the gene for enzymes working at the less active reaction on the rTCA cycle is lost or that of enzymes lacked on the less active part is not newly gained during evolution of LUCA. In short, the kinetic hypothesis we propose in this study is that a complete rTCA cycle is never observed in organisms because of its kinetic instability from the competition between the reductive and oxidative flux, as long as a carbon fixation pathway including the WL pathway yields ACOA influx into the rTCA cycle. The validity of this kinetic hypothesis is examined in the next subsection for deep-branching bacteria and archaea. TCA flux for deep-branching bacteria depends on the presence/absence of WL and rGly pathways. To assess the validity of the kinetic hypothesis, the fluxes of carbon metabolism for deep-branching bacteria were examined using the kinetic network model simulations with enzymes identified by the KEGG genome database 15 . The results for deep-branching bacteria grown autotrophically, the Firmicutes, Cyanobacteria, Aquificae, and Proteobacteria 16,18 , are shown in the SI Fig. S5. From the fluxes observed under the autotrophic growth condition (listed by Table A1 in the SI), four representative flux patterns of carbon metabolism for the deep-branching bacteria were extracted (Fig. 2). The light dashed lines without arrow in Fig. 2 and the SI Fig. S5 indicate zero flux because of either totally or partially absence of enzymes. Cyanobacteria was excluded here because it possessed an oTCA cycle 20 and was simulated with a lower reduced to oxidized agent ratio but was later incorporated in subsequent discussions. These representative flux patterns are classified according to whether the WL pathway coexists with the TCA cycle and yields an ACOA influx into the rTCA cycle. In the type-B1 pathway, the WL pathway produces ACOA influx, resulting in MAL influx through PYR into the partial rTCA pathway. This is applicable for H 2 /CO 2 -utilizing acetogens, e.g., Sporomusa termitida, which grows both autotrophically and heterotrophically 21 . Next, as an alternative pathway where a WL pathway coexists with CS is the type-B2 pathway, where the ACOA influx yielded by a WL pathway causes the partial oTCA pathway to synthesize AKG. However, because of the absence of several enzymes acting in the rTCA pathway, the competition between oxidative and reductive fluxes does not occur. This is found for acetogenic anaerobic bacteria, e.g., Moorella thermoacetica, which also grows autotrophically and heterotrophically 22 . In the type-B3 pathway, neither the incomplete WL pathway nor the incomplete rGly pathway produces an ACOA influx; hence, there is no competition between reductive and oxidative flux, and the necessary kinetic condition with respect to PC is also satisfied, thus a complete rTCA flux is produced. This is observed for sulfur-reducing bacteria, T. takaii 13 and D. acetivorans 14 , which grow chemolithoautotrophically and heterotrophically. In the type-B4 pathway, if the rGly pathway produces ACOA influx, the kinetic hypothesis does not allow a complete rTCA cycle, yielding OAA influx through PEP into the partial rTCA pathway. This is observed for anaerobic anoxygenic photoheterotrophs, e.g., Heliobacterium modesticaldum 23,24 . Interestingly, it was expected that, according to the KEGG genome database 15 , H. modesticaldum possessed an rGly pathway similar to the one recently identified for Desulfovibrio desulfuricans grown chemolithoautotrophically 2 . The absence of complete rTCA cycle in H. modesticaldum possessing no WL pathway but an rGly pathway supports the extended applicability of kinetic hypothesis. Taken together, the proposed kinetic hypothesis is basically satisfied because the network patterns observed for deep-branching bacteria do not involve impaired rTCA fluxes. Partial or impaired rTCA and full oTCA fluxes are observed for deep-branching archaea. Next, the fluxes of carbon metabolism for deep-branching archaea were examined using the kinetic network model simulations with enzymes identified by the Fig. 2 Typical flux patterns of carbon metabolism for deep-branching bacteria obtained from the kinetic network model. B1: a partial rTCA pathway coexisting with a WL pathway that yields an ACOA influx; B2: a partial oTCA pathway coexisting with a WL pathway that yields an ACOA influx; B3: a full rTCA cycle and either an incomplete WL or an incomplete rGly pathway; B4: a partial rTCA pathway coexisting with an rGly pathway that yields an ACOA influx. The light dashed lines without arrow indicate zero flux because of either totally or partially absent enzymes. The "incomplete WL or rGly pathway" also includes cases where enzymes acting in the WL or rGly pathway partially exist, indicating no ACOA influx from outside. The dashed arrows mean that there are cases where the corresponding flux is present or absent depending on species. The autotrophic growth condition with high Fdx red /Fdx ox ratio (listed by Table A1 in the SI) is used in the kinetic network simulations for bacteria B1-B4 grown autotrophically. KEGG genome database 15 . The results for the deep-branching archaea grown autotrophically, namely, Euryarchaeota, Candidatus Bathyarchaeota, and Crenarchaeota , are shown in the SI Fig. S6, except for Fig. S6f. Based on the fluxes observed under the autotrophic growth condition (listed by Table A1 in the SI), three representative flux patterns of carbon metabolism for deepbranching archaea were identified (Fig. 3). The light dashed lines without arrow in Fig. 3 and SI Fig. S6 indicate zero flux because of either total or partial absence of enzymes. In the type-A1 pathway, the WL pathway produces an ACOA influx, resulting in an OAA influx into the partial rTCA pathway. This is found for autotrophic methanogens such as Methanococcus maripaludis . In the type-A2 pathway, the absence of WL pathway does not prevent a complete rTCA cycle under autotrophic growth conditions. Nevertheless, the rTCA cycle is kinetically inhibited because of the absence of PC in this pathway. This absence led to the loss of the OAA influx from PYR into the rTCA cycle, thus causing a decrease in the concentration of CIT, which is the product of rTCA cycle, and resulted in the impairment of the reversed CS reaction. Consequently, the right half of the rTCA flux was significantly reduced (also discussed in the SI Table S4). In fact, as shown in the type-A2 pathway, instead of the rTCA cycle, the complete oTCA cycle under a heterotrophic growth condition was reproduced even in the absence of PC by an uptake of PYR and lowering the ratio of reduced agent (Fdx red ) to oxidized agent (Fdx ox ). This is found for the heterotrophic anaerobic archaeon Vulcanisaeta distributa 28 . As shown in the type-A3 pathway, the partially impaired rTCA flux was observed under the autotrophic growth condition even in the absence of the WL pathway. In addition to the absence of PC, type-A3 Crenarchaeota Thermoproteus tenax and Acidianus hospitalis possess the dicarboxylate/4-hydroxybutyrate (DC/4HB) 29 and 3hydroxypropionate/4-hydroxybutyrate (3HP/4HB) pathways 30 , respectively. These alternative carbon fixation pathways yield an ACOA influx under chemolithoautotrophic growth conditions and thus impairs the rTCA flux on the right-hand side (Fig. 3 (A3)) in the same manner as the WL pathway. The inefficiency on the partially impaired rTCA cycle should be acceptable if the DC/4HB and 3HP/ 4HB pathways more efficiently work for carbon fixation. Furthermore, these Crenarchaeota grow heterotrophically 29,30 . The oTCA flux was simulated under heterotrophic growth conditions with lowering the Fdx red /Fdx ox ratio, though not shown in Fig. 3 (A3). Therefore, it is conceivable that these Crenarchaeota possess and maintain the enzymes for the less active reductive reactions on the A1 in the SI) was applied for A1 and A3, while a heterotrophic growth condition with a lower Fdx red /Fdx ox ratio was used for A2 with PYR uptake. In A3, the DC/4HB and 3HP/4HB pathways were modeled as a combination of additional ACOA influx and the conversion of succinyl-CoA (SCOA) to ACOA. partially impaired rTCA cycle to utilize the oTCA cycle under heterotrophic growth conditions. Taken together, the kinetic hypothesis is satisfied on the carbon metabolic pathway of the deep-branching archaea, parallel to that of bacteria. ## Discussion Elementary evolution processes of carbon metabolism are derived from the kinetic hypothesis and the simulated typical flux patterns. In the pioneering work by Braakman and Smith 5 , a single connected redundant network consisting of a complete WL pathway and a full rTCA cycle was proposed as a more robust and plausible topology that LUCA or (proto)biological systems leading up to the LUCA should possess on the basis of phylometabolic analysis. In contrast, the results obtained from our kinetic network model do not support the coexistence of those pathway in one organism, unless a more dominant and selective pressure that favors the combination of those pathways exists in parallel. The key difference between their work and ours is whether the kinetic factors are taken into consideration. If it is assumed that, as widely accepted, LUCA is an anaerobic hydrothermal chemoautotroph possessing at least a WL pathway 1 , the confirmed kinetic hypothesis raises a model of evolution processes on carbon metabolism for deep-branching archaea and bacteria (Fig. 4). In this model, the simulated typical carbon metabolism pathways A1-A3 (Fig. 3) B1-B4 (Fig. 2), and type-B5 were incorporated. The type-B5 pathway is observed for Cyanobacteria, e.g., Synechocystis sp., which possesses an oTCA cycle and a Calvin-Benson-Bassham (CBB) cycle (Fig. S5d in the SI) 20 . As shown in Fig. 4, the evolution processes observed for deepbranching archaea and bacteria can be classified into two types: type-I evolution that occurs with a coexisting complete WL pathway and type-II evolution that occurs in the absence of a complete WL pathway. The incomplete WL (iWL) pathway often lacks, at the very least, CO dehydrogenase/acetyl-CoA synthase, which is the final step of ACOA synthesis and is known as one of the most oxygen-sensitive enzymes in the biosphere 5,31 . Braakman and Smith proposed that oxygen toxicity to the enzymes act as a selection force on the iWL pathway 5 . First, we discuss how type-I evolution occurs in the presence of a WL pathway from the viewpoint of the kinetic hypothesis. The phenotype that possesses a partial rTCA pathway coexisting with a WL pathway is commonly observed for both archaea and bacteria (A1 and B1) and is thus predicted to be a plausible candidate for the carbon metabolism phenotype of LUCA. Alternatively, if LUCA does not possess a partial rTCA pathway, such ancestral chemoautotrophs would gain enzymes necessary for a partial rTCA pathway that coexists with the WL pathway during evolution toward A1 or B1. On the other hand, in cases where ancestral chemoautotrophs gained CS during the early stages of type-I evolution, they would gain enzymes necessary for a partial oTCA pathway to synthesize AKG, as seen in B2. This corresponds to acetogenic anaerobic bacteria growing autotrophically and heterotrophically (Fig. S5d in the SI). Evolutionary divergence of ancient carbon metabolism not coexisting with WL pathway. Here it should be noted that the kinetic hypothesis does not exclude type-II evolution from LUCA, which is a direct evolution from an ancestral chemoautotroph coexisting with an incomplete WL pathway. Therefore, type-II evolution occurs following type-I or could take place directly from LUCA as well. As seen in Fig. 4, type-II evolution of bacteria depended on whether they obtained PC, which was necessary for maintaining a full rTCA cycle, if no ACOA influx was supplied by the other carbon fixation pathway. If ancestral bacteria could gain PC or keep that was previously gained, they would gain enzymes Fig. 4 Evolution process model from LUCA on carbon metabolism pathway for deep-branching archaea and bacteria. In this model, it is assumed that LUCA is an anaerobic hydrothermal chemoautotroph that possesses a WL pathway 1 . Evolutionary processes observed for deep-branching archaea and bacteria are classified into type-I and type-II evolution: the former and latter occur in the presence and absence of a WL pathway, respectively. During type-I evolution, the WL pathway yields direct ACOA influx into the rTCA cycle, while it does not do so in type-II. Type-II evolution occurs following type-I evolution or could take place directly from LUCA as well. The phenotypes B1-B4 and A1-A3 are the same as shown in Figs. 2 and 3, respectively. necessary for a full rTCA cycle to increase the rTCA flux (phenotype B3); otherwise, they would undergo alternative evolution. On one hand, as seen in B5, because the ancestor of Cyanobacteria without PC acquired a CBB cycle that enabled of photoautotrophic growth, utilizing neither WL nor the rTCA carbon fixation pathway, they would gain enzymes for an oTCA cycle to produce energy, reducing agents, and metabolites so that they grew photomixotrophically as well (Fig. S5d in the SI) 32 . On the other hand, as seen in B4, some ancestral bacteria acquired an rGly pathway 2 , which worked as an alternative carbon fixation pathway and provided an ACOA influx, and they either gained enzymes necessary for partial rTCA or kept the ones already possessed so that they synthesized AKG. H. modesticaldum 24 with the type-B4 pathway possesses enzymes for the remaining part of the rTCA cycle, except for CS and PC. Thus, H. modesticaldum might have evolved from type-B3 pathway and, during this process, possibly lost PC and a part of the less active enzymes of the impaired rTCA cycle. Turning to the phenotype of deep-branching archaea, those with type-A3 have rTCA flux partially decreased by both the ACOA influx from a DC/4HB or 3HP/4HB pathway and the lack of PC. Therefore, from a kinetic point of view, it would not be expected that they directly gained the enzymes with significantly low activity during evolution. Type-II evolution of ancestral archaea is thus interpreted according to the following plausible scenario: as seen in A2, ancestral archaea without PC would first gain enzymes necessary for a complete oTCA cycle under heterotrophic growing conditions so that they evolved to be heterotrophic anaerobic archaea. Subsequently, the heterotrophic anaerobic archaea would acquire enzymes necessary for a DC/4HB or 3HP/4HB carbon fixation pathway and gain the ability to grow not only heterotrophically but also chemolithoautotrophically, as seen in A3. Notably, V. distributa 28 with type-A2 partly possesses enzymes necessary for the DC/4HB carbon fixation pathway. Thus, it might have evolved into a heterotroph from type-A3 and, during the process, possibly lost some of the less active enzymes for the DC/ 4HP pathway under heterotrophically growing conditions. In summary, the model of elementary evolutionary processes on carbon metabolism presented here satisfies the following two kinetic conditions that are necessary for maintaining a full rTCA cycle: avoiding a direct ACOA influx into the rTCA cycle and the reinvestment of the outflux equivalent of ACOA into the rTCA via the OAA influx. These kinetic hypotheses and necessary conditions are expected to play crucial roles in revealing the carbon metabolic network that LUCA possesses and understanding the evolution processes from LUCA toward deepbranching archaea and bacteria. However, the selective forces such as energy optimization, oxygen toxicity, and the alkalinity in environments 5 , and influences of lateral gene transfer 33 are not taken into consideration in the evolution processes presented in Fig. 4. Investigation of the relation among the presented kinetic insights, the abovementioned selective forces, and evolutionary convergences are the scope of future research. ## Methods Kinetic network model. The enzymatic reactions identified for the rTCA cycle of T. takaii and the reactions related to the TCA cycle including anaplerotic ones (Eqs. B1-B17 in the SI) were incorporated into the kinetic network model for carbon metabolism described by the ordinary differential Eqs. 1-16 listed in the SI (also see Fig. A1 in the SI). The information on the enzyme-coding genes were obtained by browsing on specific organism pages in the KEGG database 15 . In addition, biomass synthesis reactions for cell growth (Eq. B18 in the SI) were also incorporated into the kinetic network model (see Fig. A1 in the SI). Thus, the five universal precursors of anabolism, ACOA, PYR, PEP, OAA, and AKG, are consumed by the biomass synthesis. Kinetic models for enzymatic reactions on the rTCA cycle were, in part, taken from the literatures including the modeling paper of Beard and co-workers 34 , and those that were not available from literatures were developed by ourselves on the basis of biochemical experimental data. The details of model equations and parameters used in the simulations are given below. The key point that should be especially stressed on the kinetic network modeling was the apparent equilibrium constants upon total concentrations of substrates and products for each enzymatic reaction, which were precisely determined according to the method and database developed by Beard and co-workers 35 . The calculation of the apparent equilibrium constant and the apparent reaction Gibbs free energy is explained in "General aspects on kinetic modeling of enzymatic reactions" of the SI. Simulations. The ordinary differential equations (Eqs. 1-16 in the SI) for the kinetic network model comprised of 22 reaction equations and 29 chemical species were solved using COPASI biochemical system simulator (ver. 4.28) 36 . The steadystate concentrations and fluxes for all the organisms of which results were discussed in the main text and SI were determined by the kinetic network model simulation based on Eqs. 1-16 in the SI. The concentrations of chemical species used as the model parameters in the kinetic network model are listed by Table A1 in the SI. Several model parameters in Table A1 were estimated so that the kinetic network model simulation reproduced the concentration of metabolites experimentally determined for D. acetivorans grown autotrophically (Table S2 in the SI) 14 .

chemsum

{"title": "Kinetics of the ancestral carbon metabolism pathways in deep-branching bacteria and archaea", "journal": "Nature Communications Chemistry"}

simple_tyrosine_derivatives_act_as_low_molecular_weight_organogelators

## Abstract: the gelation of L-tyr(tBu)-oH in tetrahydrofuran (tHF) was discovered serendipitously. It was noted that this tremendously low molecular weight (LMW) compound has the ability to gel a wide variety of organic solvents (e.g., N,N-Dimetylformamide (DMF), tHF, butanol, toluene), even in very low concentrations (i.e., 0.1 wt/v% in DMF). Addition of bases such as NaOH and piperidine enhanced the gel property. By changing the side-chain protecting group to tert-butyldimethylsilyl (tBDMs), a fluoride ion-responsive organogel was also acquired. This new organogelator responded fluoride ion concentration as low as 0.2 ppm. Characterization of microstructures and gel behaviours were studied by powder X-Ray diffraction spectroscopy (XRD), transmission electron microscopy (TEM), rheological measurements and molecular dynamics (MD) simulations. experimental observations and theoretical simulations consistently show a fibre-like structure of the gel, in which the organogelator molecules are held together via a dense network of hydrogen bonds, and via van der Waals interactions between hydrophobic groups.Low molecular weight (LMW) organogels have attracted significant research attention over the past two decades due to their numerous potential applications, including drug delivery, oil-spill recovery, smart electronics, and stimuli-responsive materials 1-5 . Characteristics such as the intermolecular hydrogen bonding, π-π stacking of aromatic units, hydrophobic effect, and van der Waals forces lead to the self-assembly of low molecular weight organogelators (LMWO) in fibrous structures, tapes, sheets, etc 1,6 . The entanglement of these secondary structures eventually results in the immobilization of the solvent, which can be referred to as a gel 7 . Amino acids are versatile compounds for self-assembled structures and, therefore, have found various applications as LMWO 8-14 . However, amino acids usually suffer a lack of wide range solvent applicability or require several steps to synthesize. There are also many examples of multi-component LMW gels in the literature 15 . In some cases, two components are necessary to form a gel, whereas in other cases two different components, each having gelation ability on its own, are mixed in order to obtain a gel that results from the assembly of the two components in an alternating fashion. Additionally, a non-gelling additive can be used to increase the lifetime stability of a gel and/ or to enhance its mechanical properties.The design of new fluoride-sensitive molecules is currently the focus of research efforts on the part of many groups, notably due to the implications of such molecules for dental care and for the treatment of osteoporosis [16][17][18] . However, overexposure to fluoride ions can cause kidney problems and fluorosis. The World Health Organisation has declared the appropriate amount of fluoride ions in drinking water to be 0.5-1.0 mg/L 19 . In the literature, stimuli-responsive organogels have been used to selectively detect fluoride ions by forming charge-transfer complexes [20][21][22][23] . Further, many chemosensor systems have been investigated with regards to their ability to sense fluoride ions using the fluorescence and colorimetric properties of materials 16,17 . Some of these studies depend on the cleavage of the silicon-oxygen bond of silyl ethers to promote fluoride ions [24][25][26][27][28] . In the present study, we demonstrated the organogelator properties of L-Tyr(tBu)-OH, a tremendously low molecular weight organogelators. This organogelator has a molecular weight of only 237 amu, as well as an outstanding gelation ability in relation to a wide range of organic solvents down to a concentration of 0.1 wt/v%. Then, with regards to potential chemosensor applications, we synthesized L-Tyr(TBDMS)-OH and investigated its gelation properties under various conditions. L-Tyr(TBDMS)-OH is sensitive to the presence of fluoride, since the latter can trigger the cleavage of the Si-O bond to form L-Tyr-OH. ## Results and Discussion spontaneous Gelation of L-tyr(tBu)-OH (2). L-Tyr(tBu)-OH is an inexpensive commercially available compound and is also synthetically straightforward to produce quantitatively in one step from Fmoc-L-Tyr(tBu)-OH 29 . Gelation occurs spontaneously during the following reaction (Fig. 1). The cleavage of the fluorenylmethyloxycarbonyl (Fmoc) group with piperidine in THF resulted in gelation within 15 minutes during the reaction. After the removal of the by-product, gelation was successfully achieved under the same conditions using pure L-Tyr(tBu)-OH which indicates that the by-product does not play a role in the gelation process. This reaction was initially performed aiming to synthesize unprotected tyrosine to be used in another study and eventually led us to the serendipitous discovery of this new organogelator, which triggered further investigations as reported hereafter. Same gelation properties were observed for the other enantiomer, D-Tyr(tBu)-OH, as expected. However, no gel formation was observed in the case of racemic Tyr(tBu)-OH at minimum gelation concentration. In addition, we observed that the tert-butyl moiety plays a significant role in the gelation process, since gel formation does not occur for either L-Tyr-OH, L-Phe-OH and L-Tyr(Me)-OH. Additive and solvent screening for gelation. In order to investigate the gelation properties of L-Tyr(tBu)-OH, different combinations of additives and solvents were considered using the vial inversion method (Tables 1 and 2). Table 1 summarizes the effects of the different additives (bases and alcohols) on gelation in THF. Gels were prepared in 1.0 mL of THF. After dissolution of L-Tyr(tBu)-OH in solvent with the help of ultrasonic bath at 40 °C, 10.0 μL of additive was added. The solutions were placed again into ultrasonic bath for 4-10 minutes. The formation of the gels was determined using inversion test. L-Tyr(tBu)-OH forms a gel in THF with a minimum concentration of 1.1 wt/v% without the use of any additive. Surprisingly, the addition of piperidine lowers the minimum gel concentration to 0.45 wt/v%. Gelation also occurs in the presence of other bases, such as diisopropylamine, diethylamine and imidazole, albeit with weaker gel appearances. However, when triethylamine and DBU is used as additive, a suspension and a clear solution is formed, respectively. This result suggests that the additive should also have donor-acceptor hydrogen bond ability to trigger gel formation. To further test this hypothesis, short-alkyl-chain-containing alcohols were also considered but were found to play no favourable role in the gelation process as an additive. On the other hand, 2-ethylhexanol does show significant effect on gelation, which indicates that van der Waals interactions between additive and organogelator also decrease the minimum gelation concentration. Beside these organic bases and alcohols, the addition of 1 eq. of NaOH resulted in a minimum gelation concentration as low as 0.25 wt/v% in THF. These results show that additives are not necessarily involved in the hydrogen bond network, but that the amine-type additives are also acting as a base, which deprotonates the carboxylic acid, allowing the resulting www.nature.com/scientificreports www.nature.com/scientificreports/ carboxylate group to better participate in the hydrogen bond network. This is also consistent with the higher minimum gelation concentrations observed when short-alkyl-chain-containing alcohols were used as additives, since alcohols are weakly acidic. The gelation of L-Tyr(tBu)-OH was further tested in a wide variety of solvents with or without the addition of the non-gelling agent, piperidine or NaOH. As shown in Table 2, L-Tyr(tBu)-OH forms a gel with a wide variety of organic solvents at remarkably low concentrations. Among the tested conditions, DMF appears as the best solvent for gelation, with an ability to gel at 0.1 wt/v% without any additive. Further, 2-ethylhexanol, when used as a solvent, shows promising gelation results with a concentration as low as 0.2 wt/v%. The addition of a non-gelling agent to the latter gel merely changes the appearance of the gel to transparent and does not have a significant impact on the minimum gelation concentration. Similarly, toluene, hexane, and 1,2-dichloroethane also appear to be good solvents for gelation only with the addition of piperidine. Presence of piperidine decreases the gelation concentration for the solvents tert-butylmethylether and 1,2-dimethoxyethane, of which they formed solution at the defined concentrations without additives. Surprisingly, cellulose thinner, isopropylalcohol, and n-butylalcohol, when used as solvents, result in a clear gel without piperidine, whereas the addition of piperidine forms solutions. The gelation of sunflower oil indicates a promising potential application of L-Tyr(tBu)-OH in the field of drug delivery . Similarly, the gelation of diesel by L-Tyr(tBu)-OH was observed, indicating a possible function in oil-spill recovery 3,10 . ## Gelation and Fluoride Ion Response of L-tyr(tBDMs)-oH. We synthesized L-Tyr(TBDMS)-OH (Fig. 2a) and investigated its gelation properties under various conditions. The gelation of this derivative was achieved in both THF and 2-ethylhexanol, with a minimum concentration of 1 wt/v% in both solvents. L-Tyr(TBDMS)-OH is sensitive to the presence of fluoride, since the latter can trigger the cleavage of the Si-O bond to form L-Tyr-OH, as depicted in Fig. 2b. The addition of sodium fluoride to L-Tyr(TBDMS)-OH gels in 2-ethylhexanol in concentrations of 0.2, 0.3 and 0.5 ppm resulted in a complete gel to solution transition within 44 h, 18 h, and 1 h, respectively (Fig. 2c). Fluoride ion cleave the TBDMS moiety to yield L-Tyr-OH, which, as previously discussed, does not show gelation properties in 2-ethylhexanol, explaining the gel to solution transition observed after a certain period of time. We therefore suggest L-Tyr(TBDMS)-OH as a potentially promising gelator that can be used for the detection of fluoride ions. www.nature.com/scientificreports www.nature.com/scientificreports/ Characterization of Microstructures and Gel Behaviours. The characterization of the microstructure of the L-Tyr(tBu)-OH gel as well as the molecular packing at an atomic scale was performed using transmission electron microscope (TEM) imaging (Fig. 3a), X-ray powder diffraction (XRD) measurements (Fig. 3b), and molecular dynamics (MD) simulations. In Fig. 3a, the TEM images of the L-Tyr(tBu)-OH in THF with the addition of piperidine show the formation of nanofibres with an approximate width of 40 nm and a length of several micrometres. The XRD patterns reported in Fig. 3b for the samples prepared with and without piperidine additive indicate that piperidine does not participate in the molecular packing, since the two patterns are nearly identical. MD simulation of L-Tyr(tBu)-OH in THF resulted in the spontaneous formation of aggregates stabilized by strong hydrogen bond interactions between the carboxylate and ammonium groups of the gelator (Fig. S7a). A similar simulation including piperidine molecules also resulted in a spontaneous self-assembly, though with a significantly different interaction motif (Fig. S7b). As such a different molecular packing would result in a different XRD pattern, we therefore definitely rule out the structural role of piperidine in the formation of the gel. The corresponding results are further discussed in the Supporting Information. The subsequent microsecond-long MD simulation of pure L-Tyr(tBu)-OH highlights the formation of long networks of interacting molecules, which form a series of parallel packed fibre-like structures (Fig. 4), in agreement with the TEM imaging. As highlighted in Fig. 4c and further detailed in Fig. S7, each fibre is composed of a hydrophilic core with a strong and compact network of hydrogen bonds, while the hydrophobic side chains point outward and ensure interactions between fibres. We also noted the sporadic occurrence of hydrogen bonds branching between the fibres, which are likely to participate in the overall stability of the molecular assembly. The analysis of radial distribution functions (RDF; Fig. S8) for specific pairs of atoms along the dynamics revealed a series of well-defined peaks centred at distances consistent with the diffractions seen in the XRD pattern. Our simulations showed no evidence of stable π−π stacking interaction between aromatic rings of L-Tyr(tBu)-OH, which is most likely due to the large steric hindrance by tBu groups. Instead, the sharp and intense peak at 2θ = 26.65° (dhkl = 3.34 ) in the XRD pattern can be attributed to the tight interaction observed between carboxylate and ammonium groups along the simulation, with a carbon-nitrogen RDF strongly peaking at a distance of 3.33 . Additional discussions, simulations details, and parameter files are also available in the Supporting Information. The gelation process is triggered by driving forces similar to those at play in the formation of reverse micelles. The hydrophobic character of the solvent enhances the tendency of the polar part of L-Tyr(tBu)-OH molecules to interact with each other, in an isolated core with hydrophobic tBu groups pointing towards the solvent. Unlike micelles, however, the small size of the hydrophilic amino acid backbone together with the highly directional and dipolar character of the interacting chemical groups favours the formation of a linear network rather than spherically shaped vesicles. As predicted by MD simulations (Figs 4 and S8) this linear network can branch through the interconnection of fibres via hydrogen bonds and van der Waals interaction. This last observation leads us to postulate that the gelation results from the formation of a three-dimensional grid of interconnected fibre-like structures in solution. www.nature.com/scientificreports www.nature.com/scientificreports/ In summary, L-Tyr(tBu)-OH and L-Tyr(TBDMS)-OH are forming gels due to a dense network of hydrogen bonds between ammonium and carboxylate groups. A number of amino acid-derived organogelators have been reported in the literature, presenting substitution either on the amine or on the carboxyl group. To the best our knowledge, however, there was thus far no report of a natural amino acid-derived organogelator bearing free -NH 2 and -COOH groups, which we showed here to be the main factor affecting the three-dimensional structure of the gel. Also, tert-butyl moiety appears to play a crucial role in the gelation process by preventing the π−π interactions due to its steric hindrance. This is further showed by the incapability of L-Tyr(OMe)-OH to form a gel. The methyl group seems to be too small to prevent the π−π interactions, leading to precipitation instead of gelation in THF. Rheological measurements were conducted to investigate the gel behaviours of both compounds, L-Tyr(tBu)-OH and L-Tyr(TBDMS)-OH (Fig. 5). A large difference between dynamic storage modulus G″ and loss modulus G′ at all frequencies indicates that the organogels in question show elastic character as a soft matter dominantly. To prove the favourable effect of non-gelling additive on the gel, piperidine added gel was also investigated (Fig. 5a). Although we showed that the piperidine additives did not participate in the molecular structure of the gel, the rheological results clearly showed that such non-gelling additives enhanced the gel properties. The greater difference between G″ and G′ for the piperidine containing organogel compared to additive free organogel leads us to conclude that the presence of non-gelling additive enhanced the gel properties of these soft material. ## Conclusion The present work reports the serendipitous discovery of a new LMW organogelator, L-Tyr(tBu)-OH, which is an inexpensive compound commercially available from many common vendors and which can be synthesized in a straightforward manner from Fmoc-L-Tyr(tBu)-OH in one step. To the best of our knowledge, L-Tyr(tBu)-OH currently represents one of the lightest LMW organogelator to have been reported in the literature and the only natural amino acid-derived gelator having free -NH 2 and -COOH groups. Other amino acid-derived organogelators known in the literature form gel in limited types of solvents or require synthesis involving several steps. Our analysis showed that the molecule has the ability to form a gel in a wide range of organic solvents (polar, www.nature.com/scientificreports www.nature.com/scientificreports/ protic, apolar etc.), with a minimum gelation concentration as low as 0.1 wt/v% in N,N-dimethylformamide. Its gelation in sunflower oil and diesel indicates promising potential for applications in the fields of drug delivery and oil-spill recovery, respectively. We also found that the addition of a non-gelling additive, such as piperidine or NaOH, lowers the minimum gelation concentration and increases the mechanical properties of the gel without any disruption of the molecular arrangement. Additionally, we synthesized another new organogelator, L-Tyr(TBDMS)-OH, in a single step from L-Tyr-OH. The O-silylation of tyrosine is much simpler than alkylation, because O-alkylation of L-Tyr-OH requires extra protection of COOH and NH 2 groups which makes the synthesis quite long. While slightly heavier than L-Tyr(tBu)-OH, this new low molecular weight organogelator demonstrates great potential of being used as a chemosensor for the detection of fluoride ions in drinking water in ppm quantities. ## Methods Materials. All reagents are commercially available and used without any further purification. Fmoc-L-Tyr(tBu)-OH, Fmoc-D-Tyr(tBu)-OH and L-Tyr-OH were purchased from Chem-Impex International Inc.; tert-Butyldimethylchlorosilane from TCI Chemicals; DMF, 1,2-DME, imidazole, methanol and ethanol from Sigma Aldrich; THF, Toluene, MTBE, 1,2-DCE, n-BuOH, diisopropylamine and DBU from Merck; ACN, triethylamine from Carlo Erba Reagents; Piperidine and 2-ethylhexylamine from Acros Organics; hexane and isopropylamine from Lab Scan; diethylamine from Riedel-de Haen; 2-ethylhexanol from Veskim. ## Synthesis of L-Tyr(TBDMS)-OH (2). L-Tyr(TBDMS)-OH was synthesised based on the literature synthesis of L-DOPA(TBDMS) 2 -OH with slight modifications 33 . TBDMS-Cl (1.20 g, 8.28 mmol) dissolved in anhydrous MeCN (12.5 mL) was added on L-Tyr-OH (0.100 g, 5.05 mmol). The mixture was cooled on an ice-water bath for 10 minutes. Then DBU (1.24 mL, 8.28 mmol) was added dropwise to the reaction mixture over 10 minutes. The reaction was stirred in ice bath for 4 hours and after that an additional 20 hours at room temperature. Then, the solvent was removed under vacuum. When methanol was added, undesired precipitate was formed. After filtration, solvent was removed under vacuum. The crude product was washed with water then, with ethyl acetate to obtain pure product. The proton and carbon NMR is shown in Figs S4 and S5, respectively. 1 H NMR (400 MHz, CD 3 OD) δ 7.00 (2 H, d, J = 8.5 Hz), 6.63 (2 H, dd, J = 6.6, 1.9 Hz), 3.55 (1 H, dd, J = 8.7, 4.3 Hz), 3.05 (1 H, dd, J = 14.6, 4.2 Hz), 2.75 (1 H, dd, J = 14.6, 8.7 Hz), 0.80 (9 H, s), 0.00 (6 H, s); 13 General Gelation procedure. In order to prepare 1% (w/v) gel; 10 mg L-Tyr-(tBu)-OH was weighed and placed into a vial. 1.0 mL solvent was added. The vial was placed into ultrasonic bath for 4 minutes. Then 10.0 μL additive was added for the additive containing gels. Eppendorf tube was again placed into ultrasonic bath for 4-10 minutes. The temperature of ultrasonic bath is around 40 °C. Gel formation occurs spontaneously or up to 1 hour depending on the solvent used. Gel formations were proven by using inversion test. Fluoride Ion Response of gels of L-tyr(tBDMs)-oH. After forming gel in 2-ethylhexanol as 1 wt/v%, 10 μL sodium fluoride solution was dropped on gel from the stock solution (5 mg NaF in 1 mL water) to analyze the effect of 0.5 ppm NaF. Then, it was allowed to stay without stirring. Complete gel to solution transition was observed with naked eye and inversion test. transmission electron Microscopy Imaging. FEI Tecnai G2 Spirit BioTwin CTEM microscope was used to image the fibrilar formations after self-assembly. 1 wt/v% gel of L-Tyr(tBu)-OH and Tyr(TBDMS)-OH were prepared freshly in THF/piperidine. After diluting 50-fold with water, it was applied on Cu grid. Excess solution was removed after 2 minutes and grid was stained with 2% uranyl acetate solution. www.nature.com/scientificreports www.nature.com/scientificreports/ without piperidine addition freshly for the rheological measurements. Similarly, 1 wt/v% gel of L-Tyr(TBDMS) was also prepared to investigate its rheological behavior. Physica MCR 301, Anton Paar was used. At first, the linear viscoelastic regimes of deformations of the organogels were determined by a strain-sweep experiment. With a limit of linearity of G′, the strain values that used for the strain-controlled analyses were assigned. Then, the visco-elastic character of the samples was examined by dynamic storage modulus G″ and loss modulus G′. Frequency sweep was scanned from 0.1 to 100 Hz by using a constant target strain determined before. Powder X-Ray Diffraction Measurements. XRD measurements were conducted using X'Pert³ MRD with Cu Kα X-ray radiation (λ = 1.540598 ). Gels of L-Tyr(tBu)-OH in THF and THF/piperidine were prepared as 3 wt/v% and allowed to dry in air for overnight and then vacuum was applied to obtain xerogels.

chemsum

{"title": "Simple Tyrosine Derivatives Act as Low Molecular Weight Organogelators", "journal": "Scientific Reports - Nature"}

electrophilic_sulfur_reagent_design_enables_catalytic_syn-carbosulfenylation_of_unactivated_alkenes

## Abstract: A multi-component approach to structurally complex organosulfur products is described via the nickel-catalyzed 1,2-carbosulfenylation of unactivated alkenes with organoboron nucleophiles and tailored organosulfur electrophiles. Key to the development of this transformation is the identification of a modular N-alkyl-N-(arylsulfenyl)arenesulfonamide family of sulfur electrophiles. Tuning the electronic and steric properties of the leaving group in these reagents controls pathway selectivity, favoring three-component coupling and suppressing side reactions, as examined via computational studies. The unique synstereoselectivity differs from traditional electrophilic sulfenyl transfer processes involving a thiiranium ion intermediate and arises from the directed arylnickel(I) migratory insertion mechanism, as elucidated through reaction kinetics and control experiments. Reactivity and regioselectivity are facilitated by a collection of monodentate, weakly coordinating native directing groups, including sulfonamides, alcohols, amines, amides, and azaheterocycles. Main Text: Organosulfur compounds have diverse functions and find applications as pharmaceuticals and functional materials. 3 Organosulfides, in particular, can also be readily converted into other functional groups, making them versatile building blocks in synthesis. 4 While reliable approaches for catalytic twocomponent C−S bond formation via cross-coupling reactions 7 and C−H functionalization 8 have emerged during the past two decades, multicomponent C-S bond-forming reactions remain less explored. In this context, reactions that merge an alkene, a sulfur-based reaction partner, and a carbogenic group combine C-C skeletal formation and C-S installation into a single operation, representing an attractive means of generating complex, stereochemically dense organosulfur products from simple chemical inputs. Pioneered by Trost , Denmark 13 and others 14 , the most well-established method for 1,2-carbosulfenylation of alkenes involves formation of a thiiranium ion intermediate 15 through electrophilic sulfenyl transfer and subsequent nucleophilic ring opening, with anti-stereoselectivity arising from the SN2 nature of the ringopening step and regioselectivity dictated by alkene substitution pattern. While enabling in its own right, existing methodology for three-component anti-1,2-carbosulfenylation of alkenes is limited to a small collection of carbon-nucleophiles, namely cyanide 11 , acetylides 12 , and organozinc reagents 14 , and in the latter case, the transformation is only compatible with styrene substrates. Complementing this existing methodology with a syn-stereoselective counterpart that proceeds via a distinct mechanistic pathway would be highly enabling. By bypassing thiiranium ion formation, we envisioned that it would not only be possible to achieve the opposite stereochemical outcome but also to expand the alkene scope to include nucleophilic functional groups that are prone to intramolecular cyclization in established thiiranium chemistry. 16 Recently, nickel-catalyzed redox-neutral 1,2-functionalization of unactivated alkenes has emerged as a powerful method for joining together a nucleophile, an electrophile, and an alkene in a selective fashion. 19 Depending on the identity of the coupling partners, two closely related yet distinct mechanisms can operate. A Ni(0)/Ni(II) cycle occurs by an oxidative-addition-first pathway, where the electrophile is incorporated distal to the directing group. Meanwhile, a Ni(I)/Ni(III) cycle involving a transmetalation-first mechanism typically manifests in reversed regioselectivity. In both cases, synselectivity is dictated by inner-sphere migratory insertion. With these considerations in mind and informed by our previous experience in developing nickel-catalyzed 1,2-difunctionalization of alkenes directed by native functional groups 22 , we reasoned that it would be possible to develop a syn-selective 1,2carbosulfenyation by employing sulfur-based electrophiles within this mode of catalysis. While attractive in principle, the envisioned three-component coupling would require surmounting several challenges, including premature sulfenyl transfer to the alkene or carbon nucleophile, competitive β-hydride elimination from the alkylnickel intermediate, and the potential inhibition from the sulfur-containing products or leaving group generated upon sulfenyl transfer. Herein we report a nickel-catalyzed syn-selective 1,2-carbosulfenylation reaction of simple unactivated alkenes 32 . Critical to the success of this reaction is the identification of N-sulfenyl-N-alkyl sulfonamides as N-S electrophiles. Taking inspiration from various electrophiles in the literature that contain nitrogen-based leaving groups for fluorination , trifluoromethylthiolation 35 , cyanation 36 , and acylation 37 , this design takes advantages of the sterically and electronically tunable nature of the sulfonamide leaving group (NLG), which can be readily modified along each of its two vectors to control pathway selectivity and thereby maximize the yield of desired three-component coupling. ## Results and Discussion: Reaction discovery. To begin the investigation, we focused on the model reaction of alkenyl sulfonamide 1 38 with p-tolylboronic acid neopentyl glycol ester (2a) and various N−S electrophiles. In effort to find a suitable sulfenyl electrophile with the proper reactivity profile-appropriately tempered to not transfer the sulfenyl group to the alkene or arylnickel(I) intermediate yet sufficiently reactive to engage the alkylnickel(I) intermediate with a rate faster than that of β-hydride elimination-we first attempted established sulfenyl electrophiles from the literature (S1−S7). Modest success was found with N-sulfenyl lactams (17-51% yield, S5−S7) 9 ; however, desired three-component coupling was accompanied by formation of nearly equimolar quantities of oxidative Heck byproducts, indicating similar rates of intramolecular β-hydride elimination and intermolecular N-S oxidative addition. Extensive screening of the reaction conditions did not improve yields or product selectivity with leaving groups NLG5−NLG7. In contrast, with an N-Me benzamide as leaving group, better yields (53-64%) and product selectivities (7-11% combined byproducts) were observed (S8−S10). We then examined S11−S23 with N-substituted arylsulfonamides as leaving groups, taking advantage of the easily tunable nature of this scaffold, which allows numerous modifications to be quickly assayed. Although electron-withdrawing groups (as in S11−S13) 39 proved deleterious for the reaction, improved yields (69-94%) were observed with S14−S17. When a para-methoxy group was used (S17), the desired product was obtained in 94% yield and with <5% oxidative Heck byproducts. Steric modification to either the arylsulfonyl or N-alkyl vectors revealed no further improvement (S18−S22). Interestingly, when the N-alkyl group was changed to a N-phenyl group, no desired product was observed (S23). 40 We performed computational parameterization of the tested reagents to understand the origins of the high product yield with S17. As N−S reagents were expected to be involved mainly during the oxidative addition step, LUMO energy and bond dissociation energy (BDE) 41 were computed and plotted against percentage yields. The former value relates to kinetics and thermodynamics of organonickel coordination to the N−S reagent, while the latter reflects the subsequent oxidative addition process. By visualizing the data using a 3D-plot, we were pleased to find the maximum yield is obtained when both parameters are within certain threshold (50 to 52 kcal/mol for BDE and −0.6 to −0.1 eV for LUMO energy). Within this data set, analysis of several poor-performing electrophiles is informative. Reagents S11−S13 possess BDEs near the optimal range but have low LUMO energy (mean of −0.9 eV), reflecting the presence of the electron-withdrawing groups. On the other hand, S23, which contains the same leaving group as a previously reported radical fluorinating agent, 40 has a LUMO energy within the typical range of high-yielding variants but features a BDE of 39.6 kcal/mol-as low as the well accepted SOMO-phile S1 (39.8 kcal/mol) 39 . In this case, singleelectron transfer (SET) may take place to generate off-cycle species (i.e., a stable nitrogen centered radical and a sulfur radical). This also agrees with our hypothesis that S17 functions as a 2e − covalent electrophile. The moderate yields obtained with S5−S10 could reflect the slightly higher than optimal BDE of 52.7 kcal/mol, with the sulfur possessing diminished electrophilicity due to the weaker electron-withdrawing effect of the N-alkylamide versus N-alkylsulfonamide leaving group. Interestingly, S4, which has been applied in Lewis-base-catalyzed sulfenyl transfer reactions involving thiiranium intermediates 13,16 , was too stable to react (BDE = 79.2 kcal/mol). X-ray crystallographic analysis of representative N-S electrophiles revealed a non-pyramidalized N-atom 42 stemming from delocalization of the lone pair electron from nitrogen to the S=O π* orbitals (S17, S18). Meanwhile, in the case of S23, a more pyramidalized nitrogen and longer N−S bond indicates diminished delocalization and thus a weaker N−S bond. S6 on the other hand, features an approximately trigonal planar structure which is responsible for its higher BDE (55.3 kcal/mol) and lower reactivity. ## Substrate scope. Having identified an optimal class of N-S electrophiles, we examined the scope and limitations of the method. Regarding the electrophile scope, arylsulfenyl moieties with electron-neutral or -withdrawing substituents (3aa−3ag) could be efficiently introduced with the optimal leaving group (NLG17). On the other hand, the highly electron-donating p-methoxy case (3ab) benefited from using a less electron-rich leaving group (NLG15), indicating that it is possible to further fine-tune the leaving group to match the electronic properties of the arylsulfenyl moiety as needed (see SI for comparative data). Sterically bulky arylsulfenyl groups did not significantly inhibit the reaction (3ah−3ak), and halo-substituted arylsulfenyl coupling partners were tolerated (3ac−3ae, 3ai−3aj), offering the opportunity for further derivatization by cross-coupling. Next, we surveyed different organoboron nucleophiles. Arylboron coupling partners substituted at the paraor meta-positions with electron-donating or -withdrawing groups were generally well tolerated, with the former slightly higher-yielding in general (3ba−3bo). Potentially inhibitory or reactive groups, such as -NHBoc, -CHO, and -CN, were all compatible (3bg−3bi, 3bm−3bn). Furthermore, with an ortho-methyl substituent, 3bp was obtained in 56% yield. Bicyclic coupling partners also proved suitable, giving 3bq−3bt in moderate to good yield. Six-membered heteroaryl nucleophiles, including quinoline, pyridine, and purine moiety, could be installed though a substituent at the 2-position was required to attenuate the coordinating ability of the basic N(sp 2 ) atom (3bu−3bw). On the other hand, unmasked pyridine and purine as well as electron-rich five-membered heterocycles were incompatible in this reaction. Alkenyl-B(nep) reactants constitute another limitation, as only 20% of the corresponding product was observed in the case of a representative styrenyl coupling partner. A series of substituted alkenyl sulfonamides were evaluated. Sulfonamides derived from both homoallyl and bishomoallyl amines were tolerated (4a−4m), with the former giving slightly lower yield. Gratifyingly, benzenesulfonyl groups with different substituents at the para-position (4a−4f) and a methanesulfonyl group (4g) all furnished the corresponding products in moderate to good yield. 4-Cyanobenzenesulfonamide (Cs) can serve the dual role of directing group and amine activating group (see below) and gave 4e and 4f in 53% and 85% yield, respectively. 43 Moderate yield and diastereoselectivity were obtained with substrates containing α-or β-branching (4h−4j). With internal alkenes, Z-configured alkenes provided higher yield, as well as greater and opposite diastereoselectivity, compared with the corresponding E-isomers (4k−4l'). A sulfonamide derived from ortho-allylaniline furnished 4m in 59% yield with no chain-walking byproducts detected. We next investigated alkenes containing other synthetically useful native functional groups. This seemingly straightforward extension is complicated by the fact that minor changes to the directing group structure have a significant impact on coordination strength and metallacycle stability. To our delight, after brief optimization of base and solvent (see SI for details), a diverse collection of native directing groups was found to facilitate this transformation. Starting from homoallyl alcohol, 5a and 5b were obtained in 64% and 72% yield, respectively. 27 A large-scale experiment on 4 mmol scale furnished 5b in 59% yield. Substituents at the α-position with respect to the hydroxyl group furnished 5c and 5d in 49% and 61% yields, respectively, with negligible diastereoselectivity. Late-stage functionalization of allylestrenol, a progestin medication, furnished the corresponding product in 34% yield with 1.5:1 diastereoselectivity (5e). Free alkenyl amines bearing a basic N(sp 3 ) atom are traditionally challenging substrates owing to their ability to sequester the catalyst off-cycle. Gratifyingly, 5f−5j were obtained in moderate to good yield. With a chiral benzyl amine directing auxiliary, 5h was obtained in 60% yield and 2:1 diastereoselectivity. In addition, an alkene bearing an amide group proved compatible, giving 5k in 35% yield. Motivated by our previous success in developing a 1,2-allylmethylation reaction of alkenes directed by diverse azaheterocycles 44 , we attempted these useful substrate families and were pleased to find that pyridine (5l), pyrazoles (5m−5o), triazoles (5p−5q), tetrazole (5r), indazoles (5s−5t), and benzotriazole (5u) were all compatible directing groups. Diastereomeric ratios were determined by 1 H NMR analysis of the isolated products. For 5b, the yield in brackets was obtained from an experiment performed on 4.0 mmol scale. ## Synthetic versatility. The N−S reagents used in this study can readily be prepared on decagram scale using modular and operationally convenient chemistry. If desired, the sulfonamide leaving group can be recovered in nearly quantitative yield and recycled (see SI for detail). The reaction can be performed outside of the glovebox using the air-stable precatalyst Ni(cod)(DQ) under Schlenk technique, as demonstrated using the model reaction in Fig. 5a (80% yield, entry 1a). 45 A series of stress tests showed that this protocol is fairly robust. Simply purging with nitrogen over the solvent surface was sufficient (entry 1b), and even running the reaction under air gave 44% yield (entry 1c). When 1 mol% catalyst loading was used with extended reaction time of 40 h, 70% yield was obtained (entry 2). Reducing the electrophile and nucleophile loading led to lower yet synthetically useful yields (entry 3). A series of product derivatizations were performed to showcase the synthetic utility of the method (Fig. 5b). From a -NH(Cs) product , N-alkylation and -arylation followed by mild deprotecting furnished 6a and 6b. 43 This sequence allows straightforward incorporation of groups that are problematic in the nickel-catalyzed 1,2-carbosulfenylation (e.g., nitroarenes). Hofmann−Löffler−Freytag (HLF) cyclization furnished 1,2-disubstituted pyrrolidine 6c in 62% overall yield over 2 steps. 46 Controlled oxidation of the thioether to the sulfoxide or sulfone and successive oxidation to the sulfonyl carboxylic acid using different oxidants furnished 6d−6f in excellent yields 9 , with 6e representing a key intermediate in the synthesis of matrix metalloproteinases (MMP) inhibitor. 47 Finally, a thianthrene-based radical precursor 6g was prepared over 3 steps in 72% overall yield. Mechanistic studies. A control experiment with N-methyl alkenyl sulfonamide (7a) under standard conditions did not lead to product formation, suggesting that the sulfonamide group coordinates through nitrogen (Fig. 5c). Similarly, no desired product was formed from styrenyl sulfonamide (7b), which rules out a stepwise oxidative Heck/hydrosulfenylation pathway. The relative stereochemistry of 4l' (major) was determined by single-crystal X-ray diffraction of an oxidized derivative, indicating a syn-addition mechanism. Kinetic studies were performed using the conversion of 1 to 3bd as the model reaction applying the method of initial rates (Fig. 5d, see SI for detail). The reaction was positive order in [Ni(cod)2] and ## Conclusion: In conclusion, a novel family of N−S reagents with N-alkylsulfonamide leaving groups was developed to enable 1,2-carbosulfenylation of unactivated alkenes under nickel catalysis. The synthetic versatility of the method stems from the broad scope of compatible alkenes bearing different native directing groups. The identification of N-S reagents based upon modular components allows for fine tuning of the electronic and steric properties, which ultimately affect stability, reactivity, and catalytic performance. DFT calculations indicate that optimal yield is achieved when parameters such as BDE and LUMO energy fall within a narrow range of values. This finding is expected to guide the development of other heteroatom-based electrophiles for use in multicomponent catalytic couplings.

chemsum

{"title": "Electrophilic Sulfur Reagent Design Enables Catalytic syn-Carbosulfenylation of Unactivated Alkenes", "journal": "ChemRxiv"}

covid-19:_attacks_the_1-beta_chain_of_hemoglobin_to_disrupt_respiratory_function_and_escape_immunity

## Abstract: Investigating poor respiratory function and high immunological escape in COVID-19 patients may aid in the prevention of additional deaths. The conserved domain search method was used to evaluate the biological roles of specific SARS-COV-2 proteins in this present study. The research findings indicate that the SARS-COV-2 virus contains domains capable of binding porphyrin and synthesizing heme. S and ORF3a can bind to hemoglobin. The S protein possesses hemocyanin-like function since it contains copper-oxygen binding, immunological agglutination, and phenoloxidase domains. ORF3a's Arg134 and E's Cys44 have heme-iron binding sites, respectively. The ORF3a protein has a region that degrades trapped heme into iron and porphyrin. Hemoglobin that has been attacked by ORF3a may preserve the majority of its native structure but with decreased oxygen delivery function. By targeting hemoglobin and destroying heme, the ORF3a protein caused varying degrees of respiratory distress and coagulation symptoms in COVID-19 individuals. ORF3a of Delta and Omicron variants also retained its capacity to target hemoglobin and heme. But the S protein's hemocyanin-like domain transported oxygen to enhance the patient's respiratory condition. Through a large load of hemocyanin-like proteins, the mutant virus achieved effective oxygen transport and alleviated the symptoms of respiratory distress in patients. Simultaneously, the variant S protein's immunological agglutination and phenol oxidase functions were decreased or eliminated, resulting in a decrease in the strength of the immune response and an increase in immune evasion ability, culminating in increased virus transmission. ## Background COVID-19 patients exhibit symptoms such as diarrhea (1,2), hypotension (3), and electrolyte abnormalities (4). Early asymptomatic infections cause taste and smell loss (5,6). Patients with severe COVID-19 also have a higher risk of developing rare neurological illnesses such as epilepsy (7,8) and encephalitis (9,10). Cytokine storm causes organ failure in the later stages of severe inflammatory infection, including the heart (11), liver (12), and kidneys (13). Patients with COVID-19 will suffer varying respiratory distress symptoms during the early and severe stages. Numerous ground-glass pictures and piercing shadows occur on both sides of the lungs (14,15). Ground-glass images are frequently linked with severe hypoxia. Despite significant hypoxemia, some individuals with COVID-19 pneumonia experience no dyspnea, demonstrating the "happy hypoxia" paradox. Severe patients rescued with Extracorporeal Membrane Oxygenation (ECMO) exhibit unusual clinical characteristics, including low oxygen, low blood oxygen saturation (16,17), and high dissolved oxygen levels. Thus, elucidating the mystery respiratory issues and immune response of COVID-19 patients may be a critical step for increasing patient rescue success rates (18). Colistin B and coronavirus may result in significant skin darkening in patients with COVID-19 who are rescued with ECMO (19). Dermatological symptoms of COVID-19 include (20): urticaria, erythema confluent/maculopapular/measles-like rash (21), papular rash, chilblain-like lesions, livedo reticularis/racemoid pattern, and purpura" Vasculitis". The erythema associated with certain severe skin diseases might be converted to hyperpigmentation. Patients with COVID-19 ocular infection develop unilateral acute posterior multifocal lamellar pigment epitheliopathy (22). COVID-19 patients have bilateral pigmentation of the corneal endothelium, pigment dispersion in the anterior chamber, iris depigmentation with iris transillumination abnormalities (23). Melanin also has a vital role in the coloration of the oral mucosa (24). Oral hyperpigmentation caused by HIV can affect any area of the oral mucosa (25). Melanin affects the inflammatory response directly or indirectly by influencing the generation of host cytokines/chemokines (26). Melanin alters the signaling cascades mediated by cytokines. It boosts the release of pro-inflammatory mediators such as interleukin (IL)-1, IL-6, interferon gamma , and tumor necrosis factor (26). Living creatures develop pigments to protect themselves from UV rays. Melanin is the most strong and efficient of all pigments. When UV radiation strikes melanin's chemical chain, the chemical chain vibrates at an incredibly high rate, turning dangerous UV light into innocuous heat. When the retina is activated by damaging light, harmful free radicals are created, initiating a sequence of destructive oxidative events on the retina. Melanin can act as an antioxidant and protect the eyes by neutralizing these damaging oxidation events. Melanin also protects microorganisms from enzymatic destruction, radiation (UV, sunlight, gamma), and heavy metals, as well as increasing their thermal tolerance (to heat and cold) (27). Melanin enhances virulence by shielding fungal cells from phagocytosis by immune effector cells (28). Bacterial melanin-like pigments have been shown to be capable of scavenging superoxide anion free radicals and inhibiting monocytes' respiratory burst response (29). Through melanin, the SARs-COV-2 virus may achieve immune evasion and radiation protection. However, in COVID-19 patients, an overabundance of synthetic melanin might result in hyperpigmentation. Pigmentation also occurs in some pathogen-infected lower species. Hemocyanin's extracellular phenoloxidase (PO) is the primary source of hyperpigmentation (melanosis) in shellfish (30,31). Phenoloxidase activity resulting from hemocyanins contributes to brown algal hyperpigmentation. Hemocyanin is the pigment that causes hyperpigmentation in N. norvegicus (32). Hemocyanin is homologous to phenoloxidases such as tyrosinase because both proteins have a type 3 Cu active site coordination (33). Enzymes of phenoloxidase (PO), such as tyrosinase (EC 1.14.18.1) and catechol oxidase (EC 1.10.3.1). Catecholamine (CA) is a catechol derivative. Epinephrine (E), norepinephrine (NE), and dopamine (D) are all endogenous CA. The enzymatic reaction catalyzed by phenoloxidase results in chromogen, melanin, and other pigments (34). The phenoloxidase (PO) enzymes significantly contribute to hyperpigmentation (35). Insect prophenoloxidase (PPO) is also an important innate immune protein as it is involved in cellular and humoral defenses (36). Hemocyanin comprises three domains (37): a N-terminal domain, an active site containing binuclear copper ions, and a C-terminal domain (38). Conformational changes in the N-terminal domain can activate hemocyanin's phenoloxidase activity. The C-terminal domain contribute in organisms' immunological agglutination activity (39) and increase blood cell phagocytosis (40). Hemocyanin is found only in hemolymph and appears in hexamers or hexameric oligomers (41). The active site catalyzes the chelation of two copper ions and the binding of an oxygen molecule (42). Hemocyanin is more than twice the size of hemoglobin (43). Hemocyanin reversibly binds to oxygen molecules via altering the valence of copper ions(Cu 2 +↔Cu+) (44). It binds 96 oxygen molecules, whereas hemoglobin only binds four (43). Hemocyanin molecules float freely in the blood, whereas red blood cells contain millions of smaller hemoglobin molecules (43). Hemocyanin and hemoglobin have complementary distributions in some insect orders (45) and crustaceans (46). Hemocyanin and hemoglobin are both present in the crustaceans, Hymenoptera and Hemiptera. Hemocyanin works as a physiological supplement, compensating for poor oxygen transport in the trachea and assisting insect embryos in aerobic respiration (47). Hemocyanin expression was dramatically increased in Baifutiao (48) and locust embryos (49) under hypoxia conditions compared to normoxic settings. Increased hemocyanin content assists locust embryos in obtaining adequate and steady oxygen in hypoxic high altitude locations (49). Hemocyanin expression was also significantly elevated in the blue crab Callinectes sapidus (50) and the ecliptic crab Cancer magister(51) under hypoxic conditions (52). Altitude sickness' physiological characteristics and symptoms at high altitude are comparable to those of various illnesses related to COVID-19 (53). The SARs-COV-2 virus may contain a hemocyanin-like protein that binds oxygen and activates the phenoloxidase activity of the virus. The phenoloxidase structure synthesizes melanin. COVID-19 patients had hemoglobin and hemocyanin-like co-transport oxygen patterns. Hemocyanin-mediated oxygenation leads to effective O2 transport under hypoxic settings (54). In this mode, oxygen supplied by hemocyanin-like compensates for bodily hypoxia even when red blood cells or hemoglobin were not working adequately. However, the current approach of light absorption in oxygenation detection considered hemoglobin solely and ignored oxygen-carrying hemocyanin-like. Present blood gas detection technologies presupposed that the oxygen molecules in the blood were dissolved oxygen which was not bound to hemoglobin (55). But the implicit condition that hemocyanin-like could also bind oxygen molecules was ignored. pO2 determined how much oxygen was bound by hemoglobin and hemocyanin-like proteins. Existing assays overestimated hemoglobin's oxygen supply status. The patient's oxygenation curve may remain unchanged. The actual oxygenation curve should be moved to the right, lowering hemoglobin's affinity for oxygen. Thus, when viruses attacked hemoglobin, it exhibited "high dissolved oxygen". In contrast, hemocyanin-like proteins exhibited an abnormal oxygen transport function, resulting in "happy hypoxia." To collect iron and heme from host proteins like hemoglobin in red blood cells, pathogens have evolved various iron absorption methods (56). According to a computational research, the SARs-COV-2 virus infects erythrocytes via the Plasmodium falciparum-like and complement-like domains. Anupam Mitra discovered a viral leukocyte response in COVID-19-infected patients (57). According to Bhardwaj, pRb and its interaction with Nsp15 affect coronavirus infection (58). Electron microscopy revealed the presence of the Colorado tick fever virus in red blood cells (59). Plasmodium, Babesia, and trypanosomes also infect red blood cells, eliciting symptoms similar to COVID-19. Trypanosome extracellular vesicles fuses with mammalian erythrocytes (60), resulting in simple erythrocyte removal and additional anemia. Babesia invasion of red blood cells requires adhesion proteins and apical membrane antigens associated with thrombin sensitivity (61). Pathogenic E. coli strains can use hemoglobin as a source of iron (62). Hemoglobin protease (Hbp) is a proteolytic enzyme similar to IgA1 protease. Immunoglobulin A1 protease (IgA1 protease) is a serine protease (S6 family) that is produced by several pathogenic bacteria (63), including those that cause bacterial meningitis, Haemophilus influenzae, Neisseria meningitidis, and Chain pneumonia Cocci (63). The IgA1 protease colonizes human mucosal surfaces, affecting specific immune responses (63). This serine protease autotransporter, released from E.coli, destroys hemoglobin, binds the liberated heme, and transports it to both bacteria (56). When Staphylococcus aureus's Isd protein degrades hemoglobin, the 1-beta chain is targeted first, releasing heme and initiating a series of heme release events. Heme transfer from met-Hb to IsdH/B is slower than from met-Hb to full-length Hb receptors (64). It is also governed by simple heme dissociation from met-Hb (65). By binding to αHb via the IsdH or IsdB domains, the rate of efficient interactions between the Hb chain and the heme receptor domain is increased (66). This binding is used to specifically target the heme receptor domain, regulating the sequential release of heme from βHb and αHb chains (66). The αHb•βHb dimer releases heme from a single subunit (half-Hb) while retaining the majority of its natural structure (67). It remains linked with the Hb receptor until all heme is released. IsdH does not bind to free apo-Hb. The IsdH-Hb complex dissociates only when the heme in Hb is completely removed (66). The SARs-COV-2 viral protein may interact with αHb via the Isd domains and then break βHb using the IgA1 protease structure. This attack on the 1-beta chain of hemoglobin initiated the sequential release of heme from hemoglobin. The IsdC protein from S. aureus uses a flexible binding pocket to capture heme (68). The crystal structure of the heme-IsdC complex is the central conduit of the S. aureus Isd iron/heme uptake system (69). IsdA , IsdB, IsdH, and IsdC share the same heme-binding module, termed the NEAT (near transporter) domain (69). But the iron-regulated surface proteins IsdA, IsdB, and IsdH may be not required for heme iron utilization in S. aureus (70). Heme is a kind of metalloporphyrin. Heme activates signaling pathways independent of iron and reactive oxygen species (ROS), including those involved in redox metabolism (71). Heme metabolism is required for Plasmodium, the causative agent of malaria, to infect red blood cells (72). Numerous Plasmodium genes encoding heme-binding proteins have been found (73). Hemozoin is frequently referred to as malaria pigment in malaria parasites. Phagocytosis for the malaria pigment heme inhibits CD54 and CD11c expression in human monocytes (59). Hemozoin (malarial pigment) impairs the differentiation and maturation of dendritic cells generated from human monocytes (57). In vitro, the malaria-specific metabolite hemozoin promotes the production of three powerful endogenous pyrogens (TNF, MIP-1, and MIP-1), while in vivo, it changes thermoregulation (58). Hemozoin is involved in various processes that may contribute to Plasmodium pathogenicity. Numerous antimalarial medicines, including chloroquine, act by selectively inhibiting this hemozoin detoxification/hemozoin production pathway (74). Iron is most frequently liberated from heme's oxidative breakdown by heme oxygenase (HO). Most of the time, iron is released from heme due to the protoporphyrin ring degrading (75). C. diphtheriae's heme oxygenase degrades heme to generate α-biliverdin, carbon monoxide, and free iron (76). IsdG and IsdI (heme oxygenase) of Staphylococcus aureus cleaves the tetrapyrrole ring structure of heme in the presence of NADPH cytochrome P450 reductase, thereby releasing iron. (77). However, E.coli's deferrochelating activity does not destroy the tetrapyrrole backbone (78). It is the case with Yersinia enterocolitis HemS(79) and E. coli O157:H7 ChuS (80) . HemS is used by E. coli Bartonella hensii to deal with oxidative stress caused by H2O2 (81). Iron is released from heme by the HemS of E. coli Bartonella henii without causing damage to the tetrapyrrole backbone (81). HemS protein also degrades heme in the presence of electron donors, ascorbate, or NADPH-cytochrome P450 reductase (81). HemS protein stimulates the release of iron from heme, leaving behind hematoporphyrin. Most porphyrin molecules are hydrophobic and agglomerate in water (82). Porphyrin photosensitizers with a higher hydrophobicity penetrate mammalian cell membranes (83). Porphyrins diffuse through the phospholipid bilayer and accumulate in the cytoplasm due to concentration gradients (84). Porphyrin chemicals, such as synthetic photosensitizers, are frequently utilized to treat cancers by photodynamic therapy (85). Porphyrin derivatives get in cancer cells through endocytosis and concentration gradient osmosis (86). Porphyrins produce reactive oxygen species (ROS) that kill tumor cells (87). The SARS-CoV-2 viral proteins bound porphyrin (or heme) to get energy and cell membrane penetration. It generated reactive oxygen species (ROS) to damage the cell membrane. Porphyrins are mostly preserved in the human body as heme on hemoglobin. The virus's high requirement for porphyrin and iron developed to attack hemoglobin and broke heme into iron and porphyrin. The disorder of the body's porphyrin metabolism causes Porphyria. Atypical porphyrins have been identified in patients with acquired immunodeficiency syndrome (88). Chronic hepatic porphyria, not delayed cutaneous porphyria, is the form of porphyrin metabolic illness caused by the hepatitis C virus (88). Numerous clinical observations have revealed that patients with COVID-19 also have skin and nervous system symptoms consistent with porphyria. Genetic abnormalities in the heme biosynthesis enzyme uroporphyrinogen III synthase also cause congenital erythropoietic porphyria (89). The SARs-COV-2 viral protein may has uroporphyrinogen III synthase activity and contribute to infection by producing comparable heme. This type of uroporphyrinogen III synthase inhibited the metabolism of porphyrins. We employed the conserved domain search method to study the SARs-COV-2 viral proteins in this work. The results indicated that the SARS-COV-2 virus could synthesize heme from porphobilinogen and encodes all necessary enzymes for this process. The SARs-COV-2 ORF3a protein could target 1-beta chain hemoglobin and dissociate heme to iron and porphyrin. These attacks by the SARs-COV-2 virus and its variants could cause severe damage to respiratory tissues and organs. Meanwhile, the SARs-COV-2 virus's S protein exhibited hemocyanin-like activity. The S protein's hemocyanin-like domain transported oxygen to enhance the patient's respiratory condition and affect immune response. ## Data set 1. The sequences of SARS-COV-2 proteins. The SARS-COV-2 protein sequences came from the NCBI database. Including: S, E, N, M, ORF3a, ORF8, ORF7a, ORF7b, ORF6, ORF10, orf1ab, orf1a. Among them, the orf1ab and orf1a sequences also included corresponding subsequences. SARS-COV-2 variant's (Delta and Omicro) ORF3 and ORF8 sequences came from the NCBI. 2. Related sequences. The related sequence was downloaded from UniProt data set (Table 1 ## A localized MEME tool to identify conserved domains. The following are the steps involved in the analysis: 1. Downloaded MEME from the official website and installed it in a virtual machine running Ubuntu. VM 15 was the virtual machine. 2. Downloaded the SARS-COV-2 protein sequence from the National Center for Biotechnology Information's official website. 3. Obtained the fasta format sequences of the related protein from the official Uniprot website. 4. Generated fasta format files by MEME analysis for each sequence in all related proteins and each SARS-COV-2 protein sequence. 5. To create multiple batches of the files generated in Step 4, a batch size of 50000 was used. It was limitedby the virtual ubuntu system's limited storage space. 6. Using MEME tools in batches, searched for conserved domains (E-value<=0.05) in SARS-COV-2 and related proteins in Ubuntu. 7. Collected the conserved domains' result files. Located the domain name associated with the motif in the UniProt database. 8. Analyzed the activity of each SARS-COV-2 protein's domains. ## Viral proteins possessed the ability to bind to porphyrin and generate heme. Porphyrin biosynthesis is the process in mitochondria and cytoplasm. The initial stages of production is in Mitochondria. Three enzymes are involved in the production process: aminolevulinic acid synthase, ALAS1, and ALAS2. In the cytoplasm, the intermediate synthesis stage occurs. Porphobilinogen synthase, porphyrinogen deaminase, uroporphyrinogen III synthetase, and uroporphyrinogen III decarboxylase work sequentially in the synthesis process. The final stage of synthesis occurs in the mitochondria. The production process is carried out sequentially by coproporphyrinogen III oxidase, protoporphyrinogen oxidase, and ferrochelatase. The final product of ferrochelatase is heme The majority of mitochondrial synthesis sites are transmembrane areas. If viral proteins are capable of synthesizing porphyrins, these enzymes should be systematically possessed. The viral membrane is analogous to the mitochondrial transmembrane area, while the viral cytoplasm is analogous. As a result, viruses may be able to generate porphyrins via structural proteins. Because non-structural proteins contain both transmembrane and non-transmembrane proteins, they may also produce porphyrin. We obtained porphyrin-related sequences from the UniProt database and then compared the viral proteins to the porphyrin-related sequences using the local MEME tool. We combined the motif sequences by protein and domain due to the vast number of motif pieces. Table 2 and Table 3 summarizes the search results. Table 2 shows structural proteins possess enzymes synthesized heme. Table 3 Porphyrinogen deaminase, uroporphyrinogen III synthase, uroporphyrinogen III decarboxylase, and coproporphyrinogen III oxidase domains are present in both structural and non-structural proteins, as shown in Table 2 and Table 3. There were ferrochelatase domain present. It indicates that structural and non-structural proteins can employ porphobilinogen as a starting material for synthesizing heme, respectively. Multiple viral proteins collaborate to finish the porphyrin production pathway. Due to the limitations of our analysis capabilities, we cannot determine the relation of porphyrin synthesis between viral proteins. These enzyme activity domains are linked to porphyrin. As a result, viral proteins can also interact with porphyrins via these enzyme domains. ## The Haem_bd domain of viral proteins could bind heme. The UniProt database was used to find cytochrome C-related sequences. Then we adopted the local version of MEME to match the viral proteins and cytochrome C-related sequences. Due to a large number of motif fragments, we organized motif sequences by protein and domain. According to the motif CXXCH of the linked sequences, we formed a one-to-one correlation with the results in Table 4, and then sorted out the Haem motif (Table 5). Only E and ORF3a carry the CXXC motif, as shown in Table 5. Hematoporphyrin binds both C molecules. E has C44 as Fe binding sites, and ORF3a has R134. But neither does H. This mutation may make it easier for viral proteins to bind iron. E and ORF3a bind to heme in a relatively steady manner. Other viral proteins' heme-binding could be unstable. ## Viral proteins could bind hemoglobin Eryth_link_C is on the linker subunit of the giant extracellular hemoglobin (globin) respiratory complex. The linker subunit's C-terminal globular domain is involved in trimerization. It also interacts with globin and other adjacent trimers' C-terminal spherical linker domains. In Staphylococcus aureus, the NEAT domain encodes the human hemoglobin receptor. The NEAT domain recognizes a subfamily of iron-regulated surface determinant proteins that are found exclusively in bacteria. Iron-regulated surface determining protein H (isdH, also known as harA) interacts with the human plasma haptoglobin-hemoglobin complex, haptoglobin, and hemoglobin. It has a much higher affinity for haptoglobin-hemoglobin complexes than haptoglobin alone. These three domains serve distinct purposes. IsdH(N1) binds hemoglobin and haptoglobin; IsdH(N3) binds heme that has been released from hemoglobin. We downloaded hemoglobin-related sequences from the UniProt database. We then compared the viral proteins to the hemoglobin-related sequences using the local MEME version. We combined the motif sequence by protein and domain due to the vast number of motif pieces. Table 6 summarizes the search results. S, E, N, ORF3a, ORF6, ORF7a, ORF7b, ORF8, ORF10, 2'-O-ribose methyltransferase contain NEAT domains, as shown in Table 6. The domain Eryth_link_C is present in S, N, ORF6, ORF7b, ORF8, ORF10, nsp6, nsp10, and RdRP. Both Eryth link C and NEAT domains are in S, N, ORF6, ORF7b, ORF8, ORF10, respectively. Because S can form a trimer structure, it can be combined with extracellular hemoglobin. The S Eryth_link_C A is involved in receptor binding. S Eryth_link_C B is a member of the S2 family of proteins involved in membrane integration. The NEAT domains of S, E, N, ORF3a, ORF6, ORF7a, ORF7b, ORF8, ORF10, and 2'-O-ribose methyltransferase may perform distinct functions. S NEAT A and S Eryth_link_C B are mutually overlapped. E NEAT A is shorter proteins that perform the function of heme capture. S, N, ORF3a, ORF6, ORF7b, ORF8, ORF10 have a longer NEAT capable of binding and capturing hemoglobin. ## ORF3a acted as a hemoglobin protease. The autotransporter subfamily has two peptidase S6 domains (IPR030396 and PS51691). S06.001 -Neisseria type serine peptidase specific for IgA1, MEROPS accession number MER0000278. Peptidase S6 domains are found in the IgA-specific serine endopeptidase autotransporter of Neisseria gonorrhoeae, the immunoglobulin A1 protease autotransporter of Haemophilus influenzae, and the hemoglobin binding protease Hbp autotransporter of pathogenic Escherichia coli. In contrast to IgA peptidases (Family S6 contains Neisseria and Haemophilus IgA1-specific endopeptidases), Enterobacter sp. (SPATE) peptidases have not been demonstrated to cleave IgA1. SPATE proteins are all highly immunogenic, and each SPATE member is one of the pathogen's most prominent secreted proteins. Human IgA1 is cleaved by the bacterial IgA1 protease at a hinge region where IgA2 is not present. They are highly selective prolyl endopeptidases, and the site of cleavage within human IgA1's hinge region differs between strains. Hbp attaches to hemoglobin, destroys it, and then binds to the heme that is liberated. His, Asp, Ser, in that order, are the catalytic residues of the S6 family. Ser is found in the Gly-Xaa-Ser-Gly-Xaa-Pro motif, which is highly conserved across the S1-S2 families and the S7, S29, and S30 families. S6 is a member of the PA clan as a result of this. As established experimentally in the Hap protein of Haemophilus influenza, his and asp residues are positioned around 190 and 120 residues N-terminal to the catalytic Ser. By cleaving human IgA1, The protective role of the powerful medium, IgA1 serine peptidase, may disrupt mucosal surface-specific immunity. We used a local version of MEM to compare the hemoglobin degrading-related sequences to the SARS-COV-2 viral protein to identify conserved structures. We combined the motif fragments according to viral protein and domain classification. Table 7 illustrates the domain distribution of Peptidase S6 in SARs-COV-2. Peptidase S6 domains are shown in Table 7 for E, ORF3a, ORF7a, ORF7b, ORF8, and ORF10. However, only the ORF3a Peptidase S6 domains share a similar Gly-Xaa-Ser-Gly-Xaa-Pro motif: C-R-S-K-N-P. C and K are unambiguously derived from Gly or G. This mutation could be used to improve ORF3a's affinity for heme. As determined by heme motif research, CR is the final two letters CH in the Haem bind domain's CXXCH motif. C denotes heme's porphyrin-binding site, while R denotes iron-binding. It indicates that ORF3a acts on hemoglobin via the Peptidase S6 domains of the hemoglobin protease. The catalytic site C-R-S-K-N-P acts on the heme on hemoglobin, and then hunts for and binds to heme directly. ORF3a, ORF8, S, and N, as well as others, have a more significant number of autotransporter domains (90). As a result, ORF3a functioned as an IgA1 peptidase and a hemoglobin protease. ## ORF3a protein possessed IsdA, IsdC, and IsdH activity for uptaking heme. ORF3a can bind to and degrade hemoglobin, as seen in Tables 6-7. ORF3a can also bind heme, as demonstrated in Tables 4-5. It indicates that ORF3a can bind to and degrade hemoglobin before collecting heme. As seen in Table 6, ORF3a has a NEAT domain. All Isd proteins involved in heme absorption contain NEAT domains. IsdA, IsdC, and IsdH/IsdB proteins are present in Staphylococcus aureus. They are capable of completing activities such as hemoglobin localization and heme capture. Table 8 summarizes the protein sequences corresponding to the ORF3a NEAT domain found in the search results. As shown in Table 8, ORF3a's NEAT domains are derived from IsdA, IsdC, and IsdH. The three proteins' motifs, notably the heme-binding motif "CWKCR," are remarkably similar, indicating that their functions are intimately tied to heme capture. The searches mentioned above show that the IsdH domain supports ORF3a in localizing hemoglobin, and then the IsdA domain attacks hemoglobin. The IsdA domain is responsible for transferring the heme from hemoglobin to IsdC. IsdC binds to heme. ## ORF3a protein was capable of converting heme to iron and porphyrin. The ABM domain (IPR007138) is a monooxygenase domain involved in the biosynthesis of antibiotics. This domain is found in the IsdG and IsdI strains of S. aureus. IsdG and IsdI are heme-degrading enzymes that resemble monooxygenases structurally. This domain is also found in the MhuD heme-degrading monooxygenase from Mycobacterium tuberculosis. HemS/ChuX (IPR007845) is a protein that degrades heme. ABM monooxygenases (IsdG and IsdI) released iron from heme by cleaving the heme tetrapyrrole ring structure in the presence of NADPH cytochrome P450 reductase. HemS catalyzes the release of iron from heme without causing damage to the tetrapyrrole backbone. After HemS binds to heme, the tetrapyrrole ring structure of heme also is broken to release iron in the presence of electron donors such as ascorbate or NADPH-cytochrome P450 reductase. We retrieved the sequences of proteins involved in heme degradation from the UniProt database. Then we compared them one by one to ORF3a using the local MEME version. We combined the discovered motif fragments according to conserved domains. The heme degradation domains of the ORF3a protein searched are listed in Table 9. ORF3a contains ABM and HemS domains, as shown in Table 9. ORF3a ABM and ORF3a HemS were homologous. The N-terminus of ORF3a ABM includes more "NFVRI" sequence fragments than the N-terminus of ORF3a HemS. The C-terminal region of ORF3a HemS has more "YDYCIPYNSV" sequence fragments than the C-terminal region of ORF3a ABM. ORF3a ABM and ORF3a HemS both include the heme-binding motif "CWKCR," indicating that ORF3a ABM and ORF3a HemS can bind heme. As shown in Table 9, the ORF3a protein could directly separate heme into iron and porphyrin via the HemS domain. ORF3a protein could also cleave the tetrapyrrole ring of heme via the ABM/HemS domain, thereby degrading heme and releasing iron in the presence of NADPH-cytochrome P450 reductase. Table 9. ORF3a protein has a heme-degrading domain. ## ORF3a protein impaired human respiratory function We retrieved the crystal structure of ORF3a from the PDB database and then annotated the conserved domains that attack hemoglobin and degrade heme using the Discovery Studio 2016 tool (Figure 1). ORF3a is a dimer in Figure 1. The heme-binding motifs "CWKCR" and "CYDYC" of a monomeric ORF3a are labeled in Figure 1.A. Coincidentally, the two phantoms are positioned adjacent, producing a clip. It indicated that the clip's function was to capture the heme and expel the iron. The NEAT domain of a monomeric ORF3a is labeled in Figure 1.B. The NEAT structure comprises densely packed IsdA, IsdC, and IsdH domains. The NEAT domain contains the heme-binding motif "CWKCR" and "CYDYC". The Peptidase S6 domain of a monomeric ORF3a is labeled in Figure 1.C. The Peptidase S6 domain spans the heme-binding motif "CWKCR" and "CYDYC". These two heme motifs contain the unusual X-X-Ser-X-X-Pro pattern CRSKNP. The Peptidase S6 domain overlaps the NEAT domain. The HemS domain of a monomeric ORF3a is labeled in Figure 1.D. The HemS domain spans the heme-binding motif "CWKCR" and "CYDYC". The HemS domain is overlaps to the NEAT, Peptidase S6, and ABM domains. It implies that the sequence fragments around the heme-binding motifs "CWKCR" and "CYDYC" are highly conserved. They are involved in hemoglobin assault and heme degradation. ORF3a may have a mechanism for impairing respiratory function. ORF3a infected erythrocytes via domains like that of Plasmodium falciparum. ORF3a formed ion channels by binding cytoskeletal proteins such as spectrin in red blood cell membranes. The HemS, NEAT, Peptidase S6, and ABM domains were positioned near the cytoplasm in the erythrocyte inner membrane. ORF3a interacted with the hemoglobin protein via the NEAT structure. The NEAT domain possessed the IsdA, IsdC, and IsdH functions. ORF3a might be linked to 1-alpha-Hb via the IsdH domain of the NEAT domain, then the IsdH domain of the NEAT domain was positioned near the heme-binding region of 1-beta-Hb. ORF3a altered the shape of beta-Hb via the Peptidase S6 domain. As a result, heme was lost from 1-beta-Hb. ORF3a IsdA was responsible for capturing shed heme and transporting it to the region of ORF3a IsdC. Then, via the HemS domain, ORF3a dissociated heme into iron and porphyrin. ORF3a cleaved the tetrapyrrole ring of heme via the HemS or ABM domain in the presence of NADPH-cytochrome P450 reductase to liberate iron. Due to the Isd domains of ORF3a, hemoglobin attacked by ORF3a might preserve the majority of its natural domain while exhibiting reduced oxygen transport function. ## S protein efficiently transported oxygen molecules through hemocyanin activity Hemocyanin is commonly found in hexamers and has three copper-binding tyrosinase sites. Because SARs-COV-2 S protein could form hexamers during membrane fusion, we examined whether S protein possesses hemocyanin activity. We obtained the hemocyanin-related sequences from the Uniprot database and used the local MEME tool to compare them to the S protein individually. Finally, we will combine the searched motifs according to conserved domains. S hemocyanin-related domains are listed in Table 10. According to Table 10, the S protein contains the Hemocyanin protein's C-terminal (Hemocyanin_C), active (Hemocyanin_M), and N-terminal (Hemocyanin_N) domains. The S protein contains three copper-binding tyrosinase regions (Tyrosinase_Cu-bd A-C). Tyrosinase_Cu-bd A-C are found in the active area of the Hemocyanin protein, Hemocyanin_M domains. Tyrosinase_Cu-bd A resides in Hemocyanin_M E. In contrast, Tyrosinase_Cu-bd B-C resides in Hemocyanin_M F. Hemocyanin_C and Hemocyanin_M are found in the S protein's N to C terminus. Hemocyanin_N is only found at the S protein's C-terminus. The schematic diagram in Figure 2 illustrates the three-dimensional placements of the domains of Hemocyanin_M, Hemocyanin_N, and Tyrosinase_Cu-bd. Hemocyanin_N A possesses phenoloxidase activity, but it is located in the transmembrane area. It is consistent with phenol oxidase being isolated from phenolic substrates. Hemocyanin_N also serves as the binding site for the enzyme Tyrosinase_Cu-bd B-C. Tyrosinase_Cu-bd A is located on the exterior of the viral membrane. When Tyrosinase_Cu-bd A binds to Cu, it may initiate a chain reaction, causing a conformational shift in S to expose the Hemocyanin_N domain ThenTyrosinase Cu-bd B-C bind to copper. After Tyrosinase Cu-bd A-C bind copper, S protein exhibits hemocyanin activity. ## Functional consequences of the Delta and Omicron mutations on the ORF3a and S proteins Delta (B1.617.2) and Omicron (B.1.1.529) mutants are SARs-COV-2 viruses that spread globally starting in 2021. All relevant variants had significantly higher viral loads than the wild type, with Omicron's average viral loads being many times that of Delta (91). Individuals infected with Omicron exhibit only subtle symptoms. Omicron mutants have a more significant number of mutation sites and a greater capacity for immune evasion. We obtained the ORF3a and S proteins from the NCBI database for the Delta and Omicron mutations. Then, the effect of mutations on the ORF3a and S proteins' attack functions was compared. Mutations in the S protein alleviated symptoms of respiratory distress in patients. We used the local MEME tool to match the S proteins of the Delta and Omicron mutants to hemocyanin-related sequences. The discovered motifs were integrated according to domains, and the resulting search results are displayed in Table 11. The S proteins of Delta and Omicron mutants contain Hemocyanin_C, Hemocyanin_M, Hemocyanin_N, and Tyrosinase_Cu-bd domains, as shown in Table 11. The hemocyanin activity-related domains of the S proteins from SARs2, Delta, and Omicro were mapped using the local version of the IBS tool (Fig. 3). As illustrated in Figure 3, the S proteins of SARs2, Delta, and Omicro include Hemocyanin C domains in both S1 and S2. As shown in Figure 3, the S protein of SARs2 and Omicro contains the Hemocyanin_N domain. S1 includes the Hemocyanin_N domain of Delta's S protein. As illustrated in Figure 3, the S proteins of SARs2, Delta, and Omicro include Hemocyanin_M domains in both S1 and S2. However, the three viruses had dramatically varied distributions of Tyrosinase_Cu-bd domains. The S protein of Omicro contains only two Tyrosinase_Cu-bd domains. Omicro S Tyrosinase_Cu-bd B is positioned within the Hemocyanin_M domain, but Omicro S Tyrosinase_Cu-bd A is not. SARs2 contains the Tyrosinase_Cu-bd domains entirely within the N-terminal Hemocyanin_M domain. Delta has the Tyrosinase_Cu-bd domains in C-and N-terminal Hemocyanin _M domains. The above distributions suggest that the S of SARs2 S, the Delta S variant, represents the cell agglutination and immune interference activity of the entire sequence. However, since Omicro's S lacks the tyrosinase_Cu-bd domain, cell agglutination and phenoloxidase activity cannot be fully achieved. On the S sequence, the Delta variant exhibits phenoloxidase activity. The copper-binding site of Delta S is highly exposed and readily possesses phenoloxidase activity. SARs2 S has a copper-binding site in the membrane fusion domain, which requires full exposure of SARs S Tyrosinase_Cu-bd A-C and binding to copper to function as a phenoloxidase. Delta S, and Omicro S all lack or lose their phenoloxidase activity. Since SARs2 S, Delta S, and Omicro S all have tyrosinase Cu binding sites, all three S proteins act as oxygen carriers. Due to the easily accessible copper binding site of Delta S1, Delta S has a high oxygen transport capacity. This may explain why Delta patients have higher viral loads and lower levels of respiratory distress than SARs2 patients. Omicro S developed into a hemocyanin-like protein capable of carrying oxygen molecules but lacking immunological agglutination and phenoloxidase functions. This may explain why Omicro patients experience fewer severe symptoms and no apparent respiratory distress. ORF3a mutations had no effect on oxygen supply with reduced hemoglobin. We drew the ORF3a's mutation sites of the hemoglobin attacked-associated domain for SARs2, Delta, and Omicro using a local version of the IBS program (Figure 4). The haeme_bd, NEAT, peptidase S6, IsdA, IsdC and IsdH, ABM, and HemS domains of the SARs2 ORF3a protein are all located within the heme-binding "CWKCR" or "CYDYC" motif, as shown in Figure 4. It suggests that these domains are involved in the degradation, capture, and breakdown of hemoglobin. Omicro ORF3a does not contain mutations. Delta ORF3a contains two single site mutation 26 and 275. Their mutations are S->L and L->F. Notably, the two Delta ORF3a mutation sites are not located in the Haem_bd, NEAT, peptidase S6, IsdA, IsdC, and IsdH, ABM or HemS domains described above. Therefore, these two point mutations are unlikely to affect ORF3a's hemoglobin attacked, heme capture, or degrade functions. ## S and E protein bind porphyrin to cause high viral infection The furtherevolution of the novel coronavirus also displays some paradoxical characteristics. The current theory reveals that the novel coronavirus binds to the human ACE2 receptor through a spike protein. The novel coronavirus enters human cells in the form of phagocytosis. The novel coronavirus pneumonia is highly contagious. What causes the high infectivity of the novel coronavirus? The structural proteins S, M, E, and N have the production and binding domains of porphyrin, according to the research.We believe that in addition to the invasive method of spike-ACE2, it should maintain the original invasive pattern. Medical workers have detected the novel coronavirus from urine, saliva, feces, and blood sincethe virus can live in body fluids. In such media, the porphyrin is a prevalent substance, and porphyrin compounds are a class of nitrogen-containing polymers. Existing studies have shown that they have a strong ability to locate and penetrate cell membranes. Therefore, the novel coronavirus may also directly penetrate the human cell membrane through linking porphyrin. ## Higher hemoglobin causes higher morbidity The novel coronavirus pneumonia might be closely related to abnormal hemoglobin metabolism in humans. The number of hemoglobin is a significant blood biochemical indicator, and the content varies with genders. The number of normal men is significantly higher than that of normal women, which might also be a reason why men are more likely to be infected with the novel coronavirus pneumonia than women. Besides, most patients with the novel coronavirus pneumonia are themiddle-aged and older adults, whilemany of these patients have underlying diseases such as diabetes. Diabetic patients have higher glycated hemoglobin which is deoxyhemoglobin and is also a combination of hemoglobin and blood glucose, which is another reason for the high infection rate for the elder people. This present study has confirmed that ORF3a could coordinately attack the heme on the beta chain of hemoglobin. Both oxygenated hemoglobin and deoxygenated hemoglobin are attacked, but the latter is more attacked by the virus. During the attack, the positions of ORF3a are slightly different, which shows that the higher the hemoglobin content, the higher the risk of disease. However, it is not sure that the disease rate incited by abnormal hemoglobin (structural) is relatively low. The hemoglobin of patients and rehabilitees could be detected for further research and treatment. ## Interfering with the normal heme anabolic pathway This article held that the virus directly interfered with the assembly of human hemoglobin. The main reason was that the normal heme was too low. Heme joins in critical biological activities such as regulation of gene expression and protein translation, and the porphyrin is an essential material for the synthesis of the heme. As the existing traces show there is too much free iron in the body of critically ill patients, it could be that the virus-producing molecule competes with iron for the porphyrin, inhibiting the heme anabolic pathway and causing symptoms in humans. It is not clear whether the spatial molecular structure of the heme and porphyrins in patients with porphyria is the same as that in healthy people. If there is an abnormal structure, it is unobvious whether this porphyrin can bind to a viral protein to form a complex, or whether a viral protein can attack this heme. It could be proved by clinical and experimental research. ## Novel coronavirus has a strong carcinogenicity Numerous published studies demonstrate that after cancer patients get COVID-19, their overall risk of cancer deterioration increases. Numerous clinical studies have established that COVID-19 patients exhibit symptoms consistent with oxidative stress damage. Under conditions of severe oxidative stress, the patient's ROS control mechanism becomes disorganized. The dysregulation of reactive oxygen species (ROS) is a critical element in carcinogenesis. The lungs of the deceased from COVID-19 were mucus-filled, and the deteriorated lung tissue resembled adenocarcinoma-like alterations. We discovered that proteins such as S have a melanoma domain during our search for antigen and membrane fusion domains. Many malignancies, including lung cancer, have pathogenic proteins that contain melanoma domains. Besides, we were conducting computer research and discovered that proteins such as S contain p53 domains. Normal cells could develop cancer cells when the P53 protein was mutated. We believe that the SARS-COV-2 virus is highly carcinogenic based on these comprehensive characteristics. ## Conclusion Investigating the factors underlying decreased respiratory function in COVID-19 patients may aid in the saving of many lives. The conserved domain search method was used to investigate the biological roles of certain SARs-COV-2 proteins. The findings indicate that the SARS-COV-2 virus proteins contains regions capable of binding porphyrin and synthesizing heme. S and ORF3a have the ability to bind to hemoglobin. ORF3a's Arg134 and E's Cys44 have heme-iron binding sites, respectively. ORF3a contains Haem bd, NEAT (IsdA, IsdC, and IsdH), Peptidase S6, ABM, and HemS domains. All are found across the "CWKCR" heme-binding site. It indicates that the ORF3a protein can bind to the hemoglobin 1-beta chain and collect and degrade heme into iron and porphyrin. The S protein exhibits hemocyanin-like activity and can transport oxygen molecules via copper binding. The S is triggered by binding to copper oxygen, producing melanin and activating the human immune response through phenoloxidase domains. In patients with COVID-19, the ORF3a protein's action to attack hemoglobin and degrade heme caused respiratory tissues and organs damage. It resulted in a range of respiratory distress and coagulation symptoms in patients, as well as a disruption of normal heme metabolism. However, the oxygen-carrying activity of the S protein's hemocyanin-like domain enhanced the patient's respiration function. The S protein's phenoloxidase function was decreased or missing in the Delta and Omicron viriant, resulting in a weaker immunological response and greater immune escape. However, the ability of ORF3a to target hemoglobin was not diminished. The mutant virus utilized the S protein's high-efficiency oxygen transportability to reduce respiratory distress symptoms in patients.

chemsum

{"title": "COVID-19: Attacks the 1-beta Chain of Hemoglobin to Disrupt Respiratory Function and Escape Immunity", "journal": "ChemRxiv"}

investigation_of_acetyl_migrations_in_furanosides

## Abstract: Standard reaction conditions for the desilylation of acetylated furanoside (riboside, arabinoside and xyloside) derivatives facilitate acyl migration. Conditions which favour intramolecular and intermolecular mechanisms have been identified with intermolecular transesterifications taking place under mild basic conditions when intramolecular orthoester formations are disfavoured. In acetyl ribosides, acyl migration could be prevented when desilylation was catalysed by cerium ammonium nitrate. ## Investigation of acetyl migrations in furanosides O. P. Chevallier and M. E. Migaud * ## Introduction Over the years, a number of methods aimed at achieving chemoselective acylation of carbohydrates have been developed. In addition to the challenges encountered in the selective acylation of compounds incorporating multiple hydroxyl groups, subsequent isomerisations of partially acylated products often arise, in particular in the case of acetyl groups. Such migrations take place under varied reaction conditions, often accounting for much of yield loss and are often proposed to involve the formation of an orthoester intermediate between adjacent or remote alcohol and ester groups. We aim to prepare analogues of 1''-O-acetyl adenosine diphosphate ribose (AcO-ADPR) incorporating modified acetylated ribosides, xylosides and arabinosides in place of the ribose moiety. The synthesis of such compounds requires access to Scheme 1: Acetyl migration products upon TBAF/THF treatment well-characterised acetylated xylose, arabinose and ribose synthetic precursors and while few acetylated furanosides have been reported in the literature, those that have, have often been obtained in modest yields and reliable compound characterisations have been limited. Whilst attempting the synthesis of 1-O-acetyl, 2,3-isopropylidene riboside 1a from the 5-O-silylated precursor 1 (Scheme 1) using mild desilylating conditions, three compounds were isolated, evidently products of acyl migration processes. In order to synthesise acetylated ADPR analogues, unequivocally, the free C5-hydroxyl derivatives of the monoacetylated furanosides must be obtained without rearrangement upon protecting group manipulation. Consequently, it was decided to carefully characterise all possible rearrangement products occurring in acetylated furanoside precursors. To the best of our knowledge, acyl migrations occurring under such mild reaction conditions in furanosides has not been investigated in contrast with the extensive work that has been carried out on pyranosides. Therefore, an investigation on the mechanisms, reaction conditions and structural features, which control this type of chemistry in furanosides was initiated and conditions under which such migration could be minimised were sought. To this end, a series of acetylated furanosides (1-5) were synthesised. Product distribution due to acetyl migration was carefully examined when these compounds were treated under neutral or mildly basic/acidic reaction conditions that promote the unmasking of a silylated alcohol. ## Results and discussion All acetylated furanosides were prepared according to synthetic routes which favoured divergent syntheses (see additional information for full experimental data). Riboside 1 was synthesised from ribose in 47% overall yield via a chemoselective silylation of the C-5 primary hydroxyl group of 2,3-O-isopropylidene ribose (Scheme 2). The acetylated xyloside 2 and riboside 3 were both synthesised from xylose in 46% and 45% overall yield, respectively. The methyl arabinoside 4 was synthesised from arabinose (Scheme 4) in 17% overall yield in 12 steps; a long synthetic sequence which was dictated by the preferred pyranoside form of arabinose in solution. The successful synthesis required three successive protecting group interconversions to allow for the base catalysed benzylation of the C-3 hydroxyl and the acid catalysed acetonide removal prior to acetylation at the C-2 position of the methyl arabinoside. While an acid-catalysed benzylation would have significantly shorten the route, it was found to be unsuccessful when carried out on 5-O-TBDPSprotected 1,2-O-isopropylidene arabinoside. Riboside 5 TBDPSCl, imidazole, DMAP, DMF (94%); c) HgO, HgCl 2 , acetone (80%); d) 2,2-dimethoxypropane, p-toluenesulfonic acid, acetone (85%); e) BnBr, NaH, THF (97%); f) TBAF, THF (87%); g) PivCl, pyridine/DCM, DMAP (85%); h) TFA/H 2 O (8/2) (87%); i) MeOH, H 2 SO 4 (80%); j) MeONa/MeOH (76%); k) TBDMSCl, pyridine/DCM, DMAP (85%); l) Ac 2 O, pyridine, DMAP (98%). During the TBAF-catalysed deprotection of the silyl ether of β-1-O-acetyl riboside 1, a mixture of three different ribosyl derivatives (1b-d, Scheme 1) was obtained. The components of this mixture were separated by chromatography and individually identified by NMR and MS analyses. The main product 1b, for which the anomeric proton is shielded by -0.7 ppm and the diastereotopic C-5 protons de-shielded by 0.46 and 0.53 ppm when compared with 1, arises by an acetyl migration from the β-anomer C1-position to the C5-primary alcohol, possibly intramolecularly via an orthoester intermediate. Isolation of the bis-acetylated riboside 1c and of the riboside 1d indicated that an intermolecular trans-esterification was also taking place. To identify the reaction conditions and the structural features of the acetylated furanosides facilitating acetyl migration upon protecting group manipulation, the various protected carbohydrates (1-5) were treated with tetrabutyl ammonium fluoride (TBAF) in THF, with KF in the presence of 18-crown-6 in benzene, with ceric ammonium nitrate IV (CAN) in aqueous acetonitrile and with H 2 -Pd/C in THF. No attempts were made to examine standard protic acid-catalysed desilylations such as AcOH in H 2 O/THF or TFA/H 2 O in DCM since we observed rapid degradation under such conditions of the synthetic intermediates of compounds 1-5, all incorporating either an acetonide or a methyl ketal moiety. The reactions were stopped as soon as complete disappearance of the starting material had occurred to minimise rearrangements subsequent to the deprotection step. The ratio of products and migration products were calculated based on the quantities of purified material isolated by chromatography, as 1 H-NMRs of the crude reaction mixtures were rarely sufficient to establish such ratios accurately. The chemical shifts and the splitting patterns along with nOe experiments were sufficiently clear to allow accurate compound identification. No isomerisation was detected for any of the isolated compounds when stored in an organic solvent such as hexane, ethyl acetate or CDCl 3 for up to 12 hrs. The chemical shifts of the 1 H-and 13 C-NMR of the desilylation products for compound 1 (Table 1 and Table 2) are representative of the chemical pattern for all investigated furanoside derivatives. When derivatives 1, 2 and 3 were reacted with a reagent source of fluoride, intramolecular acetyl migration was detected for compound 1 and 2 (1b and 2b) but not for compound 3 (Table 3). The facile intramolecular migration observed for 1 and 2 involves an orthoester intermediate which cannot form readily from riboside 3. This latter observation is in contrast with the acetyl migration often reported for the glucoside series and for which the orthoester intermediate (chair conformation) can form easily under mildly basic or acidic conditions. In riboside 3, the rigid conformation of the furanose ring prevents the formation of this orthoacetate. Similarly, no migration product 4b (Table 4) was detected when arabinoside 4 was treated with TBAF in THF while as expected, the acetyl migration between the C2 and C3 positions in the furanoside 5 occurred readily under neutral anhydrous conditions (Table 4). In riboside 5, the five membered cyclic orthoester intermediate yields a 50-50 ratio of 2-O and 3-O-acetylated products (5a and 5b, Table 4). The formation of such an orthoester did not occur in compounds 3, 4 or 5 under fluoride-catalysed desilylation conditions and as a result no intramolecular migration products were observed. However, for all sugars, a small proportion of bis-acetylated (compounds c) and non-acetylated (compounds d) compounds have also been isolated (Table 3 and Table 4). The proportion of intermolecular transesterification and de-esterification products increased when anhydrous KF conditions were used instead of commercial TBAF in THF (Table 3, condition C). Similarly, longer reaction time of the fluoridecatalysed desilylation reactions resulted in an increase in the formation of c and d type derivatives. In addition, the formation of the c and d furanoside derivatives took place, as easily when the liberated hydroxyl group was either syn or anti to the acetyl moiety and whether either five or six atoms separated the oxygen atoms of the silyloxy and the acetyl moieties. Such direct transesterification is intermolecular and while it has been observed in nucleoside-phosphoramidite and glycoside chemistry, this alkoxide-promoted inter-molecular acetyl migration process has been overlooked in furanosides. The isolated quantities of 1c and 1d decreased by a factor of two upon dilution (1:10) when 1 was reacted with TBAF, while 1b formation remained rapid with no detection of 1a formation. Similarly, complete transesterification took place when riboside 6 was treated with acetyl riboside 1 under basic conditions (Scheme 6). Acetyl migration in riboside 1 was prevented when CAN was used instead of fluoride-containing reagents, indicating that the seven membered ring orthoester was not favoured under these mildly acidic conditions. This reagent was also found suitable for the silyl-removal in the other riboside 3 and 5 and arabinoside 4 with no trans-esterification occurrence. Unfortunately, the formation of a six-membered ring orthoacetate appears to be facilitated under acid-catalysed reaction conditions and total intramolecular acetyl migration occurred in xyloside 2 in the presence of CAN, showing how finally balanced things are in these systems. In conclusion, to minimise acetyl migration in furanoside derivatives, base-catalysed reactions should be conducted under dilute conditions when the formation of an orthoacetate is disfavoured to minimise intermolecular transesterification. Unfortunately, we have been unable to identify a set of reaction conditions which could minimise the readily occurring intramolecular migration resulting from the formation of a five or six membered ring orthoacetate intermediate. Under such circumstances, migration can only be circumvented by controlling the stereochemistry of the reactive centres and opting for a multistep synthetic approach. However, it can be said that in mild acidic conditions provide better control over intra-and intermolecular acetyl migration than basic conditions. Finally, CAN has been identified as a very useful alternative reagent to TBAF in preventing unwanted base-catalysed acetyl migration in the deprotection of silylated furanosides.

chemsum

{"title": "Investigation of acetyl migrations in furanosides", "journal": "Beilstein"}

hydrogenation_of_unactivated_enamines_to_tertiary_amines:_rhodium_complexes_of_fluorinated_phosphine

## Abstract: In the hydrogenation of sluggish unactivated enamine substrates, Rh complexes of electron-deficient phosphines are demonstrated to be far more reactive catalysts than those derived from triphenylphosphine. These operate at low catalyst loadings (down to 0.01 mol %) and are able to reduce tetrasubstituted enamines. The use of the sustainable and environmentally benign solvent (R)-limonene for the reaction is also reported with the amine isolated by acid extraction. ## Introduction A potentially very direct method to produce tertiary amines is by the hydrogenation of enamines. While the hydrogenations of enamides, bearing coordinating acyl substituents is probably the most developed and studied of all hydrogenation processes, studies on the hydrogenation of unactivated enamines are scarce and several important problems need to be solved. Some time ago, the enantioselective variant was highlighted by several pharma companies as one of the more important aspirational transformations for production of pharmaceuticals . Several examples of highly enantioselective and quite reactive processes have appeared for enamines that are activated by a chelating group, or can potentially isomerise to an NH imine during catalysis . A few papers have appeared with good enantioselectivity for some quite specific enamines, but despite the importance of these contributions, catalyst loadings around 1 mol % are used . Commercial applications generally require catalyst loading below 0.05 mol %. We are not aware of any achiral or chiral homogeneous catalysts that promote these reactions at this substrate/catalyst ratio, so the intrinsic lower reactivity of these substrates needs to be addressed with new catalysts. In a recent study on hydroaminomethylation, i.e., domino hydroformylation-enamine formation-enamine hydrogenation, we noted that the enamine hydrogenation was the slowest reaction in the process, and use of an electron-deficient phosphine sped up the reduction step significantly . DFT calculations revealed that in the hydrogenation of these aldehyde-derived enamines, the final stage of hydrogenation, reductive elimination was the rate-determining step. This is in contrast to nearly all studies on homogeneous hydrogenation of alkenes where oxidative addition, and probably more often migratory insertions are rate-determining and accelerated by electron-rich phosphine ligands. Prior to embarking on a quest for highly active Rh catalysts for enantioselective enamine hydrogenation, we investigated if more commonly encountered enamine substrates are also reduced much faster using Rh complexes of electron-withdrawing phosphines. In this paper, we report how a range of enamines can be successfully hydrogenated in high yield using low levels of rhodium, including some very deactivated enamines that do not hydrogenate using conventional catalysts. ## Results and Discussion The majority of the enamines produced in this study were synthesised from the parent ketones and secondary amines by adapting literature procedures (Scheme 1) . Since isolation of pure enamines is not a completely trivial task, Supporting Information File 1 gives full details for the synthesis and purification of enamines 1a-i. One of the main modifications made to the synthetic procedure is at the end of the reaction, wet diethyl ether was added in order to precipitate all titanium salts (this strategy was previously used after formation of imine bonds using TiCl 4 ) . Enamine 1g was more stable than all other enamines with disubstituted double bond studied here; no hydrolysis was observed in wet chloroform even after 6 hours. It is also worth mentioning that tetrasubstituted enamines are very stable towards hydrolysis. Consequently, enamines 1b and 1c were isolated by acid-basic work-up with purities of over 99% (see Supporting Information File 1 for details). Enamine 1j cannot be prepared using this strategy, and therefore we developed a new branched-selective hydroaminovinylation procedure . Some time ago, this enamine was detected in a product mixture with up to 39% selectivity . A key aspect that prevents better selectivity is that, in general, Rh catalysed hydroformylations of 'alkyl' alkenes of type RCH 2 CH=CH 2 give mainly the linear product . Since we had recently discovered that Rh complexes of the 'BOBPHOS' ligand unexpectedly give unprecedented branched regioselectivity in enantioselective hydroformylation of alkyland arylalkenes , we reconsidered this cyclisation reaction using the new catalyst (Scheme 2). We were pleased to find that the selectivity is increased to 78%. Since the desired product is achiral, there is no need to use enantiopure BOBPHOS for this synthesis. When the reaction was performed on 8 mmol scale, using a BOBPHOS sample made from racemic biphenol derivative 3, the enamine 1j was isolated in an overall yield of 60%. We initially wanted to establish the generality of the previous observation that electron-withdrawing ligands enhance the rate of Rh catalysed hydrogenation relative to more electron donating ligands such as triphenylphosphine. The hydrogenation of enamine 1e at a S/Rh ratio of 250 at 65 °C proceeded at a suitable rate, such that simply measuring conversion at the times given provides a meaningful measure of the relative rates of hydrogenation for Rh catalysts derived from electrondonating and electron-withdrawing ligands. A screen of monodentate ligands was performed for hydrogenation of 1e with catalysts derived from ligands 4-9 (Scheme 3). Compared to triphenylphosphine, more electron-poor ligands, particularly commercially available 7, 8 or also commercial product 9, show faster rates of hydrogenation of 1e. It can be envisaged that other less electron-donating phosphines could also be used to good effect, providing they are stable under the reaction conditions. It is possible that stability is an issue with the strong π-acceptor ligand triphenylphosphite. We note here that an earlier attempt by some of us using chiral phosphites in this type of reaction gave very low conversions to product under these conditions. The ligand electronic effect clearly supports our earlier proposal of the reductive elimination as rate determining step in this process . Using readily available simple ligand 8, combined with [Rh(COD)Cl] 2 , we also studied the hydrogenation of a range of other enamines with [Rh(COD)Cl] 2 /PPh 3 as a control. Table 1 shows very clearly the improved performance of the less strongly donating phosphine ligand for this process. For enamines 1g, 1h and 1i, experiments with much lower catalyst loadings were performed in order to prove that the rate of hydrogenation is faster when 8 is used instead of 4. Of particular note is the hydrogenation of the deactivated enamines 1b and 1c. It is well known that, even without deactivating nitrogen substituents, the hydrogenation of tetrasubstituted alkenes is not generally achieved with Rh catalysts ; Crabtree's catalyst is often used to accomplish this type of task . The ability of this catalyst combination to conduct this type of transformation, as shown in Table 1, entries 6 and 8 are of synthetic value. Trisubstituted enamine 1d (Table 1, entry 9) is slower to reduce than all disubstituted enamines (except entry 1). Enamine 1j (Table 1, entry 31) shows a much faster rate of hydrogenation than 1d (entry 9), presumably due to the fact that there is a ring strain due to the double bond in a 6-membered ring, which is released after the double bond is hydrogenated. 1a does not get hydrogenated with the catalysts studied. It is likely that this is due to the substrate binding to the catalyst via the pyridine nitrogen and deactivating the catalyst. In order to provide support for this, the normally high-yielding hydrogenation of 1d was carried out in the presence of 30 equivalents of pyridine relative to Rh, and the conversion dropped to 10%. Comparing Table 1, entries 12, 14 and 16, it is clear that electron-poor enamines get hydrogenated faster. A possible reason for the enamine 1f being reduced slower than 1g may come from the fact that 1f is a much more stable enamine (see enamine synthesis section, Supporting Information File 1). It is well known that the reductive elimination is sped up with more bulky ligands -i.e., when the bulk around the transition state is larger, this step occurs more readily. Enamines 1h and 1i are more bulky due to their N-benzyl substituents, and therefore are hydrogenated with faster rates. Table 1, entry 29 represents a TON of 4550 mol mol −1 which is, to the best of our knowledge, the highest TON in an enamine hydrogenation reported up to date. We found it convenient to carry out these reactions at 30-60 bar of H 2 gas (in order to compare reactivities of enamines with triphenylphosphine as a ligand). Full conversion is also possible at 5 bar pressure (Table 1, entry 30), but we did not observe product using a balloon of hydrogen (~1 bar). The latter observation contrasts somewhat with the results of reference , when using 1 mol % of a [Rh(diphosphine)(COD)] cation on a disubstituted enamine: complete conversion can be realised in 2-18 hours. It can be assumed that the catalysts used here are less efficient at activating hydrogen relative to more electron-rich metal systems, meaning below a certain pressure threshold, hydrogen activation does not proceed at a sufficient rate. In order to prove that toluene is not the only solvent where an electronic effect holds, a polar protic solvent (MeOH) was chosen (Table 2). The electronic effect still holds in MeOH as a solvent, although it is less pronounced, and the best rate of conversion is found in toluene. Another solvent explored in this study was (R)-limonene. This solvent is now being used as a green alternative to hexane in the cleaning industry and extraction , but barely has been exploited in synthetic chemistry so far . While being an environmentally benign, fairly cheap, waste-derived chemical, it might seem counter-intuitive to use it in hydrogenation since it contains 2 double bonds itself. However, in hydrogenation of enamines, as was shown above, enamine hydrogenation benefits from electron-poor ligands, so the hope was that the enamine hydrogenation would be competitive over limonene hydrogenation. In addition, this solvent enables us to study the relative reactivities of these C=C bonds, as well as possibly giving a greener procedure. Examples of hydrogenation of 1h in limonene as a solvent are shown in Table 3. The results shown in Table 3 suggest that limonene is a promising solvent for this process. As expected, triphenylphosphine shows higher selectivity in hydrogenation of the limonene's disubstituted double bond, and low conversion to amine. Ligand 8 allows good conversion to amine with relatively low amounts of limonene hydrogenated at the least substituted double bond. While this solvent is not completely inert, it is envisaged that the mixture of limonene and dihy- drolimonene (10) would be a perfectly suitable solvent mixture to recycle and reuse. (R)-Limonene is a high boiling solvent, creating a disadvantage for processes where the solvents are removed by evaporation. However, in amine synthesis in general, amines are isolated by extraction into acid, and this was demonstrated here (see Supporting Information File 1). We suggest that (R)-limonene is worth considering as a sustainable, benign solvent for amine synthesis in the future. ## Conclusion Overall, the primary outcome from this study is to demonstrate that highly active Rh catalysts for enamine hydrogenation are a possibility, but they require quite different ligands to those needed for enamide hydrogenation. From a synthetic perspective, large scale reduction processes generally prefer the use of hydrogen gas to any other reductant, since it potentially saves on cost, waste, atom-economy, solvent and water use; the catalysts identified here could be useful in this regard. While it is possible some heterogeneous hydrogenation catalysts could accomplish enamine reductions, the issues with functional group tolerance would be problematic in many cases. From a more general synthetic viewpoint, the use of reagents such as sodium triacetoxyborohydride or sodium cyanoborohydride can be appealing at small scale where the practical issues noted above are not so important. However, the formation of tertiary amines from aryl ketones using hydride reagents has been reported to be problematic . In addition, the hydride reductions, whether carried out as reductive amination or reduction of enamines need stoichiometric acetic acid to promote the formation of the iminium ion that is the substrate reduced in hydride reductions, which might not be compatible with other functional groups. To the best of our knowledge, the hydrogenation of tetrasubstituted enamines has not been carried out before. The use of green, non-toxic and renewable solvent (R)limonene is introduced here as a potentially promising solvent for amine synthesis. This solvent could prove a particularly useful green solvent for any reaction that involved an aqueous/ organic work-up as purification step, particularly if catalysts could be recycled, although that is likely to be challenging in moisture sensitive catalytic hydrogenation chemistry. The ligand electronic effects seem counter intuitive at first glance, but they support the finding by DFT calculations that enamine hydrogenation has a different rate determining step to most other alkene hydrogenations, and show that these observations are a general phenomenon of synthetic use, since some of the enamines studied here are rather unreactive using normal catalysts and/or in reductions using hydride reagents. Research on the creation of enantioselective enamine hydrogenation catalysts that can operate at industrially acceptable catalysts loadings may well benefit from chiral π-acceptor phosphines as ligands and this is being actively researched in our laboratory.

chemsum

{"title": "Hydrogenation of unactivated enamines to tertiary amines: rhodium complexes of fluorinated phosphines give marked improvements in catalytic activity", "journal": "Beilstein"}

what_accounts_for_the_different_functions_in_photolyases_and_cryptochromes:_a_computational_study_of

## Abstract: Photolyases (PL) and cryptochromes (CRY) are light-sensitive flavoproteins, respectively involved in DNA repair and signal transduction. Their activation is triggered by an electron transfer process, which partially or fully reduces the photo-activated FAD cofactor. The full reduction additionally requires a proton transfer to the isoalloxazine ring. In plant CRY, an efficient proton transfer occurs within several µs, enabled by a conserved aspartate working as proton donor, whereas in E. coli PL a proton transfer happens at a second timescale without any obvious proton donor, indicating the presence of a long-range proton transfer pathway. Unexpectedly, the insertion of an aspartate as proton donor in a suitable position for proton transfer in PL does not lead to a transfer process similar to plant CRY, but even prevents the formation of a protonated FAD. In the present work, we combine several computational methods to explain the divergent behaviour of closely related proteins towards the FAD protonation. Free energy landscapes of the proton transfer mechanisms are obtained from state-of-the-art biased QM/MM simulations. Comparison of the computed free energy barriers with kinetic and spectroscopic measurements allows a microscopic interpretation of the different proton transfer mechanisms and rate constants. Our results further illustrate the fine tuning of the electrostatic FAD environment and the adaptability of the FAD pocket to ensure the divergent functions of the members of the PL-CRY family. ## Introduction Photolyases (PL) and cryptochromes (CRY) constitute a family of flavoproteins, the photolyase/cryptochrome family (PCF), which is present in all kingdoms of life ranging from bacteria to mammals. The structures of both PL and CRY exhibit highly conserved elements as an electron transfer (ET) pathway consisting of three tryptophans, called Trp-triad, and a non-covalently bound flavin adenine dinucleotide cofactor (FAD). The photo-reduction of FAD by a cascade of electron transfer reactions (ET) generates the active form of these proteins and enables their biological functions. 1,2 The major role of CRY is to participate in the regulation of diverse biological responses as a signalling molecule. CRY contributes to the entrainment of the circadian rhythm 3 , the space orientation by interaction with the magnetic field of the earth 1,2,4,5 which was especially shown for a CRY4 of birds 6 , and the regulation growth or flowering state in plants. 7,8 Therefore, the formation of a radical pair is required which is initiated by an ET reducing FADox to FAD •-. Additionally, further proton transfer (PT) reactions can form FADH • in the presence of proton donors. Hence, CRY offers a FAD pocket with an environment supporting FAD to switch between redox states. The main task of PL is the repair of photo-induced DNA damages by an ET between FADHand the DNA lesion. To fulfil their biological task, a stabilized FADHis generated by a photoactivation of FADH • and further ET along Trp-triad. Consistently, the protonated FAD states (FADH • semiquinone or FADHhydroquinone) prevail in PL in vivo 1 , which both need to be stabilized by the environment of the FAD pocket. This FAD pocket is conserved in most PCF proteins to a large extent 9 , however, it is obviously supposed to offer different interaction patterns to enable the different functions, e.g. in CRY, a FAD •radical needs to be formed while in PL a FADHneeds to be stabilized. therefore, one specific amino acid in a position close to the N5 of FAD differs between members of the PCF: in most plant CRY, e.g. CRY I from Arabidopsis thaliana (CRYI), an aspartate (Asp) is found in this position 10,11 , but a non-titratable residue is often present, such as a cysteine (Cys) in entomic CRY. 12 In CRY-DASH of a cyanobacterium 13 , which shows a high similarity in structure and function to the PL, an asparagine (Asn) is found at this position This Asn is also present in PL, e.g. in PL from E. coli 14 and A. nidulans. 15 The impact of this amino acid has been studied by experimental site-directed mutagenesis, showing that it affects the behaviour of the proteins decisively: in CRYI, the Asp acts as proton donor to form FADH • from a photo-reduced semiquinone FAD •-, whereas a mutation of Asp into Cys, inspired by the entomic CRY, prevents FAD protonation. 16 The mutation into Asn leads to the formation of FADH • despite the lack of an obvious proton donor such as Asp which indicates the presence of an alternative PT pathway. Additionally, in the Asp to Asn mutant, a PL-like DNA repair activity was measured, which in fact requires the stabilization of FADH -. 10 Conversely, the presence of Cys in entomic CRY, like in Drosophila melanogaster CRY, inhibits the protonation of FAD •-. 17,18 In this CRY, the mutation of the Cys into Asp, albeit offering a suitable proton donor to FAD, does not lead to the formation of FADH • . 19 On the contrary, a mutation of Cys into Asn shows a stabilization of protonated FADH • , which does not affect the photoreceptor function. 19 In E. coli and A. nidulans PL, as observed in the entomic CRY, the mutation of Asn into Asp does not enable a PT to form FADH • . 20,21 After photoreduction, the E. coli PL Asn to Asp mutant (PL-N378D) shows a slightly different UV-VIS absorption spectrum compared to the spectrum for FAD •-, characteristic of the presence of a different FAD state, called FAD x , and the absence of FADH • . It was hypothesized that this state indicates a strong hydrogen bond network between Asp and FAD N5 or a shortly living FADH • . So, the presence of Asp blocks the stabilization of FADH • and furthermore inhibits any other PT pathway. Combined spectroscopic and computational studies on E. coli PL (WT-PL) showed that a strong hydrogen bond between FADHand Asn378 stabilizes the FADHstate, which is essential for the repair activity. 22,23 Further experimental studies investigated the replacement of Asn into Ser in the E. coli PL and a drop of the in vitro DNA repair activity from 50 % for WT to less than 1 % is observed. 24 This underlines the dominant role of Asn for DNA repair, likely by stabilizing the FADH • and FADHstates over other redox states. In summary, the reported observations highlight the fine tuning of the individual FAD pocket to stabilize the active form of the protein. 25 It appears that the nature of the residue interacting with FAD N5, which may be Asp, Asn or Cys, is strongly related to the biological most relevant FAD protonation state in WT, thus to the biological role of the respective protein. However, a mutation does not necessarily result in a functional conversion of the mutants. Consequently, other structural or dynamical parameters, like the formation of an PT path, cainfluence the protonation/redox state of FAD. To explore this pathway, Müller et al. studied the protonation of FAD in WT-PL in vitro by preparing a protein which contained FAD in its fully oxidized state, FADox. The timescale of the ET to form FAD •was below 300 ps and that of the subsequent PT was 4 s (see mechanism in Figure 1). 20 Since Asn cannot protonate FAD directly, a pathway has to exist, which seems not to involve other acidic residues in the FAD pocket. 13 Damiani et al. suggested that water molecules can enter the FAD pocket and form a PT pathway from the bulk water to FAD. 21 The formation of the FAD x state occurs in 3.3 µs in PL-N378D. Compared to that, spectroscopic studies in CRYI show that the photo-reduction of FADox occurs within 31 ps and the subsequent PT from Asp396 to FAD •occurs in 1.7 µs while the backward transfer is ten times slower (690 µs). 26,27 So, the CRYI pocket allows a faster switch between protonated and deprotonated FAD than the PL pocket. Computational chemistry methods provide a consistent and comparatively cheap approach to investigate the effect of differences between the FAD pockets from CRY or PL-WT and mutants at a molecular level. In our previous study 28 , we computed rate constants for both ET and PT 28 processes in CRYI in agreement with the experimental results. The aim of the current study is to resolve the mechanism and energetics of the PT processes taking place in PL-WT and PL-N378D, and compare to those in CRYI (see Figure 1). The combination of QM/MM free energy calculations of the PT and classical MD simulations provide a complete picture of the interplay between the FAD protonation and the structural and dynamical behaviour of the FAD binding pocket. ## Molecular Dynamics Simulations. The starting structures for MD simulations were adapted from the X-ray crystal structures of E. coli PL (PDB ID 1DNP) 33 and CRY I At. (PDB ID 1U3D) 34 by Deisenhofer and coworkers. The protonation states of relevant amino-acid side chains were established on the basis of pKa calculations performed with PROPKA3.1. 35,36 Since no crystal structure of the N378D mutant of E. coli PL is available, a starting structure of the mutant was created by modifying the WT-PL structure. In PL-N378D, two different starting structures where created containing different protonation states of Glu363. This residue has a calculated pKa value of 6, which suggests the coexistence of both the protonated or the deprotonated states. The Glu363 side chain may interact with the carboxyl group of Asp378 and the surrounding solvent molecules. Hence, different interaction patterns with the carboxyl group of Asp378 would result from these states: a protonated Glu363 is able to form a strong hydrogen bond with negatively charged Asp378, and a weaker one with neutral Asp378; a deprotonated Glu363, however, leads to a destabilization of the deprotonated Asp378 by electrostatic repulsion. All of the simulations were performed with the AMBER-SB99-ILDN force field 37,38 using GROMACS 5.0.4. 39,40 The cofactors FAD and MTHF were parametrized with the xLeap module of AmberTools 41 employing the general Amber force field (GAFF). 37,42 The atomic charges of FAD, FAD •and FADH • were fitted on the electrostatic potential (RESP) 43,44 obtained at the HF/6-31G* 45,46 level of theory with GAUSSIAN09. 47 The protein was placed in a periodic box sized 96×94×121 A³ for WT-PL and PL-N378D and 98×98×98 ų for CRYI. The box was filled with TIP3P water molecules to obtain a density of 1000 kg•m -3 , and an appropriate number of water molecules were substituted by sodium ions to achieve electroneutrality; no extra salt was added. These systems were equilibrated by means of a protocol consisting of a series of energy minimization, NVT (1 ns) and NPT (1 ns) simulations. The eventual, production simulations used the Nosé-Hoover thermostat 48 and the Parrinello-Rahman barostat 49 to maintain a temperature of 300 K and a pressure of 1 bar, respectively. The simulations employed a leap-frog integrator with a time step of 2 fs. We performed MD simulations containing FADox (100ns) for the PL-WT, PL-N378D and CRYI. Then, we used the last geometry from these simulations and changed the parameters to FAD •to start 200 ns or 50 free MD simulations for PL-N378D, CRYI, or PL-WT respectively. ## Coulombic energy calculations in PL-N378D and CRYI The GROMACS tool gmx energy was used to evaluate the electrostatic interactions between the groups of atoms defined as follows: side chain of Asp378, isoalloxazine ring of FAD, solvent and protein (excluding the side chain of Asp378). The electrostatic interactions between each pair of the groups were calculated using a cut-of radius of 10 along the last 20 ns of the MD trajectories for each protein for the FAD •state. The 20 ns trajectory was used to generate each 2 ns a snapshot used as a starting structure for individual MD simulation containing FADH • and deprotonated Asp378/394, resulting in 10 individual MD simulations containing the force field parameters FADH • and deprotonated Asp378/394. Taken together, this leads to a trajectory of an accumulated length of 25 ns, which reflects the direct structural response of the binding pocket upon the formation of the PT product. We used the structures of the individual trajectories to calculate the corresponding coulombic energies, ECoul, between the Asp-FAD complex and the environment. The averaged ECoul energy of each protonation state is compared according to eq. 1. ## Biased sampling methods Free energies of the various processes studied, structural rearrangements and PT reactions, were obtained with umbrella sampling (US) or metadynamics 50 in its well-tempered form 51 , whichever proved more appropriate in pilot simulations. Free energy was obtained by using the weighted histogram analysis method (WHAM) 52,53 in US simulations. All of the biased sampling simulations were performed with GROMACS interfaced with the Plumed 2.0.1 software. 54,55 Classical WT metadynamics simulations of the rotation of N378 side chain in PL-WT The free energies of the rotation along the dihedral angle  ≡ N-C-C-C of N378 side chain were obtained with classical well-tempered metadynamic simulations extended to 50 ns. Gaussian-shaped hills of width  = 20 ° were added every 500 ps, their initial height was 0.29 kcal•mol -1 , and the bias factor was 6. ## Classical umbrella sampling simulations of the formation of a water wire in PL-WT The free energy profiles of the penetration of water molecules into the FAD binding pocket of PL-WT were generated with classical US simulations. The reaction coordinate for the entrance of the first water molecule was the distance between the center of mass of the N5 and O4 atoms of FAD and the center of mass of a selected bulked water, being 6.5 away. The interval of 6.5 was divided into 29 windows. Harmonic biasing potentials with a force constant of 2388.5 kcal•mol -1 • -2 were applied, and each window was simulated for 50 ns. ## QM/MM Proton transfer reactions were simulated with the quantum chemistry-molecular mechanics approach (QM/MM) employing the approximate DFT method, the third-order tight-binding density theory functional (DFTB3) for the QM part. 56 The general parametrization of DFTB3 for organic and biomolecules, 3OB, which exhibits a sufficient accuracy for biophysical application. 57 In general, DFTB underestimates the barrier for PT because its parameterization is based on a GGA PBE functional. 58 Nevertheless, DFTB3/3OB 57 provides a good description of proton affinities, which are relevant for the simulation of PT reactions, as well as PT barriers. The QM/MM simulations were performed with our recent QM/MM implementation of DFTB3 in Gromacs combined with Plumed. 59 This made it possible to perform US and metadynamics simulations with a QM/MM Hamiltonian. ## QM/MM umbrella sampling simulations of proton transfer In PL-WT, the free energy profile of the PT reaction along the water wire in the protein pocket was obtained with an US-QM/MM simulation. The QM region consisted of the isoalloxazine ring of FAD, the side chain of Glu106 and five water molecules, and it was described with DFTB3/3OB. The progress of the PT process was captured with the center of excess charge reaction coordinate ζ, see below for details. 60 The distance between proton acceptor and donor was divided into 57 windows. Harmonic biasing potentials with a force constant of 2388.5 kcal•mol -1 • -2 were applied, and each simulation was extended to 300 ps. The reaction coordinate for the long-range PT in PL-WT involves the H-N5 of FAD •-, the atoms of the water wire and the external proton donor, so we used the modified centre of excess charge (mCEC) by König et al. 60 ## 𝝃 = ∑ 𝒓 𝑯 Where 𝒓 𝑯 𝒊 are the coordinates of all of the hydrogen atoms possibly taking part in the PT process; 𝒓 𝑿 𝐽 are the coordinates of the proton acceptors -donor, final acceptor as well as all of the relays, and 𝑤 𝑋 𝑗 are the numbers of protons bound to each acceptor in the least protonated state (e.g., w=2 for a water molecule). The last term is a correction of coordinate, which runs over all of the hydrogens and proton acceptors and it involves a switching function 𝒇 𝒔𝒘 (𝒅 𝑯 𝒊 ,𝑿 𝒋 ) to decide whether each couple of atoms is connected by a bond: Considered were these values of empirical parameters: rsw=1.25 and dsw=0.04 . The coordinate ξ is a vector quantity that expresses the position of proton being transferred. A scalar reaction coordinate is obtained from ξ as where 𝑑 𝝃,𝑫 and 𝑑 𝝃,𝑨 are the distances from the point ξ to the initial proton donor and to the final proton acceptor, respectively. With such a collective variable, the PT reaction proceeds from ζ=0 for the reactant to ζ=1 for the acceptor. In PL-N378D, the free energy profile of the PT reaction between N5 of the FAD and the Asp378 was also obtained with a US-QM/MM simulation. The QM region contained the side chain of Asp378 and the isoalloxazine core of FAD. The difference between the H-O distance of Asp378 and the H-N5 distance of FAD •-, dO-H-N, was considered as a reaction coordinate for the PT. These simulations were performed for the three protein variants involving a protonated Glu363, a deprotonated Glu363 or a methionine in position 363. Harmonic biasing potentials with a force constant of 2388.5 kcal•mol -1 • -2 were applied, and the simulation of each of the 20 windows was extended to 300 ps. The shown data of the PT in CRYI is based on the previous study 28 , in which just the distance between the proton and the N5 of FAD, dH-N, was considered as reaction coordinate. Rate constants were estimated from the energy barrier heights in the free energy profiles by applying the transition state theory, using the relation 61 ## 𝒌 where T is the temperature, ΔG ‡ the barrier height, and kb, h are Boltzmann's and Planck's constants, respectively. The attempt frequency ν of the transition was estimated as with kH-N ≈ 537 kcal•mol -1 2 being the harmonic force constant of the N5-H bond in the Amber force field and mH being the mass of a hydrogen atom. ## Results and Discussion Proton transfer in PL-N378D and CRYI The experimental time scale for the protonation of FAD •is 1.7 μs in CRYI and the formation of FAD x in PL-N378D occurs within 3.3 µs, which is four orders of magnitude slower than the FAD photoreduction. Consequently, the protein is able to relax after the formation of FAD •and before its protonation. Decades of nanoseconds MD simulations capture this relaxation. In a first step, we equilibrated the CRYI and E. coli mutant (PL-N378) structures using free MD simulations with neutral FADox. To model the FAD •states, we performed unbiased MD simulations from the respective neutral structures considering a non-protonated semiquinone FAD •in the active site. This generates the starting point for the PT reaction. Initially, the Asp378/396 backbone carbonyl oxygen faces the neutral FAD. After the photo-induced ET, forming a negatively charged FAD •-, the Asp residue in both proteins rotates within few nano-seconds, the protonated side chain now faces the O4 or N5 of FAD and forms a hydrogen bond which is visualized in Figure 2. This arrangement is a prerequisite for a potential PT forming FADH • . The distance between FAD N5 and the proton of Asp378/396 was measured along the MD trajectories of both PL-N378D and CRYI, as were the distances between FAD and several neighbouring amino acid side chains (shown in Table 1). In CRYI, the distance between the proton of Asp378/396 and N5 is on average smaller than in PL-N378D, where on the contrary the distance between the O4 of FAD and the proton is smaller. Indeed, the proton of Asp396 in CRYI faces more often the N5 of FAD than in PL-N378D. The neighbouring amino acids (Arg344-Asp372 in PL-N378D, Arg362-Asp390 in CRYI) show a conserved salt bridge and the relative positions of Arg344/362 and FAD differ in the two proteins, which leads to different electrostatic interactions, as detailed below. A further difference is seen in the standard deviations of the distances between the proton of the Asp and the Arg-Asp salt bridge, which are larger in CRYI, indicating a higher structural flexibility. Earlier experimental studies of PL and CRY also described the FAD pocket in CRY to be more flexible, which is in agreement with our simulations. 62, 63 In the crystal structure of CRYI the distance between Cβ of Arg344/362 and the methyl group of the FAD isoalloxazine ring, FADMet, is larger by 0.5 than in PL-N378D, which increases to more than 1 during the MD simulation. Additionally, the guanidine group of the Arg sidechain is on average 0.5 farther away from the N5 of FAD in CRYI compared to PL-N378D. Free energy landscape reveals an endergonic PT in PL-N378D After the reduction of FAD, the rotation the O-H proton, described above, brings Asp378 into a position where a PT to the N5 of FAD is possible. To study the PT-energetics, we computed the one-dimensional free energy landscape of the PT in PL-N378D and compared it to the previous results from CRYI, being presented in Figure 3. 28 The corresponding driving force and activation free enthalpy values are listed in Table 2. The three states labelled I., II. and III. designate respectively the reactants with the proton on the aspartic acid, the transition state (TST) of PT and the products with the proton on FADH • . An important difference is clearly apparent in the Gibbs free energy of the PT: while state III. is the global minimum in CRYI, state I. is favoured in PL-N378D. We considered here protonated Glu363 in PL-N378D because it showed the smallest ΔG 0 (the free energy profile of deprotonated Glu363 is added to the SI). Therefore, PT is an endergonic reaction in PL-N378D but exothermic in CRYI. The computed energy landscape allows an interpretation of the absorption spectra measured in CRYI and the PL-N378D. CRYI shows slow rate constants of 0.02 ns -1 and 0.002 ns -1 for the forward and backward PT respectively, and a well-defined FADH • spectrum, consistent with our free-energy landscape favouring FADH • . For PL-N378D, the reported measured spectrum was called FAD x , which strongly overlaps with the FAD •spectrum, however, showing only a slight deviation from the spectrum of a pure FAD •species. 20 This can be interpreted with the free-energy profile, highlighting an energetically favoured FAD •state. Nevertheless, the activation energy for the forward PT is just 4.9 kcal•mol -1 and only 2.9 kcal•mol -1 for the backward PT which corresponds to rate constants of 1.6 ns -1 and 47 ns -1 respectively. So, PT is feasible during the recording time of the spectra, and a superposition of both species may lead to the particular form of the absorption spectra. Since it was not possible to generate the CRYI functionality with a single mutation in PL, we look for another mutation to enable PT in PL-N378D. We observed that the protonation state of Glu363 influences the PT between Asp378 and FAD •in PL-N378D. At first sight, this residue could partially explain the different behaviour between both proteins. CRYI shows an apolar Met381 at the homologous position of PL-WT Glu363, so the N378D-E363M double mutant appears as a good candidate. We investigated this double mutant; however, no improved stabilization of the PT product was observed (see SI for numerical results). This double mutation is insufficient to convert the nature of WT-PL into CRYI and recover the PT activity. Therefore, a combination of multiple small contributions seems to be necessary to make the PT reaction feasible. ## Table 2: The free energies and rate constants calculated from the biased QM/MM simulations of PT in PL-N378D and in CRYI. ## Analysis of coulombic interactions between FAD and the pocket Since the considered PT involves a partial transfer of negative charge from the large isoalloxazine ring to a small carboxyl group, the electrostatic interactions between the environment and the active site may play an important role to stabilize the protonation states. To quantify these effects, the electrostatic interaction energies between the PT complex (isoalloxazine ring and Asp378/396) and all of the amino acid residues and the solvent, Ecoul, were calculated for the FADH • and the FAD •state. The difference between the product and reactant state in this coulomb interaction, ΔEcoul, illustrates the contributions of the individual interactions to the stabilisation of the PT product or the reactant. In CRYI, the FADH • -Asp378product is slightly favoured (the sum of ΔEcoul = -3.9 kcal•mol -1 ), while there is no significant preference for either state in PL-N378D (sum of ΔEcoul = +0.6 kcal•mol -1 ). It is interesting to see that the total difference in electrostatic energy of 4.5 kcal•mol -1 roughly resembles the difference in the reaction energy as shown in Table 2. The analysis of the different contributions (see Table 1SI) showed that the major difference between the proteins is based on the interaction between Aspand the environment. For further insight, we computed the contributions of the individual amino acid residues and the water molecules to ΔEcoul, visualized in the colouring scheme of Figure 4 to highlight the individual interactions between each amino acid of the pocket and the FAD-Asp complex. The analysis shows three major factors being responsible for the energy difference: the different position of the Arginine as mentioned above, the interaction between Asp378 and Glu363 and a different interaction with water molecules. No positive value of ΔEcoul larger than +0.3 kcal•mol -1 , and therefore disfavouring the FADH • state, was observed in CRYI. However, in PL-N378D several positive contributions occur (+0.3, +0.4 and +2.9 kcal•mol -1 for Leu375, Gly381 and Arg344, respectively). Thus, Arg344 (Arg362 in CRYI) interacts differently in PT PL-N378D: while it supports PT in CRYI (-1.74 kcal•mol -1 ), it disfavours it in the PL mutant. The distance between Arg and FAD is shorter and the structure is more rigid in PL-N378D (see Table 1), which could explain part of the preference for the negatively charged FAD •state. Negative values of ΔEcoul, favouring the FADH • formation, were observed similarly, such as the interaction with the first Trp of the Trp-Triad. Furthermore, in PL-N378D the Glu363 (-10.32 kcal•mol -1 ) is strongly hydrogen-bonded to the negatively charged Asp378, while in CRYI Met381 (-1.05 kcal•mol -1 ) just slightly interacts with the FAD-Asp complex. At first sight it may seem that this residue is not a good choice for stabilization of the FADH • . However, Met381 shows a flexibility allowing water molecules to approach Asp396 resulting in a stabilizing interaction of -5.58 kcal•mol -1 with negatively charged Asp. In contrast, the interaction with water molecules is positive (7.51 kcal•mol -1 ) in the PL mutant, destabilizing the product state. ## Proton Transfer in PL-WT As observed in the PL-N378D mutant, the mechanism of FAD protonation in the PL-WT might involve structural reorientation of the pocked triggered by the partial FAD photoreduction. Therefore, we performed 10 independent 50 ns MD simulations to assess the relaxation of protein structure in response to a photo-induced ET forming FAD •-, starting from an MD trajectory containing FADox and switching the charge state into FAD •by passing to a corresponding set of MM atomic charges. In the starting structure of the MD simulations the amido oxygen of N378 points towards N5 of FAD •-, which is shown in Figure 5 and will be call O-conformation in the following. In 8 out of 10 simulations, significant structural rearrangements are observed in the area enclosed by FAD, Asn378 and a loop close to the heterocycle of the isoalloxazine ring. These are:  Rotation of the Asn378 side chain to face the FAD N5: in the initial state, the keto group faces N5, after rotation the amino group of Asn (7 occurrences). This rotation, which will be further mentioned as N-conformation, occurred from 2 ns and 37 ns, but the reverse rotation was not seen.  Flow of water molecules into the FAD pocket, also facing the FAD N5 atom (1 occurrence). The rotation of the Asn378 side chain affects the protonation of FAD •-These free MD simulations give a first qualitative insight of the rotation of Asn378, but for a more detailed understanding, the energetics of the side chain rotation around the Cβ-Cγ bond were further quantified by means of biased sampling simulations. We performed The structural motives may explain the energetic preferences: (I) in case of FAD •the rotation into the N-conformation may be driven in order to avoid an electrostatic mismatch between the Asn378-oxygen and the negative charge on FAD, and the formation of a hydrogen bond between Asn378 and N5 of FAD can further stabilize the N-conformation (shown in Figure 5). (II) For protonated FAD (FADH • and FADH -), the N-conformation may be disfavoured due to a steric hindrance between H-N5 of protonated FAD and the Asn378 side chain. (III) In the FADox state the O-conformation is favoured by about 5 kcal•mol -1 , which is surprising since, as in FAD •-, a hydrogen bond in the N-conformation could be formed. Here, however, the stabilisation of the O-conformation is due to a hydrogen bonded network (not shown), formed by several water molecules and the Glu363 side chain. Our results are consistent with the experimental and computational studies by Wijaya et al., 64 which reported a hydrogen bond between the amido oxygen of Asn378 and N5-H of FADH • and FADHstates as well as the absence of the N-conformation in FADox. The formation of a water wire Immediately after FAD photoreduction to FAD •-, the system is in the O-conformation, since this is the state favourable for FADox. The rotation into the N-conformation may be a dead end for the FADHformation, since the hydrogen bond hinders protonation. For protonation, the rotation has to be reversed. As discussed above, the formation of a water wire may connect a proton donor at the protein surface with FAD, thereby enabling PT. In three of the unbiased MD simulations reported above, Asn378 remained in the O-conformation, and in one of them two water molecules entered the FAD pocket and are engaged in hydrogen bonding with FAD N5, connecting this cofactor with the protein exterior. In the beginning of the simulation, both water molecules were located in the bulk solvent, quite far from FAD. After 2.5 ns, one of them entered the pocket to interact with FAD N5. After 10 ns, the other water molecule also entered, and both of them stayed between FAD N5 and the side chain oxygen of Asn378 for about 2.5 ns. Then, the first water was released, and the second water stayed in the FAD pocket for the entire remaining simulation time. The distances between these water molecules and FAD N5 are analysed in the SI (Figure 3SI). The entrance of water molecules into the FAD pocket proceeds through a flexible loop composed of Met367 and Leu376. There are multiple other water molecules located in front of the loop (see Figure 8), which allow the formation of a water wire connecting FAD •with the bulk solvent. Importantly, the influx of water molecules into the FAD pocket is vital for the formation of such a wire, which represents a possible pathway for the protonation of FAD. To better characterise the kinetics of this event, free energy profiles were computed for the waters entering the FAD pocket. We used classical force field simulations in combination with US. To reduce the complexity of the possible reorientations in the pocket, we focused on just the free energy barrier for the entrance of two water molecules: (I) the reaction coordinate for the entrance of the first water is the distance of this water molecule to FAD. (II) The reaction coordinate for the second water is the distance of this water molecule to the first water. The free energy profiles are shown in Figure 7 I. and II. For a detailed discussion of the calculations, see SI. The first water molecule has to overcome a rather low barrier of 3.5 kcal•mol -1 corresponding to a rate of ktrans ≈ 18 ns -1 , finding a stable minimum in the binding pocket. The second water has to overcome a similar barrier, however, the final state is not stable, indicating that the formation of this water wire is a transient process. In addition, this water wire formation competes with the formation of the N-conformation of Asn378. Long-range proton transfer along a water wire These two water molecules form a path connecting FAD N5 with a potential proton donor at the exterior of the FAD-pocket. Inspection of the structure identifies Glu106 (with a pKa around 6) as the closest titrable amino acid. Snapshots from the free MD with the two water molecules entering the FAD pocket also show a water wire, connection Glu106 with FAD, as shown in Figure 8. Since this water wire is a transient phenomenon, we had to constrain these waters in order to compute the free energy profile for PT from Glu106 to FAD. QM/MM US simulations were performed, considering the Glu106 side chain, FAD •and the five water molecules as possible proton carriers in the QM region. The resulting free energy profile is shown in The transfer to the fifth water molecule and then to the N5 atom of FAD is a downhill process resulting in a product state which is 10 kcal•mol -1 below the reactant state. The rate limiting step is the deprotonation of the donor (I) with a rate constant of 0.13 s -1 , which corresponds to a mean reaction time of 7.7 s; this is on the same order of magnitude as the experimental figure of 4 s. 18,20 Stability of FADH vs FAD in the different proteins With respect to the biological functions of CRY or PL, a different protonation state of FAD needs to be stabilized. So, CRYs are used as light sensitive signalling molecules which are able to form a FAD • -Trp •+ or FADH • -Trp • radical pair, and switch to the FADox starting state after seconds. The protonation/deprotonation mechanism of FAD is facilitated by the presence of a close aspartate. On the contrary, Asn allows PL to stabilize an FADH • required for their biological function. The protonation of FAD •along the water wire showed a strongly exothermic PT, highlighting the strong stabilization of FADH • in the PL pocket of around 10 kcal•mol -1 while in CRYI the difference was just around 2 kcal•mol -1 . Additionally, no strong acceptor of the proton would be available in PL and the further photo induced ET forming FADHcan happen. In a similar way, we want to addresses the question why CRYI is not able to form FADHwhile it has the same Trp-Triad. In CRYI, the PT forming FADH • also results in a close negative charge at Asp396, which can be stabilized by the environment but still influences the energetics of FADH • . Furthermore, it was suggested that the negative charge at Asp396 induces conformational changes which may be required for the signal transduction. 26,65 To investigate this point, we computed the stabilization of the FADH •induced by the pocket. Therefore, we calculated the energy of the HOMO of FADH •in WT-PL, PL-N378D and CRYI (further details shown in SI). The pocket of PL-WT shows a decrease of the orbital energy of around 1.48 eV and therefore 0.36 eV and even 0.69 eV more than in PL-N378D and CRYI respectively, which is also shown in Figure 5SI. The close negative charge at the Asp in CRYI and PL-N378D may not allow the same stabilization than the Asn378 in WT-PL. Due to the similarity of the pocket in PL-WT and PL-N378D it can be concluded that the charge of the Asp378 causes this missing stabilization, additionally, it highlights that the pocket of PL or PL-N378D have a stronger stabilizing effect of a negative charge on FAD than the FAD pocket of CRYI. ## Conclusion The different functions of photolyases and cryptochromes are directly linked to the occurrence of different FAD oxidation and protonation states, and therefore, one interesting question to be addressed is how these states are stabilized by the respective protein environments. Experimental studies have identified one crucial amino acid, which is responsible for cofactor protonation in plant CRY (Asp396 in CRYI), while PL have conserved an Asn. This Asn stabilizes the biological requested FADH • /FADHstates without preventing protonation of FAD •-, whereas the mutation of Asn into Asp (PL-N378D) inhibits the stabilization of FADH • . Also mutating the nearby Glu363 into Met, which is an obvious difference between the CRYI and PL-N378D proteins, does not change the FAD protonation equilibrium, although the protonation state of Glu363 can have a decisive influence on the PT barrier. Therefore, more differences between the CRY and PL active site environments must affect the architecture of the pocket and thereby influence the FAD oxidation and protonation states. In this work, we used classical and QM/MM MD simulations to investigate in detail the FAD binding pocket and the factors, responsible for the stabilization and particular proton transfer events, summarised in Figure 9. In both proteins, a reorientation of Asp378/396 occurs after ET to FAD (Figure 9), leading to a conformation, from which a PT to FAD can occur. In CRYI, this leads to a stable FADH • species, while the electrostatic interaction on PL-N378D makes this thermodynamically unfavourable, leading to the of FAD x which dominantly resembles the FAD •species. Electrostatic interaction with the environment and hydrogen bonding can achieve stabilization of the FAD oxidation and protonation states. Therefore, differences in the Coulomb interactions between the FAD-Asp ion-pair and the flexibility of the pocket influence the PT energetics: the location, rigidity of the Arg344 in PL-N378D and the interaction with the water molecules entering the FAD pocket. So, a different water distribution in CRYI allows a stabilization of the deprotonated Asp396, which results to an exothermic PT, while it is endothermic in PL-N378D. Protonation of FAD •requires a proton donor, which is adjacent to the FAD in CRYI, while possible donors are located outside of the FAD binding pocket in PL, leading to much longer proton transfer times. So, our simulations suggest a potential proton donor Glu106 outside the FAD binding pocket. The proton transfer to FAD then needs the transient formation of a water wire, a step which is endothermic occurring on a ns time-scale and competing against the rotation of Asn378 (Figure 9), which is exothermic in presence of FAD •-. The proton transfer along the wire is exothermic, but has to overcome a large barrier of more the 19 kcal•mol -1 , leading to a time-scale in the seconds-regime, in agreement with experiments. Our results underline the strong structure-function relationship around FAD protonation in the PCF. For DNA repair, FADHis required, because this will have the proper electronic structure after excitation in order to allow ET to the DNA lesion. The presence of Asn strongly stabilizes the protonated FAD, highlighted by our exothermic PT in PL-WT, while the pocket favours a negative charge on FAD. On the contrary, in CRYI, the negative charge on Asp396 would have a repulsive interaction with the negative charge on FADH -, thereby not allowing the photoreduction of FADH • . However this aspartate residue insures fast protonation and deprotonation of FAD during the photocycle 26 , in agreement with the formation and stabilization of the radical-pair in the protein, required for the CRYI signalling function. ## Figure 9: PT and ET reactions in CRYI, PL-WT and PL-N378D The experimentally explored ET and PT reactions of PL E. coli by Brettel and CRYI by Kottke are extended by structural reorientations observed in biased and unbiased MD simulations.

chemsum

{"title": "What accounts for the different functions in Photolyases and Cryptochromes: a computational study of critical events in the protein active sites", "journal": "ChemRxiv"}

raman_spectroscopy_as_a_tool_for_monitoring_mesoscale_continuous-flow_organic_synthesis:_equipment_i

## Abstract: An apparatus is reported for real-time Raman monitoring of reactions performed using continuous-flow processing. Its capability is assessed by studying four reactions, all involving formation of products bearing α,β-unsaturated carbonyl moieties; synthesis of 3-acetylcoumarin, Knoevenagel and Claisen-Schmidt condensations, and a Biginelli reaction. In each case it is possible to monitor the reactions and also in one case, by means of a calibration curve, determine product conversion from Raman spectral data as corroborated by data obtained using NMR spectroscopy. ## Introduction Continuous-flow processing is used in the chemical industry on production scales. In a research and development setting, there has been increasing interest in using flow chemistry on smaller scales. To this end, a wide range of companies now produce equipment for both micro-and mesofluidic flow chemistry . Some of the advantages of these devices are increased experimental safety, easy scale-up and thorough mixing of reagents . It is not surprising, therefore, that a wide range of synthetic chemistry transformations have been reported using this equipment . When it comes to evaluating the outcome of reactions performed using flow chemistry and optimizing reaction conditions, one option is to use inline product analysis. This opens the avenue for fast, reliable assay in comparison with the traditional approach in which performance is evaluated based on offline product analysis. When interfaced with microreactors, inline analysis has taken significant strides in recent years . Spectroscopic tools such as infrared , UV-visible , NMR , Raman , and mass spectrometry have all been interfaced with success. There have been less reports when it comes to mesoflow systems. Perhaps most developed is the area of infrared monitoring. The now ubiquitous ReactIR equipment has been interfaced with commercially available flow equipment to allow for real-time analysis of reactions and on-the-fly optimization of conditions . In our laboratory we have had success interfacing a Raman spectrometer with a scientific microwave unit . This has allowed us to monitor reactions from both a qualitative and quantitative perspective. A recent report of the use of Raman spectroscopy for monitoring a continuous-flow palladium-catalyzed cross-coupling reaction sparked our interest in interfacing our Raman spectrometer with one of our continuous-flow units and employing it for inline reaction monitoring of a number of key medicinally-relevant organic transformations. Our results are presented here. ## Results and Discussion Interfacing the spectrometer to the flow unit In interfacing our Raman spectrometer with a continuous-flow reactor, our objective was to use a similar approach to that which proved successful when using microwave heating. Borosilicate glass is essentially "Raman transparent". Therefore reactions could be monitored by placing a Raman probe near the reaction vessel, without requirement to place any parts of the spectrometer inside the reaction vessel. The exposure of metallic components to the microwave field was avoided using a quartz light-pipe extending both the excitation laser and the acquisition fiber optic components of the spectrometer almost without any loss of light. The optimum distance of the lightpipe to the outside wall of the reaction vessel was found to be approximately 0.5 mm. Moving to our continuous-flow reactor, we decided to place the spectroscopic interface just after the back-pressure regulator assembly. This meant that we did not need to engineer a flow cell capable of holding significant pressure. Instead we used an off-the-shelf flow cell traditionally used in conjunction with other spectroscopic monitoring tools. The cell had screw-threaded inlet and outlet tubes of the same diameter as the tubing of the flow unit (i.d. 1 mm). The sample chamber had a width of 6.5 mm, height of 20 mm and a path length of 5 mm giving the cell a nominal internal volume of 0.210 mL (Figure 1a). We built an assembly to allow us to hold the cell in a fixed location and vary the distance of the quartz light-pipe so as to optimize the Raman signal intensity. The apparatus is shown in Figure 1b. ## Testing the interface: The synthesis of 3-acetylcoumarin As our first reaction for study, we selected the piperidinecatalyzed synthesis of 3-acetylcoumarin (1) from salicylaldehyde with ethyl acetoacetate (Scheme 1). We had extensive ## Scheme 1: The reaction between salicylaldehyde and ethyl acetoacetate to form 3-acetyl coumarin (1). experience of monitoring this reaction both qualitatively and quantitatively when using microwave heating so believed it would be a good starting point for our present study. The reaction works well when using ethyl acetate as the solvent. However, 1 is not completely soluble at room temperature. To overcome potential clogging of the back-pressure regulator as well as mitigating the risk of having solid particles in the flow cell (which would perturb signal acquisition), we leveraged a technique we developed for this and other reactions previously . Once the reaction stream has exited the heated zone, it is intercepted with a flow of a suitable organic solvent. This solubilizes the product and allows it to pass through the back-pres-sure regulator unimpeded. In the case of 1, we intercept the product stream with a flow of acetone. Our first objective was to determine whether we could observe spectroscopically a slug of the coumarin passing through the flow cell. The Raman spectrum of 1 (Figure 2) exhibits strong Raman-active stretching modes at 1608 cm −1 and 1563 cm −1 while the salicylaldehyde and ethyl acetoacetate starting materials exhibit minimal Raman activity in this area. As a result, we chose to monitor the 1608 cm −1 signal. To mimic a product mixture, we pumped a solution of 1 in acetone through our flow reactor, intercepted it with an equal volume of ethyl acetate and passed this mixture through the flow cell. We recorded a Raman spectrum every 15 s in an automated fashion as the coumarin passed through the cell by using the "continuous-scan" function on our spectrometer. By subtracting the spectrum of the solvent mixture (1:1 ethyl acetate:acetone) from the spectra recorded, we were able to clearly see the growth of the signal due to 1 followed by a plateau as it passed through the cell and then a drop back to the baseline as the final aliquot exited (Figure 3). Knowing we could observe the product as it passed through the flow cell, we next performed the complete reaction. As a starting point, we chose as conditions a flow rate of 1 mL/min through a 10 mL PFA coil at room temperature. We were indeed able to monitor the reaction as shown in Figure 4. In an effort to optimize the reaction conditions, we varied both the temperature of the reactor coil and also the flow rate, monitoring each run and then compiling the data (Figure 4). While increasing the reaction temperature to 130 °C led to a marked increase in product conversion, reducing the flow rate from 1 mL/min to 0.5 mL/min at this temperature did not have a significant impact on the outcome of the reaction. In an attempt to quantify product conversion, we needed next to obtain a calibration curve to allow us to convert units of Raman intensity to units of concentration in standard terms. To achieve this, we passed solutions of various concentrations of 3-acetylcoumarin (3) in ethyl acetate/acetone through the flow cell and collected the Raman spectrum. When the signal intensity at 1608 cm −1 is plotted against concentration, after subtraction of signals due to the solvent, the result is a straight line (Figure 5). The Stokes shift (which is being monitored) is inversely proportional to the temperature. Since the flow cell is situated after the product mixture exits the heated zone and because of the very efficient heat transfer observed using narrow-gauge tubing, the product mixture was essentially at room temperature by the time it passed through the flow cell. As a result, it was not deemed necessary to involve a scaling factor to account for temperature effects. With the appropriate calibration curve in hand, we were able to obtain product conversion values for each set of reaction conditions screened, taking into account the fact that the product concentration is halved by the interception with acetone. To determine their accuracy, we also determined product conversion using NMR spectroscopy. Comparison of the values shows a good correlation (Table 1). ## Expanding the technique to other reactions The Knovenagel condensation We turned our attention next to the Knoevenagel condensation of ethyl acetoacetate with a range of aromatic aldehydes (Scheme 2). Our objective was to optimize conditions using one aldehyde substrate spectroscopically from a qualitative standpoint and then screen other examples. We chose benzaldehyde as our initial substrate, ethyl acetate as the solvent and piperidine as a base catalyst. In order to determine the optimal spectral frequency at which to monitor we wanted to find a quick way to derive the Raman spectrum of the product 2a. As was the case with 1, this could be achieved computationally using Gaussian 09 at the B3LYP/6-31g(d) level of theory , and a signal at 1598 cm −1 selected for monitoring. Performing the reaction across a range of conditions, flowing the reaction mixture at 1 mL/min through the 10 mL coil heated to 130 °C proved to be optimal (Figure 6). A 67% conversion to 2a was obtained, as determined by GC analysis. Purification of the product mixture gave a 60% isolated yield of the Z-isomer of 2a. Using these optimized reaction conditions, we screened three para-substituted aldehyde substrates (Table 2). As expected, placing an electron-donating methoxy group on the aromatic ring led to lower product conversion as compared to benzaldehyde (Table 2, entry 2). A methyl-or fluoro-substituent has little effect on the outcome of the reaction (Table 2, entries 3 and 4). ## The Claisen-Schmidt condensation We moved next to study the Claisen-Schmidt condensation of benzaldehyde with acetophenone to yield chalcone (Scheme 3). Chalcones display interesting biological properties such as antioxidant, cytotoxic, anticancer, antimicrobial, antiprotozoal, antiulcer, antihistaminic, and anti-inflammatory activity . They are also intermediates on the way to highly fluorescent cyanopyridine and deazalumazine dyes . The calculated Raman spectrum of the product 3a shows a very strong signal at 1604 cm −1 which was selected for monitoring. Using sodium hydroxide as the catalyst, the reaction was monitored under a range of reaction conditions (Figure 7). We fast discovered that at temperatures in excess of 65 °C we observed decomposition or else formation of a highly fluorescent byproduct, as evidenced by collapse of the Raman spectrum. We also observed a significant "tail" on the plot of signal intensity at 1604 cm −1 vs time. We attribute this to the fact that the chalcone product is very highly Raman active and even a trace in the flow cell can be readily detected. It does however highlight the fact that there may be both significant dispersion along the length of the reactor and the product is slow in clearing the flow cell. Dispersion is the consequence of laminar flow and some of the material takes longer to travel through the reactor than the rest. Thus, when a flow reactor is used to process a finite volume of reagents, the leading and trailing ends of the product emerging from the end of the reactor will have mixed to some extent with the solvent that preceded or followed it. This means that there are zones at the leading and trailing ends of the product stream in which the concentration of product is variable. Our optimal conditions for the reaction were heating at 65 °C with a flow rate of 1 mL/min, this corresponding to a product conversion of 90%, as determined by GC analysis. Performing the reaction under these conditions using three substituted benzaldehydes as substrates, we obtained product conversions of 66−98% depending on how electron rich or deficient the aromatic ring of the aldehyde was (Table 3). ## The Biginelli reaction As our final reaction for study, we turned to the Biginelli reaction (Scheme 4) . This acid-catalyzed cyclocondensation of urea, β-ketoesters and aromatic aldehydes to yield dihydropyrimidines has received significant attention, these products having pharmacological activity including calcium channel modulation, mitotic kinesin Eg5 inhibition, and antiviral and antibacterial activity . The Biginelli reaction has been performed in flow previously as a route to densely functionalized heterocycles using HBr generated in a prior step as the catalyst for the reaction . Copper catalysis has also been used in flow mode for preparing PEG-immobilized dihydropyrimidines . We decided to screen a set of conditions for the reaction of benzaldehyde, ethyl acetoacetate and urea catalyzed by sulfuric acid (Figure 8). The calculated Raman spectrum of the product, 4a, shows a strong signal at 1598 cm −1 which was selected for monitoring. Using a catalyst loading of 10 mol % and a flow rate of 1 mL/min, we monitored the reaction over a temperature range from 25-120 °C. Seeing that the reaction did not reach completion within the 10 min in the heated zone, we then repeated the process at lower flow rates; first to 0.5 mL/min and then 0.25 mL/min. Our optimal conditions as determined by Raman monitoring were heating at 120 °C with a flow rate of 0.25 mL/min, this corresponding to a product conversion of 89%, as determined by GC analysis, and a product yield of 78% after purification. Performing the reaction using three other aldehyde substrates resulted in similar product conversions (Table 4). ## Conclusion In conclusion, we describe here an apparatus for real-time Raman monitoring of reactions performed using continuousflow processing. We assess its capability by studying four reactions. We find that it is possible to monitor reactions and also, by means of a calibration curve, determine product conversion from Raman spectral data as corroborated by data obtained using NMR spectroscopy. Work is now underway to expand the scope of the method to other classes of useful reactions. ## Experimental General experimental All reagents are used as received from the various vendors without purification. Sodium sulfate, MeOH, EtOH, EtOAc, DMF and Et 2 O (ACS Grade and reagent grade), were purchased from Sigma-Aldrich and used without further purification. Deuterated NMR solvents (CDCl 3 ) were purchased from Cambridge Isotope Laboratories. CDCl 3 stored over 4 molecular sieves and K 2 CO 3 . NMR Spectra ( 1 H, 13 C, 19 F) were performed at 298 K on either a Bruker DRX-400 MHz NMR, or Bruker Avance 500 MHz NMR. 1 H NMR Spectra obtained in CDCl 3 were referenced to residual non-deuterated chloroform (7.26 ppm) in the deuterated solvent. 13 C NMR Spectra obtained in CDCl 3 were referenced to chloroform (77.3 ppm). 19 F NMR spectra were referenced to hexafluorobenzene (−164.9 ppm) . Reactions were monitored by an Agilent Technologies 7820A Gas Chromatograph attached to a 5975 Mass Spectrometer or 1 H NMR. Flash chromatography and silica plugs utilized Dynamic Adsorbants Inc. Flash Silica Gel (60 porosity, 32-63 µm). ## Apparatus configuration The Raman system used was an Enwave Optronics Spectrometer, Model EZRaman-L . The continuous-flow unit used was a Vapourtec E-series. A Starna 583.65.65-Q-5/Z20 flowcell (width: 6.5 mm, height: 20 mm, path length: 5 mm) was placed inline after the back-pressure regulator using 1 mm i.d. PFA tubing (the void volume between the flow reactor and the flow cell was 4.79 mL). The flow cell was secured in place in a custom-made box and the fiber-optic probe from the spectrometer inserted so it touched the wall of the flow cell. During a reaction, spectral data was recorded at pre-determined time intervals using the EZ Raman software provided with the instrument. The data was then exported to Excel for processing. Typical procedure for monitoring the formation of 3-acetylcoumarin (1) Performing the reaction: Into a 50 mL volumetric flask was added salicylaldehyde (6.106 g, 50 mmol, 1 equiv) and ethyl acetoacetate (6.507 g, 50 mmol, 1 equiv). Ethyl acetate was added to bring the total volume to 50 mL (1 M) and the reagents were thoroughly mixed. An aliquot of this solution (10 mL) was transferred to a 20 mL vial equipped with a Teflon-coated stir bar. The flow reactor was readied using the equipment manufacturer's suggested start-up sequence. Ethyl acetate was pumped at 1 mL/min to fill the reactor coil. The back-pressure regulator was adjusted to 7 bar and the reactor coil heated to 65 °C. After the heating coil, the product stream was intercepted with a stream of acetone (1 mL/min) by means of a T-piece to ensure complete solubility of the product. The Raman probe was inserted into the box containing the flow cell and was properly focused. A background scan of the ethyl acetate/acetone solvent system was taken. This background was then automatically subtracted from all subsequent scans, thereby removing any signals from the solvent. The Raman spectrometer was set to acquire a spectrum every 15 s throughout the run, with 10 s integration time, boxcar = 3, and average = 1. When the flow unit was ready, piperidine (0.099 mL, 0.1 mmol, 0.1 equiv.) was injected all at once into the vial containing the reagents and, after mixing for 15 s, the reaction mixture was loaded into the reactor at a flow rate of 1 mL/min. After the reaction mixture had been completely loaded into the reactor, ethyl acetate was again pumped through the coil at 1 mL/min. After the product had been fully discharged from the flow cell, the scans were halted. While the product mixture was passing through the flow cell, a drop of the exit stream was removed and an NMR spectrum recorded to obtain product conversion for comparison with data obtained by Raman spectroscopy. NMR conversions were determined by comparing signals from the starting salicylaldehyde (9.84 ppm) and the coumarin product (8.45 ppm) . Obtaining a relationship between signal strength and concentration: To obtain a calibration curve, spectra of 3-acetylcoumarin in 1:1 ethyl acetate/acetone were recorded at a range of concentrations by passing the solutions through the flow cell. A plot of signal strength due to the peak at 1608 cm −1 versus concentration of 1 was constructed (Figure 5). From this, units of Raman intensity could be converted to units of concentration in standard terms and hence product conversion determined. ## Typical procedure for monitoring the Knoevenagel reaction An analogous approach was used to prepare the Knoevenagel product as for the case of 1, benzaldehyde (5.306 g, 50 mmol, 1 equiv) being used in place of salicylaldehyde and there being no need for acetone interception of the product mixture. The Raman spectrometer was programmed to take continuous scans using the same parameters as in the case of 1. After the product had been fully discharged from the flow cell, the scans were halted. The resulting clear yellow solution was poured over aqueous 2 M HCl and extracted with ethyl acetate. The combined organic layers were washed with brine, dried over sodium sulfate, and the solvent was removed in vacuo by rotary evaporation affording the crude product. The crude product was loaded on a 15-cm silica gel column (55 g silica gel) and a gradient eluting system (99:1, 95:5, 90:10; Hex:EtOAc) was used to obtain (Z)-ethyl 2-benzylidene-3-oxobutanoate (2a, 1.3095 g, 60%) as a clear yellow oil. Typical procedure for monitoring the Claisen-Schmidt reaction Into a 50 mL volumetric flask was added 4-fluorobenzaldehyde (1.551 g, 12.5 mmol, 1 equiv) and acetophenone (1.637 g, 12.5 mmol, 1 equiv). Ethanol was added to bring the total volume to 50 mL (0.25 M) and the reagents were thoroughly mixed. An aliquot of this solution (10 mL) was transferred to a 20 mL vial equipped with a Teflon-coated stir bar. The flow reactor was readied using the equipment manufacturer's suggested start-up sequence. Ethanol was pumped at 0.5 mL/min to fill the reactor coil. The back-pressure regulator was adjusted to 7 bar and the reactor coil heated to 65 °C. After the heating coil, the product stream was intercepted with a stream of acetone (0.5 mL/min) by means of a T-piece to ensure complete solubility of the product. The Raman spectrometer was configured as in the case of monitoring formation of 1. When the flow unit was ready, 2 M NaOH (0.125 mL, 0.25 mmol) was injected all at once and after mixing for 15 s the reaction mixture was loaded into the reactor coil at a flow rate of 0.5 mL/min. After the reaction mixture had been completely loaded into the reactor, ethanol was again pumped through the coil at 0.5 mL/min. After the product had been fully discharged from the flow cell, the scans were halted. The yellow product solution was poured into a beaker containing ice (100 g) causing an immediate precipitation of the product. To ensure complete precipitation, the solution was stirred at 0 °C. The solid product was collected via vacuum filtration and washed with cold ethanol. The material was dried in air to yield (E)-3-(4-fluorophenyl)-1-phenylprop-2-en-1-one, (3d, 0.5421 g, 91%) as a pale yellow solid. ## Typical procedure for monitoring the Biginelli reaction In a 50 mL volumetric flask was dissolved urea (3.003 g, 50 mmol, 1 equiv.) in methanol (~30 mL). Into the flask was then added benzaldehyde (1.306 g, 50 mmol, 1 equiv) and ethyl acetoacetate (6.507 g, 50 mmol, 1 equiv). Methanol was added to bring the total volume to 50 (1 M) and the reagents were thoroughly mixed. An aliquot of this solution (10 mL) was transferred to a 20 mL vial equipped with a Teflon-coated stir bar. The flow reactor was readied using the equipment manufacturer's suggested start-up sequence. Methanol was pumped at 0.25 mL/min to fill the reactor coil. The back-pressure regulator was adjusted to 7 bar and the reactor coil heated to 120 °C. After the heating coil, the product stream was intercepted with a stream of N,N-dimethylformamide (0.25 mL/min) by means of a T-piece to ensure complete solubility of the product. The Raman spectrometer was set to acquire a spectrum every 25 s, with 20 s integration time, boxcar = 3, and average = 1. When the flow unit was ready, 6 M H 2 SO 4 (0.2 mL, 0.1 equiv) was injected all at once and after mixing for 15 s the reaction mixture was loaded into the reactor coil at a flow rate of 0.25 mL/min. After the reaction mixture had been completely loaded into the reactor, methanol was again pumped through the coil at 0.25 mL/min. After the product had been fully discharged from the flow cell, the scans were halted. The reaction mixture was transferred to a separatory funnel, diluted with diethyl ether and quenched with satd. sodium bicarbonate (100 mL) and deionized water (100 mL). The layers were separated and the aqueous layer was extracted with diethyl ether (3 × 100 mL). The combined organic layers were washed with brine (2 × 100 mL) and dried over sodium sulfate. The solvent was removed in vacuo by rotary evaporation affording the crude product. The resulting solid was transferred to a filter funnel and was washed with cold methanol. The solid was isolated and air dried to afford 5-ethoxycarbonyl-6-methyl-4-phenyl-3,4-dihydropyrimidin-2(1H)-one, (4a, 2.030 g, 78%) as a fluffy white solid. 1

chemsum

{"title": "Raman spectroscopy as a tool for monitoring mesoscale continuous-flow organic synthesis: Equipment interface and assessment in four medicinally-relevant reactions", "journal": "Beilstein"}

acid_dissociation_in_(hx)_n_(h_2_o)_n_clusters_(x_=_f,_cl,_br,_i;_n_=_2,_3)

## Abstract: The interactions between two or three hydrogen halide molecules and the same number of water moieties are investigated through a systematic exploration of the corresponding potential energy surfaces using a stochastic methodology in conjunction with density functional theory computations. Our results indicate that HF, the weakest acid in the series, is partially dissociated. Similarly, HCl, HBr, and HI undergo dissociation in the presence of three, two, and two water molecules, respectively. The decrease in the number of water molecules required for dissociation, when compared with clusters with one single HX molecule, suggests cooperative effects. Interestingly, the hydrogen-bridged bihalide anions (XHX -) are present in the global minimum of (HX)n(H2O)n clusters with X = Br, I and n = 2, 3. These anions are even persistent at room temperatures. Concomitantly, a broad spectrum of interactions is found, among them water⋯water, HX⋯water, and HX⋯HX hydrogen bonds, halogen bonds, ionic and long-range X⋯H interactions. ## Introduction Acid dissociation is a complex process affected by several factors, most prominently the environment and the strength of the proton-containing bond to be broken. In the aqueous phase, acidic strength is dictated by the pKa of the species in question. The hydrogen halide series covers a spectrum from the weak to the very strong acidity, as it is clear from Table 1. Interactions between hydrogen halides and water molecules are relevant because they participate in numerous biochemical and atmospheric reactions (Keesee, 1989;Müller et al., 1992;Rokita, 1992;Peter, 1997;De Haan et al., 1999;Domcke and Sobolewski, 2003). Moreover, the involved processes, such as proton transfer, acid dissociation, and the interactions that stabilize HX-water complexes, epitomize some central questions in chemistry. The nature of the interactions escapes descriptions in simple terms, rather it is necessary to invoke a complex mixture of electrostatic, covalent, induction, and dispersion contributions, varying in relative weights. Table 1. Acidic power and properties of the H-X bond in the hydrogen halide series. pKas for dissociation in aqueous environments, bond lengths in , and dipole moments in Debyes (Cotton and Wilkinson, 1988;Weast et al., 1989;Sato and Hirata, 1999). Microsolvation studies of a single acid molecule with one or more water molecules have expanded our understanding of acid dissociation and, in some cases, have pinpointed 4 the number of solvent molecules needed to achieve the formation of the ion pair (Laasonen and Klein, 1994;Schindler et al., 1994;Lee et al., 1996;Ando and Hynes, 1999;Conley and Tao, 1999;Gertner et al., 1999;Cabaleiro-Lago et al., 2002;Devlin et al., 2002;Hurley et al., 2003;Voegele and Liedl, 2003;Kuo and Klein, 2004;Odde et al., 2004;Gutberlet et al., 2009;Vargas-Caamal et al., 2016a). These studies correspond to ideal situations that do not account for crucial factors such as increasing acid concentration via the explicit addition of new molecules. Indeed, there are fewer reports about microsolvation of clusters with more than one acid molecule compared with those for a single molecule, and many of them are focused on HCl. For instance, Morrison et al. reported an infrared spectroscopy study of mixed (HCl)m(H2O)n clusters (m:n = 1:1, 2:1, 2:2, and 3:1) immersed in helium nanodroplets (Morrison et al., 2010). They discussed infrared spectra in the HCl stretch region (2600-2900 cm -1 ) and determined that the 2:2 cluster has a nonalternating cyclic arrangement (Figure 1a). Chaban and coworkers analyzed the (HCl)2(H2O)2 and (HCl)4(H2O)4 aggregates during their discussion on the transition from hydrogen bonding to ionization (Chaban et al., 2001). The hydrogen-bonded isomer of (HCl)2(H2O)2 (see Figure 1b) is 6.9 kcal mol -1 lower in energy than the ionic one (Figure 1c), both with an alternating cyclic structure. Conversely, the 4:4 complex adopts a cubic form, where the full dissociated isomer is almost 16 kcal mol -1 more stable than the hydrogen-bonded one. The authors consider that the incorporation of anharmonic effects is a crucial factor for the prediction of reliable vibrational spectra and that cooperative effects in the solvation of hydrogen halides are extremely important. Recently, Zakai et al. performed classical molecular dynamics simulations to illustrate the transitions between the H-bonded and ionic isomers of (HCl)2(H2O)2 (Figure 1b and 1c) and of (HF)4(H2O)4 (Zakai et al., 2017). They reported that proton transfers were fully concerted in all trajectories for [Cl -⋯H3O + ]2, whereas for [F -H3O + ]4, the fully concerted 5 mechanism is dominant but partially concerted transfers of two or three protons at the same time also occur. The authors remarked that the high symmetry of the ionic and the H-bonded structures plays a key role in the collective propagation of the protons and cooperative effects. and c) ionic arrangement (Chaban et al., 2001). In order to investigate some important questions such as how many water molecules are required to onset dissociation and underline the cooperativity in hydrogen-bonded systems, we systematically explore the potential energy surfaces (PES) of (HX)n(H2O)n clusters, with X = F, Cl, Br, I and n = 2, 3. This work contributes to the understanding of the process of microsolvation of acids supported by a large variety of structures and interactions, ranging from hydrogen bonds of different strengths to halogen interactions. Perhaps, one of the most remarkable results is the stabilization of hydrogen-bridged bihalide (XHX -) moieties in the global minimum of (HX)n(H2O)n clusters with X = Br, I. We hope that the formation of hydrogen-bridged bihalide anions and the increasing relevance of the X⋯H⋯X interactions in the heavier halogens found here stimulate new perspectives on the general subject of acid microsolvation. ## Computational details In order to systematically explore the PES of the (HX)n(H2O)n clusters (with X = F, Cl, Br, I and n = 2, 3), a modified kick algorithm as is implemented in the GLOMOS code was used (Grande-Aztatzi et al., 2014;Ramirez-Manzanares et al., 2015;Cabellos et al., Cinvestav Mérida, Yuc. México. 2013). This heuristic has been employed by us to explore the PESs of several molecular clusters (Grande-Aztatzi et al., 2014;Vargas-Caamal et al., 2015;Vargas-Caamal et al., 2016a;Vargas-Caamal et al., 2016b;Murillo et al., 2017). A hierarchical screening is established, the initial structures were optimized at the PBE0/D95V level and then refined at the PBE0-D3/def2-TZVP level (Weigend and Ahlrichs, 2005;Grimme, 2006;Grimme, 2011). Each structure was characterized as a true minimum by harmonic vibrational frequency analysis. All computations were done using the Gaussian 09 (revision D.01) package (Frisch et al., Gaussian, Inc., Wallingford, CT. 2013). Temperature effects were analyzed using Eyringpy 1.0 (Dzib et al., 2019). To examine the interactions, Wiberg bond indices (WBI) (Wiberg, 1968) were computed. Stern-Limbach (Limbach et al., 2009) plots were used to correlate the symmetry and length of hydrogen bonds. In these plots, q1 measures the displacement of the proton from the center in idealized linear hydrogen bonds, while q2 quantifies the distance between the two electronegative atoms (see Figure 2 for definition of variables). For hydrogen bonds involving equally electronegative atoms, q1 = 0 indicates the point at which the proton is equally shared by both atoms, while positive or negative values denote that one of the electronegative atoms in the hydrogen bond is more strongly bound to the proton. ## Structures and Energetics Table 2 contains the energy ranges and the number of structures obtained, which are 155 for the 2:2 systems and 815 for the 3:3 clusters, including enantiomers. As for other systems interacting with water molecules, the number and diversity of structures is huge in a relatively small energy range (Cabaleiro-Lago et al., 2002;Odde et al., 2004;Perez et al., 2008;Ramirez et al., 2011;Acelas et al., 2013;Vargas-Caamal et al., 2016a;Hadad et al., 2018). Table 2. Number of structures and ZPE corrected electronic energies in kcal mol -1 of the (HX)n(H2O)n clusters. The lowest ZPE-corrected energy structures for each molecularity are presented in Figure 3. The most stable forms for the F and Cl clusters and n = 2 adopt a nondissociated H-bonded alternating structure, in agreement with previous reports (Chaban et al., 2001). In the case of the bromine 2:2 complex, the global minimum can be described as a dissociated H-bonded nonalternating cyclic form in such a way that there is a transfer of a hydrogen atom to generate a H3O + cation. The most beautiful structure is the global minimum of (HI)2(H2O)2 (see Figure 3), in which the IHIanion and the hydronium cation coexist! This Cs structure is a transition state for bromine, and it is not detected for clusters with fluorine and chlorine. Conversely, for 3:3 clusters, the lowest ZPE corrected energy clusters adopt cage-like arrays, but again the hydrogen bihalide anions are also present in the Br and I systems. It is important to comment that for the (HF)3(H2O)3 cluster, there is a planar hexagonal structure at less than 0.1 kcal mol -1 from the putative global minimum, so both arrangements are degenerate. Note that hydrogen bihalide anions are stabilized only for the stronger acids (HBr and HI). These anions are linear with a hydrogen atom placed between two halide atoms and form strong intramolecular hydrogen bonds (Jiang and Anderson, 1974;Emsley, 1980;Kemp and Gordon, 2010;Grabowski, 2016;Grabowski et al., 2016). Grabowski et al. studied the XHXanions as a way to compare the similarity between hydrogen and gold (Grabowski et al., 2016). Although their analysis was based exclusively on the anions and our clusters include anions, hydronium, and water molecules, it seems plausible to compare their computed X-H bond lengths (1.70 and 1.90 for Br-H and I-H, respectively) against our results. Thus, while the IHIanion in the (HI)2(H2O)2 global minimum has an I-H bond length of 1.93 and it is symmetric, in 3:3 complexes there are differences between the X⋯H distances in transient symmetry, with ranges between 1.65-1.78 and 1.87-1.97 for BrHBrand IHI -, respectively. These asymmetries are due to the interactions with the solvent molecules (Pylaeva et al., 2017). Kemp and Gordon conducted a study of BrHBr -(H2O)n and IHI -(H2O)n clusters with n =1 to 6 in order to determine the preferred solvated structures (Kemp and Gordon, 2010). They found that for n = 1,2, water molecules prefer to donate their hydrogen atoms for hydrogen bonding. Conversely, in this work, we started with neutral HX and H2O and found the spontaneous formation of hydrogen bihalide anions! ## Cooperativity Leopold published a comprehensive review related to the dissociation of acids in aqueous environments, in which collects theoretical evidence about the number of water molecules needed to dissociate some simple acids, suggesting 4, 3, and 3 water molecules for HCl, HBr, and HI, respectively. For the case of HF, the acid is partially dissociated for n > 7. Remarkably, our computations indicate dissociation with just two (for HBr and HI) and three (for HCl) water molecules. This is, on one hand, consistent with the pKas listed in Table 1, and on the other hand, the decrease in the number of water molecules required for dissociation, when compared with clusters with one single HX molecule, suggests cooperative effects. Cooperativity has been reported as an important effect in studies about acid ionization and hydration of ions (Tielrooij et al., 2010;Zakai et al., 2017). But even when cooperativity is one of those concepts that seems intuitively clear, it is not rigorously defined. Nonetheless, it seems generally accepted that it is related to non-additivity of the energy of individual contacts, thus, it implies that the total interaction energy of a system is larger than the simple addition of the individual interactions. Several methods attempt to get a quantitative account of the cooperative effects by decomposing the interaction energy of a system of n bodies and by the inclusion of the many-body terms in the analysis (Dannenberg, 2002;Esrafili et al., 2011;Guevara-Vela et al., 2016;Mahadevi and Sastry, 2016). Here, just to illustrate cooperative effects, we take the total interaction energy of two structurally related geometries, that is, the same patterns of hydrogen bonds but with different sizes: the global minima for (HF)2(H2O)2 and the planar hexagonal structure for (HF)3(H2O)3 (see Figure 3), for those, total interaction energies are 32.2 and 54.0 kcal mol -1 , respectively, which afford 8.1 and 9.0 kcal mol -1 per hydrogen bond, which clearly highlight the cooperative effects because if there were no cooperation, interaction energies per hydrogen bond would be the same. ## Temperature Effects Are these clusters stable at room temperature? The dissociation energies (∆Ediss) of the (HX)2(H2O)2 and (HX)3(H2O)3 clusters into water and HX are shown in Table 3. Clearly, all complexes are viable at 0 K with ∆Ediss in the range of 19.5-54.0 kcal mol -1 and the 3:3 clusters require more energy to separate them than the 2:2 ones. The increase in temperature (or the increase of the entropic effects) causes havoc in the stability of the clusters, to such an extent that the dissociation into water and acid molecules tends to be exergonic for some of them at room temperature. This is consistent with the experimental evidence that the acid strength is inversely proportional to temperature for HF, HCl, and HBr aqueous solutions (Ayotte et al., 2005). From Table 1, it is also apparent that range (Tmax) in which the complexes are viable depends on the molecularity and on the acid strength. Note that the hydrogen bihalide anions are persistent for the 3:3 systems even at room temperature. Table 3. Dissociation energies, Ediss, and free dissociation energies, Gdiss, (298.15 K and 1 atm) of the (HX)2(H2O)2 and (HX)3(H2O)3 clusters into water and HX in kcal mol -1 , and maximum temperature (K) for stability of the (HX)n(H2O)n clusters. ## Bonding analysis Wiberg bond index is a parameter related to the electron density shared between atoms, providing a reasonable quantification of bond order (Wiberg, 1968;Mayer, 2007). We provide in Figure 4 the distributions of the WBIs for H⋯X contacts for 2:2 and for 3:3 clusters. For both molecularities, most HF clusters fall in the 0.45-0.65 range, which denotes partial dissociation and just a few dissociated clusters are present for n = 3 (Figure 4b). For all the other acids, a clear separation between dissociated and undissociated complexes is noted. Dissociated/undissociated gaps are more defined for n = 2 than for n = 3. This is consistent with a greater variety of geometries and more advanced dissociation stages as n grows. The halogen atom also plays an important role because the difference between the 13 number of dissociated and undissociated groups is more evident in going from chlorine to iodine. We now focus in the clusters with HX molecules exhibiting some degree of dissociation, say, those for which q1 > -0.8. Unlike the plots of q1 vs q2 reported previously (Gonzalez et al., 2013;Vargas-Caamal et al., 2016a), there are now two curves according to the type of interaction, X⋯H⋯O or X⋯H⋯X, the latter becoming increasingly important for each one (see Figure 6-SI). For the (HBr)2(H2O)2 clusters, the separation of curves corresponding to Br⋯H⋯O and Br⋯H⋯Br interactions is evident. Concomitantly, the number of structures with hydrogen bonds decrease (~73% for water-water and HX-water interactions), while those with dissociated forms and halogen bonds increase (~13% and 40%, respectively). The global minimum is located on the curve of Br⋯H⋯O interactions and possesses a cyclic ion-pair geometry, which belongs to the Zundel-like form (X⋯H⋯O), which represent ~13% of population. For the iodine analogs, the percentages of water-water and HX-water interactions decrease to roughly 60% and the I⋯H⋯I interactions become the most important since the global minimum belongs to this Zundel-type (notice that it is the only one structure of this category in Figure 6-SI), with the formation of a hydrogen bihalide anion (Zundel-like structure), supported by the fact that the global minimum is located at q1 = 0, corresponding to the central-symmetric anion. Structures with halogen bond interactions increase considerably (almost 74%), while the Zundel-like I⋯H⋯O structures decreases at ~9% compared with HBr-water clusters. The zoo of forms found for the (HX)3(H2O)3 clusters is more diverse. The plot for HF (Figure 6) shows that the F⋯H⋯O and F⋯H⋯F curves are closer to each other. Hydrogen bonds of the F⋯H⋯O type are longer than F⋯H⋯F. The global minimum is undissociated (an alternating six-membered ring) and belongs to the type of HF-water hydrogen bonds, the dominant interaction (95%, Figure 7-SI). The lowest energy dissociated form is a quasi-Eigen cation and has a relative free energy of 4.6 kcal mol -1 . There are no structures with halogen bonds, but some examples of dissociated (2.8%) and quasi-Eigen forms (2%) were found. F-H-O Zundel-like structures are 22.4% of the total. The gap between curves is more evident for the HCl species (Figure 6). Again, a reduction of hydrogen bond interactions (about 77% and 45% for HX-water and water-water interactions, respectively) and a significant increment of the dissociated complexes (~57%) are noted. The quasi-Eigen cations represent almost 10% of population (including the global minimum) and the structures containing Zundel-like cations is now present (about 5%), while Zundel-like Cl-H-O and halogen interactions contribute with ~5 and 1%, respectively. For the (HBr)3(H2O)3 clusters, the highest percentages belong to HX-water interactions (~77%) and dissociated complexes (66%). Although the Zundel-like X-H-X interactions have a low percentage of occurrence (about 6%), leading to the same motif for the global minima, a cage-like geometry. Contributions increased in structures with halogen bond (about 12%), Zundel X-H-O type (~14%), Zundel (10%), and quasi-Eigen (~14%) cations. Finally, in (HI)3(H2O)3 clusters, water⋯water and HX⋯water interactions decreased (about 29 and 62%, respectively). The percentage of dissociated structures is slightly lower (~61%) than those with bromine, but this is compensated by the increase of the special dissociated forms: quasi-Eigen (~18%), Zundel (11.5%), and Zundel I-H-I type (~9%). In this case, there are no Zundel-like I-H-O structures and the halogen interactions represent ~59% of the entire population. Another important remark is that the gap around q1 = 0 in Figure 6 is decreasing from clusters with Cl to I, and for the latter, there are structures populating the whole range, including zero, which accounts for structures with bihalide ions being more symmetrical than those with bromine. ## Summary and Outlook The exploration of the PES for the microsolvation of hydrogen halides with more than one HX (X = F, Cl, Br, I) molecule reveals that it is viable to dissociate the hydrogen halides with few water molecules. Particularly, HCl is dissociated with three water molecules while HBr and HI only needed two. Since two water molecules may not be sufficient to dissociate a single hydrogen halide molecule, and the same number of water molecules do that in the presence of two HX molecules, strong cooperative effects are suggested. Intriguingly, for the stronger acids, bihalide anions (BrHBrand IHI -) are formed. These fragments are persistent even a room temperature, indicating that at high acid concentrations, such species could be detected. A diverse zoo of interactions and forms is found for (HX)3(H2O)3 clusters, such as water⋯water, HX⋯water, and HX⋯HX hydrogen bonds, halogen bonds, ionic and longrange X⋯H contacts, as well as quasi-Eigen, Zundel, and Zundel-like structures. The first three types are dominant in HF-water complexes and in going from F to I, these decrease to make way for the other types of interactions, so that the dissociated forms, although smaller in percentage, can become very important because they contain the global minimum, as is the case of Zundel-like structures.

chemsum

{"title": "Acid dissociation in (HX) n (H 2 O) n clusters (X = F, Cl, Br, I; n = 2, 3)", "journal": "ChemRxiv"}

photochemistry_and_the_role_of_light_during_the_submerged_photosynthesis_of_zinc_oxide_nanorods

## Abstract: Recently, metal oxide nanocrystallites have been synthesized through a new pathway, i.e., the submerged photosynthesis of crystallites (SPSC), and flower-like ZnO nanostructures have been successfully fabricated via this method. However, the photochemical reactions involved in the SPSC process and especially the role of light are still unclear. In the present work, we discuss the reaction mechanism for SPSC-fabricated ZnO nanostructures in detail and clarify the role of light in SPSC. The results show that both photoinduced reactions and hydrothermal reactions are involved in the SPSC process. The former produces OH radicals, which is the main source of OH − at the ZnO crystal tips, whereas the latter generates ZnO. Although ZnO nanocrystals can be obtained under both UV irradiation and dark conditions with the addition of thermal energy, light promotes ZnO growth and lowers the water pH to neutral, whereas thermal energy promotes ZnO corrosion and increases the water pH under dark conditions. The study concludes that the role of light in the submerged photosynthesis of crystallites process is to enhance ZnO apical growth at relatively lower temperature by preventing the pH of water from increasing, revealing the environmentally benign characteristics of the present process.Metal oxides are one of the most widely investigated inorganic substances because they are ubiquitous in nature and frequently used in technological applications. The wide range of nanoscale forms of these materials, called metal oxide nanocrystals (NCs) that can be formed as nanowires, nanotubes, and nanorods, have gained much attention in recent years owing to their anticipated properties and application in different areas, such as photoelectron devices, sensors, catalysts, and photovoltaic devices 1-8 . Among these metal oxide NCs, zinc oxide is one of the most important natural n-type semiconductors with a direct wide band gap in the near-UV spectral region (3.36 eV at room temperature) [9][10][11] . ZnO is a promising material for fabricating devices for optoelectronics and photonics applications [12][13][14] . The optical and electrical properties of ZnO are affected by many factors, such as the structure 15,16 , size 17 , shape 18,19 , and defect concentration 19,20 , which makes ZnO a very important and interesting subject to examine the controlled growth of novel materials.In a previous study, Jeem et al. 21 reported a new pathway for the synthesis of a variety of metal oxide NCs via submerged illumination in water, called the submerged photosynthesis of crystallites (SPSC). This method is completely different from typical synthetic methods for nanoparticles, such as the hydrothermal method 22-24 , solvothermal synthesis 25,26 , and chemical vapor deposition (CVD) [27][28][29] . In the SPSC method, the initial metal is surface treated by a submerged liquid plasma process, which creates a metal nano oxide semiconducting layer with surface protrusions. After that, the growth of metal oxide NCs is assisted by a 'photosynthesis' reaction, where the metal surface is irradiated with ultraviolet (UV) light in water 21 . Thus, the SPSC process requires only light and water and does not require the incorporation of impurity precursors. Moreover, this method is applicable at low temperature and at atmospheric pressure, producing only hydrogen gas as the by-product. These characteristics give rise to the potential application of SPSC as a green technology for metal oxide NC synthesis.At present, flower-like NCs of zinc oxide 21,30 and cupric oxide 31 have been successfully synthesized using the SPSC method. Previous work 21,30,31 has discussed the reactions in the SPSC process and shown that the process is photocatalytic, accompanied by hydroxyl radical generation via water splitting. The shape of ZnO nanorods (NRs), from tapered to capped-end, could be controlled by the SPSC process, and oxygen vacancy point defects near the tip-edge of the NRs were found to be opto-electrical hotspots for light-driven formation 30 . As is well known, Zn metal itself can react with water to form Zn 2+ ions and OH − with H 2 gas generation even in dark conditions, according to the Pourbaix diagram of the Zn-H 2 O system 32 , and ZnO precipitates in alkaline aqueous environments. Therefore, ZnO NCs can also be obtained from zinc and water without illumination in the dark by controlling the pH of the water. Comparing the SPSC process with the reactions under dark conditions raises the following question: what is the role of light in the SPSC of ZnO NCs? However, this question has not been addressed, and the photochemical reactions that occur in the complicated hydrothermal process are still unclear. The present study of the photochemistry of SPSC for ZnO NC fabrication on a zinc surface was conducted by irradiation with UV light in ultrapure water, and the photochemical reaction mechanism was elucidated. Furthermore, the reactions involved in the SPSC process and under dark conditions are analyzed by monitoring the pH and temperature changes of the water, based on which a photoinduced enhancement factor is introduced to discuss the role of light in the SPSC process. ## Results and Discussion Figure 1a shows the time dependence of pH and temperature during a 72-h UV-irradiation SPSC experiment. The water temperature increased rapidly from room temperature (18 °C) to 39 °C during the first 3 h of irradiation due to energy from the UV light. After that, the temperature became relatively stable (39 ± 1°C). The pH of the water exhibited a sharp peak in the initial 4 h, after which the pH decreased to 7.0-7.6. To confirm the existence of the first sharp peak, the change in the water pH with the untreated Zn plate under UV-irradiation SPSC conditions and dark conditions was measured, and the pH curves showed similar trends. Therefore, the sharp peak was not caused by the alkaline electrolyte K 2 CO 3 solution used in the plasma pretreatment process nor by UV irradiation. This result indicates that Zn undergoes dissolution in water. According to the Zn/H 2 O Pourbaix diagram 32 , the reactions involved in the sharp pH peak are considered as Zn corrosion reactions, which will be discussed in detail later. Figure 1b shows the SEM images of the surface of the Zn plate after plasma pretreatment. Distributed over the surface are micrometer-size protrusions and numerous nanobumps. The nanobumps act as seeds for ZnO NCs and are approximately 10-20 nm in size 21 . After 1 h of UV irradiation, the nanobumps grow into ZnO nanorods (NRs) with lengths of approximately 200-400 nm (Fig. 1c, Figure S1 in the Supplementary information). The NRs show apical growth with increased irradiation time (Fig. 1d-f), and flower-like ZnO nanostructures were observed when the UV irradiation time was longer than 24 h. Figure 2 shows the XRD patterns of the specimens after surface pretreatment (Fig. 2a) and after different UV irradiation times (Fig. 2b-e). After plasma pretreatment, wurtzite structured ZnO (Zincite, JCPDS 5-0664) was observed, which confirmed that ZnO forms on the surface of the Zn plate (Zinc, JCPDS 4-0831) via the plasma pretreatment process. After UV irradiation, ZnO peaks were enhanced, and the peak intensity increased with irradiation time, which is consistent with the SEM images shown in Fig. 1c-f. When the UV irradiation time was 72 h, the highest ZnO ratio was obtained. Moreover, Zn(OH) 2 (Zinc Hydroxide, JCPDS 48-1066) peaks were also detected by XRD. As the peak position of Zn(OH) 2 is very close to that of ZnO 33 , XPS spectral analysis was utilized to confirm the formation of Zn(OH) 2 . Figure 3 displays the XPS O 1 s spectra of the specimens after different UV irradiation times. The Shirley method was performed to subtract the background. Three Gaussian-Lorentzian fitting peaks, denoted O1, O2, and O3, were used to fit the experimental data. The O1 peak located at the lower binding energy of 529.9-530.1 eV is assigned to O 2− ions involved in Zn-O bonding of the wurtzite structure of ZnO 34 . The O2 peak located at 531.1-531.8 eV is typically assigned to loosely bound oxygen on the surface OH groups 35 . The O3 peak is attributed to adsorbed water 36 . Therefore, the O2 peak proved that Zn(OH) 2 is present on the surface of the specimens as a result of the SPSC process. TEM observations of the ZnO NRs after 24 h of UV irradiation are shown in Fig. 4. The NRs have a tapered top with lengths in the range of 1-4 μm. The SAED pattern (Fig. 4d) was obtained along the direction, and the growth direction is along the c-axis, with a (001) O-terminated polar surface, which is in accordance with previous reports 30,37 . The ratios of O and Zn were obtained from their electron diffraction spectroscopy (EDS) profile, and the results are shown in the inset table in Fig. 4e. As shown in the table, the ratio of O is slightly lower than that of Zn, which presumably results from the presence of oxygen vacancies 30 . According to the above results, the photochemical reactions that occur during SPSC were deduced and a schematic illustration of the NC growth reaction mechanism is shown in Fig. 5, in which both photochemical and hydrothermal reactions contribute to the SPSC process 21,30,31 . First, electrons and holes are generated by photosemiconducting reaction (1) when UV light hits the nanobumps 38 . The generated electrons build up at the apical portion of the nanobumps to generate a cathodic environment, whereas the holes left at the bottom of the concave nanobumps create a local anode 21,39 . Photochemical water-splitting reactions (reaction (2)) then build up holes at the bottom, which subsequently contribute to OH radical generation and to the photocorrosion of ZnO (reaction (3)) 37,40 . Meanwhile, electrons accumulated at the tip. The formation of hydrated electrons (reaction (4)) induces the transform of OH radicals to OH − ions 41 and contributes to the generation of an alkaline atmosphere at the end of nanobump tip 21 aq Thus, the local separation of OH − at the apex and H + at the bottom occurs on the surface of the nanobumps (as shown in Fig. 5), which results in the apical growth of ZnO via hydrothermal reactions. The first stage occurs immediately after submerging the specimen in water. As it was previously mentioned, the peak of pH curve was observed in the first several hours under both light and dark conditions; therefore, a corrosion micro-cell was assumed to form due to the nanobumps on the surface of the Zn plate, and allowing the anode and cathode reactions 42 OH − generated by reaction (7) disperses in water and increases the water pH. At the same time, photoinduced reaction (3) also generates OH − near the surface of the nanobumps through reaction (5). Moreover, hydrogen gas generation was confirmed by gas chromatography (GC-14B, Shimadzu, Japan), as shown in Figure S2 in the Supplementary information. The photocorrosion of ZnO (reaction (3)) also occurs at this stage. Along with these reactions, the pH temporarily increases for approximately 1-2 h (Fig. 1a). Through the OH − generated by reactions ( 5) and ( 7), the following hydrothermal reaction to generate zinc hydroxide occurs, and the pH decreases. Then, the reaction of Zn → Zn(OH) 2 can be written as follows: In the second stage, after 6-10 h of UV irradiation, the pH slightly increases (Fig. 1a). The OH − generated by reactions ( 5) and ( 7) causes the pH increase to alkaline levels, thereby generating the zinc hydroxide complex ion 40,43 as shown below. In the third stage, the following ZnO crystallization reaction occurs in alkaline solution 43 . Accordingly, the net reaction of ZnO growth via SPSC is represented as follows: Thus, ZnO NRs were formed on the Zn substrate, along with hydrogen gas generation, via the SPSC process. Because of the aforementioned local separation of OH − at the apical and H + at the bottom on the surface of the nanobumps/NRs by light-driven reactions 30 , synthesis of ZnO occurs by the hydrothermal reactions between Zn 2+ and OH − at the tip of the NRs. Therefore, ZnO NRs show an apical crystal growth. Furthermore, based on thermodynamic calculations using the HSC Chemistry software (Outokumpu Research Oy, Pori, Finland), ZnO formation can occur without illumination at room temperature because the Gibbs free energies (ΔG) of reactions ( 9) and ( 12) are approximately −79 kJ and −84 kJ, respectively. We obtained similar ZnO morphologies from the dark condition experiment. Figure 6 shows a comparison of the morphology of the surface of the Zn plates under dark conditions and under UV irradiation for 24 h. Two kinds of Zn plates, untreated specimens and plasma-pretreated specimens, were used. NRs were found on the surface of the untreated specimens under both dark and UV irradiation conditions (Fig. 6a,c). Comparing the morphologies of the two specimens, the NRs formed by UV irradiation have faceted characteristics, whereas those formed under dark conditions have a leaf-like structure. By contrast, when the specimen undergoes the plasma pretreatment process, the morphologies obtained under the two conditions are very different, and flower-like structures were observed. These results show that the plasma pretreatment process produced the aforementioned protrusions on the surface of the specimen, making the fabrication of flower-like structures easy. Additionally, in Fig. 6b, most of the specimen surface is covered with a dark fibrous film, which was detected by XPS as zinc hydroxide Zn(OH) 2 . Therefore, UV light greatly affects the NC morphology. These results raise the following question: as NCs can be obtained under dark condition, what is the effect of UV irradiation on ZnO NCs, and what is the role of light irradiation in SPSC? To solve this question, additional controlled experiments were conducted. Plasma-pretreated specimens in ultrapure water were placed in different lightproof chambers. After the pH of the water stabilized, some of the submerged specimens were heated to 33-47 °C by a heater, and some were irradiated by UV light of different intensities. The pH curves of these experiments are shown in Figure S3 in the Supplementary information. Typical changes in the pH of water in these experiments are illustrated in the graph in Fig. 7a. The pH curves are divided into five stages. At room temperature and under dark conditions, the pH increased during the initial 2-4 h (stage I) and then decreased (stage II), which is similar to the results presented in Fig. 1a. After the initial sharp peak, the pH remained almost constant, and the water was slightly alkaline. In stage III, the pH and temperature changed upon UV irradiation or heating. If the water was heated to 33-47 °C under dark conditions, the pH tended to increase. By contrast, if the submerged specimens were exposed to UV light after the pH stabilized, the pH of the water decreased and trended toward neutral. At stage IV, the water temperature was stable. The pH measured in the experiments reflect the OH − concentration in the water. The photoinduced reaction formed OH − on the NR surface, which mostly aggregated near the tip of the NRs 30 , whereas the pH of the medium water was relatively low. Additionally, the balance between ZnO growth and corrosion also affects the OH − concentration in the water. Hydrothermal reactions (8), ( 10) and (11) can be written as the following net reaction (13): Reaction ( 13) is a reversible reaction. The forward reaction shows the ZnO growth, and the backward reaction is the ZnO corrosion reaction. ZnO growth consumes OH − , and corrosion generates OH − . At thermodynamic equilibrium, the forward growth reaction rate is same as the backward corrosion reaction rate, and the OH − concentration is stable, hence the pH of the water remains almost constant. In Fig. 7a, the pH is stable at the end of stage II, which shows the thermodynamic equilibrium of reaction (13). Upon increasing the water temperature by heating or UV irradiation (stage III in Fig. 7a), the change in environment breaks the thermodynamic equilibrium, which is reflected by the change in the pH of water. The increase in pH reflects that corrosion appears to dominate the reaction process. By contrast, a decrease in pH reflects that ZnO growth is dominant over corrosion. Thus, the change in pH under different conditions, as shown in Fig. 7a, shows that light irradiation makes ZnO growth dominant, and the pH of water decreases to near neutral, whereas the thermal energy from heating makes ZnO corrosion dominant, and the pH of water increases. The difference in the ZnO NC morphologies formed under the different conditions (Fig. 7b,c) shows that the NRs formed under UV light have faceted characteristics, whereas the NRs formed by heating the water under dark conditions show a corroded surface, which is in accordance with the above analysis. Moreover, the pH dependence of the water temperature in the controlled experiments is summarized as pH-T curves in Figure S4a and b, in the Supplementary information. The pH-T curves in the two figures have similar shapes. Consequently, their significant features are highlighted in Fig. 8a. When the pH is in the range of 7-11, reaction (8), , occurs according to the Pourbaix diagram of the Zn/H 2 O system 32 (Fig. 8b). The pH-T relations of reaction (8) are also given by the dotted lines in Fig. 8a, as determined from thermodynamic calculations of different Zn 2+ activities in water (the calculation is shown in the Supplementary information). As shown in the figure, the pH-T curves can be divided into four stages, which correspond with the stages in Fig. 7a. Stage I and II represent the stage at which the pH increases and decreases, respectively, similar to the pH peak in Fig. 7a, when the specimen was submerged in ultrapure water in dark conditions at room temperature. After the pH returns to a steady state, UV irradiation or heating under dark conditions will propel the pH-T curve to stage III. In stage III, the temperature sharply increases, while at the same time, the pH slightly decreases, and the curves of stage III are approximately parallel to the calculated dotted line. In stage IV, the water temperature is relatively stable, whereas the pH values of the experiments change in opposite directions. in thermodynamic equilibrium can be written as followings (the calculation is shown in the Supplementary information): Zn 2 , which is related to the Zn 2+ activity. Comparing the stage III and IV curves under the two conditions with the dotted lines in Fig. 8a, UV irradiation and heating greatly influenced the stage III and IV of pH-T curves: the former shows an increase in the Zn 2+ activity, whereas the latter shows a decrease in the Zn 2+ activity. The detail data of + a Zn 2 could be obtained from equation ( 14) by experiment data of temperature and pH. Under UV irradiation condition, + a Zn 2 is in the range of 10 −5.0 -10 −2.9 , and α is in the range of −2.5 to −1.5. Under dark condition, + a Zn 2 is in the range of 10 −5.0 -10 −2.9 , and α is in the range of −3.6 to −2.0. Equation ( 15) also shows that the reaction temperature under UV irradiation is lower than that under dark conditions at the same pH condition because of the larger α value by UV irradiation. To examine the different influences of UV irradiation and heating in dark conditions on the Zn 2+ activity, the factor α can be rewritten as , where + a Zn 0 2 is the initial Zn 2+ activity before the start of stage III (the Zn 2+ activity at controlled point in Fig. 8a), which is at a steady state. β is a factor influenced by UV irradiation or heating. Under UV irradiation, the Zn 2+ activity is almost constant in stage III and slightly increases in stage IV, and the factor β is a positive value and changes in the range of 0.2 to 0.5 in the present experiments. By contrast, under dark conditions, as the water temperature increases upon heating, the Zn 2+ activity greatly decreases in stages III and IV; thus, the factor β is a negative value and changes in the range of −1.1 to −1.3 in this work. Therefore, UV irradiation and heating under dark conditions oppositely affect β. Because β increased with UV irradiation, we define β as the photoinduced enhancement factor. Thermodynamically, the increased Zn 2+ activity can enhance the ↔ + Zn Zn(OH) 2 2 reaction, which indicates that both ZnO growth and corrosion can be activated by UV irradiation. As mentioned above, ZnO corrosion and growth occur at different positions on the nanobumps/NRs upon UV irradiation. The enhancement in these reactions by UV irradiation could enhance the apical growth of the NRs. Additionally, according to the Pourbaix diagram of Zn/H 2 O system 32 , both increasing the Zn 2+ activity and increasing the system temperature can move line c in Fig. 8b to the left. Therefore, UV irradiation could allow the ↔ + Zn Zn(OH) 2 2 reaction to occur at lower pH by increasing the + Zn 2 activity, as shown by the direction of the arrow in Fig. 8b, without greatly increasing the system temperature. Thus, from the view point of pH and temperature changes of water in the SPSC process, UV irradiation could enhance the photocorrosion of ZnO and induce the apical growth of NRs by increasing the Zn 2+ activity at relatively low temperature by preventing an increase in the pH of water, to sustain a neutral environment. This result reveals the environmentally friendly characteristics of the SPSC process. For the above reason, we observed the growth of ZnO NCs along the c-axis with respect to the UV irradiation time, as shown in the SEM images in Fig. 1. However, when the UV irradiation time was longer than 72 h, further ZnO growth was not observed. (Figure S5, Supplementary information). Accordingly, the NC growth was examined by evaluating the weight change of the specimens. Figure 9 shows the weight change of the Zn plate during UV irradiation. The black line is the average value from the interrupted experiment. For both the interrupted and uninterrupted experiments, the weight of the samples increased with UV irradiation time during the initial 60-96 h in increments of less than 0.2 wt%. Most of the interrupted specimens show a lower weight increment compared with the uninterrupted specimens. This effect was caused by the delamination of the NCs from the substrate Zn plate during the removal of the samples every several hours. During the stage at which the weight increased, the NC growth reaction was dominant. However, the weight increase of all the samples did not persist, and the weight decreased after 60-96 h of UV irradiation. After 144 h of UV irradiation, many exfoliated NCs were observed on the specimen, and flower-like ZnO NCs were no longer present (Figure S5, Supplementary information). In a previous report 30 , oxygen vacancy point defects were found to exist near the tip-edge of the NRs, and the ratio of O to Zn in the NRs increased with UV irradiation time. After 72 h of UV irradiation, the tip of NRs had a flat, capped shape, and the ratio of O to Zn was very close to 1. Therefore, the ZnO growth in the apical direction is dominant compared with ZnO photocorrosion until 72 h of UV irradiation. After that, further UV irradiation makes the ZnO corrosion reaction dominant on both the tip and bottom of the NRs, which result in the NCs having a corroded surface (Figure S5, Supplementary information). Further photocorrosion by UV irradiation leads to the exfoliation of NCs from the substrate Zn plate (Figure S5b,c, Supplementary information). Based on the above discussion, SPSC involves photoradical reactions, which is the main source of OH − for ZnO apical growth. To elucidate such photoradical reactions in the SPSC process, two additional experiments of gamma-ray irradiation and SOD addition were conducted. Gamma-ray irradiation can efficiently generate radiolysis products of water, such as OH, e − aq, H(H 2 ), H 2 O 2 , and H 3 O + 44 . Figure 10 shows the morphologies of the Zn specimens in the gamma-ray irradiation SPSC experiment. The dose rate was in the range of 0.9-10.0 kGy•h −1 , and the irradiation time was 24 h and 30 h. At a dose rate of 0.9 kGy•h −1 , a dark fibrous Zn(OH) 2 structure was observed on part of the surface of the specimen, and NRs were not observed after 24 h of irradiation. When the irradiation time was increased to 30 h, flower-like nanostructures grew from the Zn(OH) 2 substrate. When the dose rate was increased to 3.5 kGy•h −1 , fine nanoparticles could be observed after 24 h of irradiation, and flower-like NRs were generated within 30 h of irradiation. However, when the dose rate was increased to 10.0 kGy•h −1 , most of the nanoparticles were covered with a dark fibrous Zn(OH) 2 film after 24 h of irradiation. In addition, a large-area Zn(OH) 2 layer over the NRs was also observed when the irradiation time was increased to 30 h (Fig. 10f) and 48 h (Figure S6a, Supplementary information) under 10.0 kGy•h −1 , and the re-deposition of ZnO was not observed. Therefore, the Zn(OH) 2 layer on the NRs restricted the further growth of ZnO nanoparticles. As mentioned earlier, when Zn(OH) 2 formed on the tips of the NRs, reactions ( 9) and (10) for the process Zn(OH) 2 → − Zn(OH) 4 2 →ZnO occurred. In this case, a local alkaline environment near the nanobumps is needed. However, after gamma-ray irradiation condition, the pH of water was in the range of 4.1-6.6. Once the Zn(OH) 2 layer formed over a relatively large area, the local OH − concentration decreased, and the further formation of − Zn(OH) 4 2 was difficult. We speculate that this effect is the main reason for the restriction of the further growth of NRs. Therefore, under gamma-ray irradiation at a relatively lower dose rate, fine nanoparticles were obtained, and the nanoparticles grew with irradiation time. In general, fine crystal submerged synthesis can be induced by gamma rays with high linear-energy-transfer (LET) radiation. Additionally, Fig. 11 shows the SEM images of the specimen surfaces with SOD addition after UV (Fig. 11a) and gamma-ray irradiation (Fig. 11b). No NCs were found in either UV-irradiated specimen or the gamma-ray irradiated specimen with SOD addition. The SOD reagent is known to capture O 2 − 45, 46 . Compared with the specimens without SOD reagent addition (Fig. 1e and Figure S6b in the Supplementary information), the specimens without SOD addition showed NCs formation on the surface, whereas the crystal formation was suppressed in the specimens with SOD addition. This result suggested that O 2 − plays an important role in ZnO growth and revealed that crystal formation via SPSC was caused by photolysis or radiolysis in oxygenated water. ## Conclusion The SPSC method, proposed as a new green technology that requires only water and light, can be applied to produce ZnO NCs. We carried out surface microstructural analysis and monitored the pH and temperature change of the water during the process. Based on this analysis, ZnO NCs formation mechanism was clarified. Furthermore, the effect of light irradiation during the process was elucidated by contrast experiments under dark conditions. It shows that both photoinduced reactions and hydrothermal reactions contribute to the SPSC process. The former generates OH radicals, which are one of the main sources of OH − at the crystal tips, whereas the latter involve the ZnO growth reactions of Zn 2+ → Zn(OH) 2 → − Zn(OH) 4 2 → ZnO. Furthermore, ZnO NCs can be obtained under both light irradiation and heating in dark conditions, in which ZnO growth and corrosion reactions occur simultaneously. Light irradiation makes ZnO growth dominant and the water pH close to neutral, whereas thermal energy makes ZnO corrosion dominant and the water pH increases. The role of light in SPSC process is to enhance ZnO apical growth at relatively lower temperature by preventing the pH of water from increasing, revealing the environmentally benign characteristics of the present process. ## Methods Surface Pretreatment. Surface pretreatment was conducted in a submerged liquid plasma experiment device (shown in Figure S7a, Supplementary information). A platinum wire (length: 1000 mm, diameter: 0.5 mm, 99.98 mass%, Nilaco, Japan) served as the anode. The cathode comprised the zinc plate (35 × 5 × 0.5 mm, 99.5%, Nilaco, Japan) wired with copper wire (diameter: 0.5 mm, 99.99%, Nilaco, Japan). Voltage was applied using a direct current (DC) power supply (ZX800H, Takasago). The electrolyte was a 100 mol•m −3 K 2 CO 3 solution. The electrolyte temperature was measured by a polymer-coated thermistor thermometer. The current and the voltage were measured by the DC power supply. The plasma treatment was carried out in the range of 130-140 V for 10 minutes. The pretreated Zn plates was washed by deionized water and then was cut to a length of 20 mm for subsequent SPSC experiment. SPSC Experiment. Two types of the SPSC experiments were conducted: UV irradiation and gamma-ray irradiation. During the UV-irradiation SPSC experiment, a plasma-treated Zn plate was placed into a 4-mL cuvette with ultrapure water (Wako Pure Chemical, Japan), which was deaerated by boiling, and then irradiated by a UV lamp (UVP, B-100AP, USA, λ = 365 nm, 3.4 eV) for 0-144 h in a lightproof chamber (shown in Figure S7b, Supplementary information). The intensity of the UV irradiation was 10−53 mW•cm −2 . During the SPSC experiment, the pH and temperature of the water were measured using a pH/ORP meter (Horiba, LAQUA, D-72) containing a micro ToupH electrode (Horiba, LAQUA, 9618 S) and a long ToupH electrode (Horiba, LAQUA, 9680S-10D). In the gamma-ray irradiation SPSC experiment, the plasma-treated Zn plate was placed into a test tube with 3 mL ultrapure water and then irradiated with gamma-rays, which was performed at the 60 Co irradiation facility of the Institute of Scientific and Industrial Research (ISIR) at Osaka University. The gamma-ray dose rate was determined by the distance from the sample to the ray source. The absorbed dose was calculated by Fricke dosimetry. Dark Condition Experiment and Superoxide Dismutase (SOD) Addition Experiment. The dark condition experiment was conducted in a lightproof chamber without illumination, in which the Zn plate was immersed in deaerated ultrapure water for several hours. In the dark condition experiment, a heater was used to control the water temperature. The SOD addition experiment was performed using both UV and gamma-ray irradiation SPSC experimental procedures. SOD from bovine erythrocytes (Sigma-Aldrich, USA) was dissolved in ultrapure water at a concentration of 1.4-3.0 g/100 mL. Weight Change Measurement. Two types of weight change measurements, interrupted and uninterrupted experiments, were conducted by a microbalance (AEM-5200, Shimadzu, Japan). The interrupted experiment involves the weight measurement of a sample at different UV irradiation times. The sample was removed from the cuvette for measurement after a certain period of UV irradiation and then returned to the cuvette to continue the irradiation. The uninterrupted experiment measures weight change of different samples UV irradiated for different times. Physical Characterization. X-ray diffraction (XRD) patterns of the samples were obtained using an X-ray diffractometer (Rigaku, Miniflex) equipped with a Cu Kα source operating at 40 kV and 15 mA. The surface morphologies were observed by field emission scanning electron microscopy (FE-SEM, JSM-7001FA, JEOL). Transmission electron microscopy (TEM) and selected area electron diffraction (SAED) patterns for the crystal were obtained using a conventional transmission electron microscope (JEM-2000FX, JEOL) operated at 200 kV. The nanoparticles were characterized using an X-ray photoelectron spectroscopy (XPS, JEOL, JPS-9200), equipped with a monochromatic Al Kα X-ray source (1486.6 eV). The analyzed area of the samples was 3 mm × 3 mm (large scale). The peak positions and areas were optimized by a weighted least-squares fitting method using 70% Gaussian and 30% Lorentzian line shapes. All XPS spectra were calibrated to the C (1 s) core level peak at 286.0 eV.

chemsum

{"title": "Photochemistry and the role of light during the submerged photosynthesis of zinc oxide nanorods", "journal": "Scientific Reports - Nature"}

aqueous_mechano-bactericidal_action_of_acicular_aragonite_crystals

## Abstract: Nanoneedle structures on dragonfly and cicada wing surfaces or black silicon nanoneedles demonstrate antibacterial phenomena, namely mechano-bactericidal action. These air-exposed, mechano-bactericidal surfaces serve to destroy adherent bacteria, but their bactericidal action in the water is no precedent to report. Calcium carbonate easily accumulates on solid surfaces during long-term exposure to hard water. We expect that aragonite nanoneedles, in particular, which grow on TiO 2 during the photocatalytic treatment of calcium-rich groundwater, exhibit mechanobactericidal action against bacteria in water. Here, we showed that acicular aragonite modified on TiO 2 ceramics prepared from calcium bicarbonate in mineral water by photocatalysis exhibits mechanical bactericidal activity against E. coli in water. Unmodified, calcite-modified and aragonite-modified TiO 2 ceramics were exposed to water containing E. coli (in a petri dish), and their bactericidal action over time was investigated under static and agitated conditions. The surfaces of the materials were observed by scanning electron microscopy, and the live/dead bacterial cells were observed by confocal laser scanning microscopy. As a result, the synergistic bactericidal performance achieved by mechanobactericidal action and photocatalysis was demonstrated. Aragonite itself has a high biological affinity for the human body different from the other whisker-sharpen nanomaterials, therefore, the mechanobactericidal action of acicular aragonite in water is expected to inform the development of safe water purification systems for use in developing countries.It was recently revealed that the surfaces of dragonflies, cicada wings, and the skins of gecko's feature nanoneedle structures that show antibacterial activity 1-6 . Reported artificial reproductions of these biomimetic structures, demonstrate similar antibacterial activity, known as the mechano-bactericidal effect [7][8][9][10][11][12][13][14][15][16][17] . Studies are steadily elucidating these mechano-bactericidal mechanisms. The main consequences of mechano-bactericidal actions are the sterilization of solid surfaces and the inhibition of biofilm formation. A significant advantage of the mechano-bactericidal mechanism is the absence of chemical reagents; this property of nanoneedle biomimetic structures is garnering the attention of scientists for their potential to feature in environmentally friendly and sustainable bactericidal technologies.Natural water typically contains various mineral components. Many are familiar with the white precipitate (scale) that forms and adheres to faucets; it develops when the concentrations of certain mineral components in tap water are relatively high. This white precipitate is calcium carbonate, with the crystal structure of mainly aragonite or calcite; significantly, aragonite may have a needle crystal habit [18][19][20] . In a previous study, we found that calcium carbonate forms on the surface of TiO 2 photocatalysts during the photocatalysis of calcium bicarbonate contained water, moreover, the calcium carbonate mainly demonstrated the crystallinity of aragonite with nanoto micrometer-sized needles 21 . We expected that these aragonite nanoneedles would continuously kill any bacteria in flowing water upon contact through the mechano-bactericidal action associated with their topography. As mentioned above, the most famous mechano-bactericidal efficiency is the sterilization of bacteria adhered to the solid surface, and almost unknown the elimination of bacteria in flowing water by mechano-bactericidal effect. It would be a new discovery if the aragonite needle crystal has a mechano-bactericidal effect in the water. The mechano-bactericidal action of acicular aragonite in water will be expected to inform the development of water purification systems for use in developing countries in which countries have a problem with safe water access. One of the reasons is that calcium carbonate itself is a non-toxic compound, in addition, this is because calcium bicarbonate, which is a raw material for calcium carbonate, is relatively universally contained in groundwater. That is to say, it is also expected that the calcium bicarbonate in groundwater will use to photocatalytic repair ## Results and discussion Growth of aragonite and calcite on TiO 2 ceramic surfaces. Calcium carbonate accumulates on the TiO 2 ceramic surface, as shown in Fig. 1, during the long-term circulation of mineral water containing calcium bicarbonate under UV irradiation (Fig. 1A). Mineral water containing only calcium bicarbonate produced a precipitate characterized by hexagonal crystal (including various large and small disphenoids and 8-faced scalenohedron crystals) (Fig. 1B right), while mineral water (such as Evian and Contrex) containing not only calcium bicarbonate but also magnesium and strontium ions produces a precipitate characterized by orthorhombic acicular crystals (Fig. 1B left). As shown in Fig. 1C, the X-ray diffraction (XRD) (a and b) and laser-Raman spectra (c), respectively, reveal that the crystal structures of the hexagonal and orthorhombic acicular calcium carbonate are calcite and aragonite, respectively. Acicular aragonite crystals did not form on the TiO 2 ceramic surface when mineral water was circulated over the photocatalyst in the absence of UV irradiation. The lengths of the acicular crystals ranged from 10 nm to a few micrometers, depending on the duration of circulation and the concentration of calcium bicarbonate in the mineral water. It is expected that the large (micrometer) acicular crystals will capture the bacteria in a water-flow system, while the small (sub-micrometer) acicular crystals exert a mechano-bactericidal effect. shown in Fig. 2. The number of E. coli cells did not change significantly in a saline water system. In a system containing unmodified TiO 2 ceramics, the number of E. coli cells decreased slightly. In a system containing aragonite-modified TiO 2 , the number of E. coli cells decreased remarkably with or without shaken. The decrease in the number of bacterial cells under static conditions indicates that E. coli was captured by acicular aragonite owing to its own motor function, and the improvement in the antibacterial action achieved by agitation was less significant than the antibacterial action associated with the motor function of the bacteria. Figure 3 display SEM images of the surfaces of unmodified, calcite-modified, and aragonite-modified TiO 2 ceramics that were immersed in saline water containing E. coli for 16 h. The distinctive shapes of E. coli cells on the unmodified (Fig. 3 (1)) and calcite-modified TiO 2 ceramic surfaces (Fig. 3 (2) and (2′)) did not change after 16 h. In contrast, the distinct shape changes of E. coli cells were observed on the aragonite-modified TiO 2 ceramic surfaces. In Fig. 3 (3) and 3 (3′), the cell membrane was observed to be stretched and leathery. In Fig. 3 (3″), a situation was observed in which a substance that appeared to be protoplasm was ejected from the stab wounds of E. coli. 11,22 , resulting in sterilization. Wu et al. considered the relationship between the length of the nanoneedle that penetrates the bacterial-cell wall and the inter-needle distance 9 . They reported that the stretching of the cell membrane increases with the increasing density of the nanoneedles. Their experiment investigated the bacteria adhere to mechano-bactericidal solid surface; however, we expect that the bactericidal mechanism of us in water phase should be similar to them because we observed bacteria impaled on the acicular aragonite. We also investigated the mechano-bactericidal performance of acicular aragonite with different nanoneedle sizes. As shown in Fig. 5, the mechano-bactericidal performance of acicular aragonite was dependent on the size of the nanoneedles. This result is shown in Table 1. This needle-size dependence of mechano-bactericidal performance is consistent with the results of earlier studies for the solid surface 7,9,11,17,22 . Figure 6 reveals the mechano-bactericidal action in circulation systems. Unmodified, calcite-modified, and aragonite-modified TiO 2 ceramics were packed into Pyrex glass tubes (300 mm × 10 mm (internal diameter)) and water containing E. coli (5 × 10 3 CFU/mL) were circulated through these tubes at a rate of 50 mL/min. Figure 7a-c show images of the live/dead E. coli cells on the surfaces of the unmodified, calcite-modified, and aragonite-modified TiO 2 ceramics, respectively, after circulating the aqueous phase for 3 h. A significant amount of viable E. coli cells was observed on the surface of the unmodified TiO 2 ceramics (Fig. 6a). The total number of bacterial cells on the surface of the calcite-modified TiO 2 ceramics (Fig. 6b) was less than that on the surface of the unmodified TiO 2 ceramics; however, both viable and dead bacteria were observed. As shown in Fig. 6c, very few dead bacterial cells were observed on the surface of the aragonite-modified TiO 2 ceramics, despite the presence of some viable bacteria; this result was consistent with the result obtained under static conditions (Fig. 4). The most significant finding of this research is the underwater mechano-bactericidal action of acicular aragonite. Almost all of the reported nanoneedle, nanopillar, and whisker-shaped materials exhibit mechano-bactericidal action. However, it is known that asbestos, potassium titanate, carbon nanotubes, and metal nanowires demonstrate lung toxicity due to oxidative stress induced by their shape . On the other hand, it is known that acicular aragonite does not show toxicity toward lung tissue, unlike the aforementioned nanoneedle, nanopillar, and whisker-shaped materials 18 . Aragonite, which is composed of calcium carbonate, easily dissolves in the www.nature.com/scientificreports/ living tissue and, as a result, morphology-induced toxicity does not manifest. In the course of water treatment through the mechano-bactericidal action using aragonite-modified TiO 2 ceramics, we must assume the defluxion of nanoneedles into the water as a result of fracturing. Unlike body-soluble acicular aragonite, nanoneedle, nanopillar, and whisker-shaped nanomaterials are unsuitable for drinking water treatment because of the risk of fragment effluence. We have already mentioned the possibility of damage to acicular aragonite during water treatment; however, it is also expected that the acicular aragonite will recover/self-repair using components in natural water (especially groundwater). More specifically, the self-replication ability of the system, and its mechano-bactericidal action, is anticipated. In fact, we found that acicular aragonite nucleates from dead E. coli cells during long-term circulation in Contrex system (Fig. 7). Combination performance of mechano-bactericidal effect and photocatalytic sterilization. Photocatalytic environmental purification can only be performed during the daytime and, for optimum performance, under sunny conditions . However, ideal photocatalytic drinking-water purification systems for developing countries should achieve absolute performance under all weather conditions, not only sunny but also cloudy or rainy; possibly by combining photocatalytic activity and mechano-bactericidal action. Figure 8 shows the change in number of E. coli cells in circulation systems featuring tubes containing unmodified, aragonite-modified and calcite-modified TiO 2 ceramics under dark and UV light conditions. The number of E. coli cells in the unmodified TiO 2 ceramic system decreases significantly more under UV-A irradiation than dark conditions due to photocatalysis 30,31 . However, the number of E. coli cells in the aragonite-modified TiO 2 ceramic system under dark conditions is lower than that in the unmodified TiO 2 ceramic system under UV-A irradiation. The point to note is that the absolute rate of E. coli cell reduction in the aragonite-modified TiO 2 ceramic system under UV-A irradiation (Sterilization rate const. = 1.22 h −1 ) was ~ 2.0 times of under dark (Reduction rate const. = 0.809 h −1 ), and ~ 3.0 times of the UV light conditions in the unmodified TiO 2 ceramic system (Photocatalytic rate const. = 0.618 h −1 ) as shown in Table 2. In a previous study, we revealed why the photocatalytic activity of TiO 2 does not decrease with increasing aragonite accumulation on its surface 21 . The photocatalytic reaction mainly involves the formation and migration 32 , is considered sufficient to allow it to migrate into the aqueous phase through the layer of aragonite. The densification of the aragonite layer is expected to impede the migration of active species. In reality, the aragonite layer on the TiO 2 surface is very porous and a few micrometers thick. This porosity facilitates the migration of the active species, generated by photocatalysis, to the surface of the aragonite layer. This is one of the reasons why the photocatalytic sterilization performance of the aragonite-modified TiO 2 ceramic photocatalyst exceeds that of the unmodified TiO 2 ceramic photocatalyst. Moreover, the mineralization of bacteria on the aragonitemodified TiO 2 ceramic photocatalyst is promoted by photocatalysis; as a result, the surface of this material will maintain a clean condition unlike that of an aragonite system without photocatalytic materials. In this experiment, we also carried out using the calcite-modified TiO 2 ceramic. However, photocatalytic sterilization could not be achieved not only in dark conditions but also in UV irradiation, as shown in Fig. 8. Calcite formed a large hexagonal crystal different from a case of acicular aragonite, and it is considered that this calcite covered the TiO 2 surface and prevents the diffusion of active species generated by UV irradiation. The situation in which calcite plays a lid for the TiO 2 ceramic with a large surface area can be also inferred from the specific surface area observation results in Table 2. As described above, we demonstrated the mechano-bactericidal treatment of water by acicular aragonite. However, several aspects require elucidation, such as the relationship between the optimum needle size of aragonite or flow speed and bactericidal performance, acicular aragonite associated toxicity against live body except for lung toxicity, and the mechano-bactericidal action of acicular aragonite against other bacterial species. As previously reported, the TiO 2 ceramic photocatalyst, which is used as a substrate for acicular aragonite growth, www.nature.com/scientificreports/ is extremely strong and does not deteriorate during long-term use. Therefore, it is expected that access to safe water can be achieved in many developing countries with systems that combine mechano-bactericidal action and photocatalysis, such as the proposed acicular aragonite-modified TiO 2 ceramic system, that obviate the use of disinfectants and concomitant chemical risks and high running costs. ## Conclusion Acicular aragonite nano-needle, a metastable phase of calcium carbonate, precipitated on the photocatalytic surface by the photocatalytic reaction of calcium bicarbonate contained water such as ground water, showed a mechano-bactericidal effect in water which is not still known. Calcite, which is a stable phase of calcium carbonate, did not show any mechano-bactericidal effect. From the results of phase contrast microscopy, the aragonite phase showed a low density of dead bacteria on its surface, which may be due to the leakage of protoplasm to be stained into the water. Although the mechano-bactericidal effect alone was greater than the photocatalytic effect alone, the synergistic effect was observed for the aragonite modified photocatalyst, and the bactericidal rate in water was equal to the sum of the mechano-bactericidal and photocatalytic effects. The aragonite modified photocatalyst is expected to be used for drinking water purification in developing countries, the loss of aragonite needle crystal habit may occur due to its long-term use. However, it was confirmed that aragonite was precipitated from dead bacteria after long-term use of the aragonite modified photocatalyst in hard water, indicating that the material has a self-regenerating function to maintain its mechano-bactericidal effect and is a promising material for realizing safe water access in developing countries. ## Methods Substrates for CaCO 3 growth. The preparation of a TiO 2 ceramic photocatalyst employed as a substrate for the growth of CaCO 3 crystals, such as aragonite, has previously been reported 21 . TiO 2 is known as a photocatalytic material that has generally two crystal phases, anatase and rutile (sometimes brookite is also mixed). The photocatalytic activity of anatase is high, higher than that of rutile . Unless specified otherwise, all of the experimental procedures were carried out in the absence of UV light; in particular, we used high temperature treated (750 °C) rutile TiO 2 ceramic as a substrate for aragonite modification to reduce the risk of undesired photocatalytic reactions. On the other hand, anatase TiO 2 ceramics calcined at 550 °C were used in the photocatalytic (under UV irradiation) experiments. This ceramic photocatalyst is generally more robust in water than other photocatalytic materials and, since its semipermanent use in the water is expected, very suitable for the treatment of drinking water in developing countries 36 . Growth of aragonite and calcite on TiO 2 ceramic surfaces. Commercial mineral water is the most suitable starting material for the growth of acicular aragonite crystals on TiO 2 ceramics. Certain brands of mineral waters are rich in calcium bicarbonate (as indicated by their ingredients labels of commercial products). While these mineral waters contain various minerals, only one precipitate, calcium carbonate, is generally formed. We used Contrex from France since this mineral water contains high concentrations of calcium bicarbonate and its pH is almost neutral. In contrast, the precipitation of a calcite reference sample required a pure calcium bicarbonate solution. The calcium bicarbonate solution employed in this study was the filtrate of a solution prepared by bubbling CO 2 gas through a calcium hydroxide-saturated solution 37 . To grow aragonite (or calcite) on TiO 2 ceramics, 500 mL of Contrex (or calcium bicarbonate solution) was introduced into a water flow line connected to a glass tube packed with TiO 2 ceramics under UV irradiation. The circulatory system was the same as in our previous report 21 . After circulating for 8 h, the circulation system and content were dried under airflow for 16 h; this circulation and drying process was repeated until 15 times. Since the CaCO 3 crystals grow larger by repeating this process, we obtained crystals of different sizes as shown in Fig. 5 and Table 1 by dividing the process into five steps. The crystalline structures of these precipitates were determined using X-ray diffraction (XRD; D2 Phaser, Bulker Germany) and laser Raman spectroscopy (NRS-4500, JASCO, Tokyo, Japan). The morphologies of the surfaces of these materials were observed by field emission scanning electron microscopy (FE-SEM; S-4700, Hitachi, Tokyo, Japan) 38 . The E. coli growth curve was determined from the optical density at 600 nm measured with a cell density meter (CO8000, Biochrom, Cambridge, UK). The E. coli was cultured until the system reached a stationary state at 0.5-1 × 10 8 CFU/mL. The cultured E. coli was diluted once with phosphate-buffered saline, and 0.1 mL of the diluted/undiluted culture was used to inoculate the NBNaCl culture media plates. The E. coli was cultured on the NBNaCl media at 35 °C for 24 h. The number of colonies was counted using a colony counter and the average number of colonies was calculated. If the number of colonies per dilution factor differed by more than a factor of 2, the value for the culture Fluorescent staining. SYTO9 is a membrane-permeable DNA staining reagent that stains the nuclear DNA of viable bacteria without membrane damage and emits 500 nm (green) fluorescence when exposed to light at a wavelength of 483 nm or lower. On the other hand, PI is not membrane-permeable, therefore it stains nuclear DNA of dead bacteria with membrane damage and emits fluorescence at 617 nm (red) when irradiated with light at a wavelength of 536 nm or lower. Based on these differences in fluorescence wavelengths, we determined the viability of E. coli adsorbed on the sample surface 39 . The final concentrations of SYTO9 and PI were prepared at 6 µmol/L and 30 µmol/L, respectively, and their mixed solution was used as the fluorescent staining solution (L/D reagent) in the viability determination. A drop of 0.1 mL of L/D reagent was added to the sample treated and the surface was covered, and the sample was allowed to stand for 15 min. The staining process was carried out in the dark to eliminate the effect of quenching of the fluorescent reagent. ## Biological experiments. Fluorescence observation and image analysis. Observation of the fluorescent staining samples was performed by confocal laser microscope (LSM880; Carl Zeiss, Oberkochen, Germany). Observation was performed while the samples were immersed in sterilized ion-exchange water to eliminate the possibility of E. coli being killed by drying. During the observation, the fluorescence emitted by E. coli adsorbed on the sample surface was captured by irradiating the sample with a laser of a specific wavelength. From the images obtained by combining green and red fluorescence, we determined the viability of E. coli. The percentage of L/D was analyzed by ImageJ image analysis software. The green and red areas of the obtained images were split using the "Split Channels" function, and the percentage of alive or dead was calculated from the total number of pixels. ## Mechano-bactericidal experiment (incl. photocatalysis). To analyze the mechano-bactericidal action of the materials, we performed one type of static and two types of dynamic experiments. In the static experiment, cultured E. coli (1 × 10 7 -10 8 CFU/mL) solution was fed into culture tubes (volume: 5 mL), containing one piece of unmodified TiO 2 ceramic, calcite-modified TiO 2 ceramic, or aragonite-modified TiO 2 ceramic, that was then left at rest at 4 °C for 16 h. The E. coli cells on the surface of each substrate were observed by confocal microscopy, after staining the cells with SYTO9 and PI, to estimate the live/dead bacterial cell abundance and evaluate their membrane integrity. In addition, the E. coli cells on the surfaces were observed by FE-SEM after fixation with glutaraldehyde 40,41 . The first dynamic experiment was a time-course experiment; 50 mL of a solution containing E. coli (~ 5000 CFU/mL) was poured into petri dishes containing unmodified or aragonite-modified TiO 2 ceramics and sampling was performed every 30 min for 3 h at 25 ºC under static and agitated conditions in a shaking incubator. For reference, this procedure was repeated without any substrates (blank). The initial number of E. coli cells in the respective samples, were differed between confocal microscopy and culture method. This differences are attributed to the different of the detection limits of confocal microscopy associated with these culture methods. In the second dynamic experiment, to simulate a practical application, we circulated 250 mL of saline water (0.5 g/L) containing E. coli (5000 CFU/mL) through a glass tube packed with aragonite-modified TiO 2 ceramics for 3 h at a flow rate of 50 mL/min. We also used the black-light blue fluorescent lamp (15W-BLB) for photocatalytic reaction as the light source when the photocatalytic experiment was carried out, and the UV-A (λ = 365 nm) intensity was fixed at 2 mW/cm 2 at the position of the photocatalyst-packed tube. The UV-A intensity was measured by means of a UV power meter C9356-1 (Hamamatsu Photonics, Hamamatsu, Japan). The solution was sampled every 30 min, and 0.1 mL of each sample was used to inoculate the NBNaCl culture media plates and cultured at 35 °C for 24 h. After each 3-h circulation, the apparatus was drained under air pressure and disinfected by circulating ethanol (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) for two 30-min periods. Milli-Q water was passed through (not circulated) the apparatus for 2 h; the apparatus was again drained under air pressure before the aragonite-modified TiO 2 ceramics in the apparatus was dried overnight under flowing air and UV irradiation (λ = 352 nm) to eliminate any residual organic compounds and bacteria on the surface of the TiO 2 by photocatalysis. After the circulation experiment, aragonite-modified TiO 2 ceramic was abstracted from the glass tube and the bacteria were observed by FE-SEM after fixation with glutaraldehyde.

chemsum

{"title": "Aqueous mechano-bactericidal action of acicular aragonite crystals", "journal": "Scientific Reports - Nature"}

towards_a_first-principles_evaluation_of_transport_mechanisms_in_molecular_wires

## Abstract: Understanding charge transport through molecular wires is important for nanoscale electronics and biochemistry. Our goal is to establish a simple first-principles protocol for predicting the charge transport mechanism in such wires, in particular the crossover from coherent tunneling for short wires to incoherent hopping for longer wires. This protocol is based on a combination of density-functional theory with a polarizable continuum model introduced by Kaupp et al. for mixed-valence molecules, which we had previously found to work well for length-dependent charge delocalization in such systems. We combine this protocol with a new charge delocalization measure tailored for molecular wires, and we show that it can predict the tunneling-to hopping transition length with a maximum error of one subunit in five sets of molecular wires studied experimentally in molecular junctions at room temperature. This suggests that the protocol is also well suited for estimating the extent of hopping sites as relevant, e.g., for the intermediate tunneling-hopping regime in DNA. ## Introduction The idea of using molecular wires as building blocks for nanoscale electronics has attracted the interest of both theoreticians and experimentalists, as it offers the possibility of establishing novel functionalities compared to conventional silicon-based electronics. 1 For instance, the potential application of molecular wires as single-molecule insulators with even greater insulating properties than vacuum has been pointed out recently. Moreover, approaches for exploiting the spin-polarization properties of diamagnetic helical molecules, such as proteins or DNA, have been suggested in the past, where chiral-induced spin selectivity can be used to design more efficient water-splitting or memory devices. Besides potential technological applications, the field of molecular electronics is appealing due to its significance for fundamental science, offering insights into molecules under unusual circumstances. There are continuing efforts to study charge transport processes in biomolecules like peptides, enzymes and DNA, as they are of vital importance to every living organism, for example in the context of oxidative DNA damage. A central question in such processes is the charge transfer mechanism. From conductance experiments on molecular wires built of a varying number of repeating monomer units, two main transport regimes have been identified, coherent tunneling and incoherent hopping. 1,15,16 One of the main factors determining the predominant charge transport mechanism is the molecular length. A crossover from the tunneling to the hopping regime is frequently identified as a sudden reduction of length dependence of conductance for wires longer than around 3 to 4 nm. 15,16 In the tunneling regime, the charge is transported coherently in a one-step process, while essentially not spending any time on the molecular bridge. 1,20 This transport mechanism is characterized by an exponential dependence of the conductance on molecular length. In contrast, as the molecule becomes longer, the transition to incoherent hopping is marked by a linear dependence of the conductance on the length in accordance with Ohmic behavior. 1,15,16 As the latter process is strongly dependent on temperature in contrast to tunneling 1 , the transition is also marked by a change in temperature dependence. 1 In the hopping regime, the charge transport timescale approaches the one of molecular vibrations, where dephasing processes associated with electron-phonon coupling can result in the formation of polarons, i.e. charge-localized deformations on specific molecular subregions. 1,20,22 As a consequence, the charge migrates through the molecule by subsequently moving through so-called hopping sites. Experimental data also point to an intermediate regime, in which hopping sites extend over large portions of the system, e.g. in DNA The groups of Elstner and Kleinekathöfer presented a comprehensive multiscale approach to the description of charge transport in molecular junctions without the need of assuming an underlying transport mechanism a priori. 27,28 Related approaches have been put forward in the context of charge transport in organic crystals. There is also a substantial body of work on such comprehensive approaches in molecular junctions employing simple parametrized model Hamiltonians. 25, However, charge delocalization in molecules can depend strongly on choices for approximations in the electronic structure description. 44? -48 If an accurate first-principles treatment of the electronic structure is desired, such comprehensive schemes would be computationally quite expensive. It also requires a computational implementation combining electronic structure calculations, 49,50 Greens function techniques 51,52 and molecular dynamics simulations. 53,54 For a computationally efficient and simple-to-implement first-principles description of charge transport processes, the identification of the crossover length can therefore be a valuable alternative, as it allows assuming either of the two transport regimes, leading to simpler theoretical descriptions of transport. Here, we are aiming at a predictive approach for crossover lengths that is computationally efficient and easily applicable to any kind of molecular wires. The methodology presented in this study relies on a connection between the predominating charge transport mechanism and charge localization properties: One could link tunneling transport to a completely delocalized charge, while increasing charge localization could be associated with hopping transport. 20,22 The charge is then predominantly located on specific subregions of the molecule, where polarons are formed. 15,55,56 The molecular wires under study are considered in their singly oxidized radical-cationic state, as holes rather than electrons dominate the transport in the wires under study here. In contrast, there are much fewer examples of charge transport through molecular wires dominated by electrons (corresponding to radical-anionic organic mixed-valence species 57 or electron-deficient thiophene-1,1-dioxide oligomers). 58 In a previous study on first-principles approaches, we validated different computational protocols based on Kohn-Sham density-functional theory (KS-DFT) regarding their capability of describing length-dependent charge localization in comparison with experiments for organic mixed-valence systems. 48 This type of donor-bridge-acceptor systems has the advantage of representing atomistically better-defined model systems than molecular junctions, while still being closely related to them as the electron transfer properties as expressed in their Robin-Day class are also length-dependent. 59,60 We found that, in contrast to other DFT protocols, a combination of the B1LYP hybrid functional with 35% Hartree-Fock exchange (BLYP35) and a polarizable continuum model (PCM), previously proposed by the group of Kaupp, works well in this case. 48 This was even though it can be assumed that entropic effects are playing an increasingly important role for the degree of charge localization in molecular wires as they get more flexible with increasing length, thus leading to a greater variety of possible structures with potentially different charge localization properties. In this study, we want to check the transferability of the validated DFT protocol to predicting transport mechanisms in molecular junctions with the primary goal of establishing an easily applicable and efficient, yet predictive approach based on first-principles. If our approach succeeds, it will most likely be due to error compensation, as it has proven very effective in vibrational spectroscopy, where a systematic good agreement between experimental and harmonic wavenumbers calculated with the BP86 exchange-correlation functional was found due to the partial compensation of the harmonic approximation by an inaccurate potential energy surface. 61 In our approach, no conformational sampling is performed, and therefore molecular fluctuations are neglected. Furthermore, it lacks an explicit description of electronic effects arising from the electrodemolecule-interface, which are particularly important in self-assembled monolayers (SAMs) due to the formation of dipole layers. Since molecular conductance experiments are often performed on SAMs where molecular wires are closely packed, it is likely that the degree of charge localization is affected by the local environment, for example by charge-stabilizing effects arising from adjacent wires. In our simulations, environmental effects from a solvent or from adjacent molecular wires are approximated via a PCM. Here, we intend to compensate these effects by the choice of the approximate exchangecorrelation functional and electrostatic embedding via PCM in order to correctly predict transport mechanisms, even though only single molecular structures at 0 K are considered. 2 We aim at the prediction of the tunneling to hopping crossover observed in molecularconductance experiments on conjugated organic wires on the basis of our static DFT approach. 46,48 In Section 2, we describe our procedure of assessing and quantifying the degree of charge delocalization in more detail. In Section 3, we present our results from DFT calculations on charge localization properties for a variety of molecular wires and compare them with respect to their agreement with transition lengths from experiments. A conclusive summary of our results is provided in Section 4, where our findings are discussed in the light of possible applications and limitations of the method. ## Assessment of Charge Delocalization 2.1 Definition of Charge Delocalization To investigate hole transport in the radical cationic molecular wires under study, we evaluated different approaches of determining the degree of charge delocalization in a well-defined manner. As the free electron and the excess positive charge are located on the same position of the radical cationic systems studied here, and free local spins are less basis-set dependent than local charges, 65 the assessment of localization is based on the analysis of the local spin density. In our previous studies on organic mixed-valence systems, the ratio between the local charges on the donor and acceptor moieties connected by a bridge served as a measure for the degree of charge localization, where complete charge delocalization would be expressed by a ratio of one, whereas a ratio of zero would indicate a fully localized charge on one redox centre. 48 Since here, the redox centers are replaced by non-redox-active anchoring groups, we define the degree of charge delocalization r deloc by relating the smallest possible subregion on which a suitably chosen large percentage of spin density is located to that percentage, Fraction of smallest possible subregion on which x % of spin density are located x/100 , where the percentage of spin density x% is chosen such that it represents a majority, e.g. 70%, but not 100% as this would in practice almost always need the full molecule to be included into the subregion. In case of complete delocalization, the spin would be evenly distributed across the entire molecule, while with growing localization, the majority of the spin density would be located on an increasingly smaller fragment of the molecule. Therefore, a molecule would be described as fully delocalized in case of a ratio of one, where for example 70% of the spin density are spread across 70% of the molecule, and so on, leading to r deloc = 1. In contrast, increasing localization results in a decrease of the ratio due to the subregion hosting the predefined percentage of the spin density becoming smaller. Ideally, a semi-localized molecular wire right on the borderline between localization and delocalization, where (in an idealized case) 100% of the spin density populates around one half of the molecule, would be characterized by a ratio of r deloc = 0.5. One could therefore set this value as a cut-off for defining a system as predominantly localized or delocalized, depending on r deloc being smaller or larger than 0.5, respectively. Since we will see in the following that one can hardly define a clear cut-off value for the prediction of the crossover from delocalization to localization that is consistently valid across a broad range of molecules, we rather assigned a molecule to either of the regimes based on the similarities of r deloc for consecutive wires. Accordingly, a relatively strong decrease of r deloc from a shorter to a longer wire would indicate distinctly stronger localization in the latter species, particularly if the values for r deloc at smaller and larger molecular lengths than these two wires would be relatively stable. In principle, one could set any value between 50 and 100 percent as the majority of the spin density, provided that it is not too close to any boundary of this range. We therefore tested several threshold values to evaluate the effect on the calculated delocalization measure and compared the predicted crossover lengths from different thresholds with the experiments. ## Illustrating the Charge Delocalization Measure on the Example of OPTI n Wires In Figure 1, the procedure of assessing the degree of delocalization is illustrated on the example of the shortest member (OPTI 4 ) of a series of conjugated oligophenylene-thiopheneimine (OPTI n ) 3 wires, investigated by Frisbie and coworkers. 15 The local spin density as obtained from natural population analysis (NPA) 66 is plotted per atom along the junction, where the subregions hosting the majority of the spin density are highlighted in different shades of yellow according to the indicated threshold. In Figure 2, the calculated delocalization measure r deloc obtained from Equation (1) for OPTI n wires is plotted as a function of the number of subunits for different fractions of the spin density, where the experimental crossover between tunneling and hopping from OPTI 6 to OPTI 7 is indicated by the black dotted line. We predicted the crossover from the change in the delocalization measure r deloc between subsequent wires, marked by red bars in Figure 2. Therefore, molecular wires were considered in the same transport regime in case r deloc decreased continuosly by the same amount or less, while the crossover was determined from the greatest change of r deloc relative to subsequent wires. The crossover predicted from theory is indicated by a second dotted line, coloured according to the agreement with experiments, see Figure 2. As mentioned earlier, the crossover was expected to occur approximately around r deloc = 0.5, which was used as secondary criteria in case the evaluation based on the change of r deloc was not definite. As can be seen from the plotted delocalization measure based on a threshold of 65% and 70% spin density, the crossover is predicted between OPTI 5 and OPTI 6 in each case and therefore one monomer unit earlier than from experiments. While in the former case, r deloc decreases relatively constantly within a range of 0.04 to 0.08, a more distinct change in r deloc by 0.1 at most is observed in case of a higher threshold of 70%. In contrast, the crossover predicted from a 75% and 80% threshold for the spin density is in good agreement with experiments and the changes in r deloc are qualitatively similar in both cases. Here, a relatively strong change in r deloc of 0.07 marks the crossover between OPTI 6 and OPTI 7 , while r deloc only changes by 0.02 at longer lengths between OPTI 7 and OPTI 8 , suggesting them being in the same transport regime. While for the OPTI n wires better agreement with the experiments was obtained in case of higher thresholds for the spin density, we could not deduce a clear trend of this being generally true for any kind of wires. In three out of five cases, the same results were obtained for all of the tested thresholds, while differences regarding the predicted crossover were observed only in case of OPTI n and OAE n wires (see Supporting Information, Section S3.2). For example, less agreement with the experiments was obtained for OAE n wires when a threshold of 75% spin density was applied. However, best agreement was obtained consistently across all five tested series when a threshold of 80% spin density was applied, which therefore provides the basis for our analysis of the following results. A comparative table of the performance of different methods in predicting the tunneling-hopping crossover from experiments is provided in the Supporting Information (Section S3.1, Table S3), where also a slightly different approach in determining the subregion is presented, where the wires are fragmented into uniformly sized monomer units that are successively included into the subregion until the minimum percentage of spin density is reached (see Section S3.4 and S3.5 for results). In addition to the assessment of charge delocalization based on the calculated delocalization measure, we visualized the corresponding spin density distributions of the molecular wires under study. Ideally, the quantitative measure defined by Equation ( 1) should reflect our intuitive classification based on a visual assessment of these spin densities, which will be checked in the following. ## Predicting Length-dependent Crossover from Static DFT Calculations To validate the capability of the BLYP35+PCM protocol in correctly describing charge transport mechanisms, we applied it to the calculation of charge localization properties of conjugated wires that previously had been investigated in molecular conductance experiments by various groups. First, our computational approach to predicting the length-dependent crossover from DFT is thoroughly discussed on the example of the thiophene-based OPTI n wires, investigated by Frisbie and coworkers. 15 In these cases, our protocol turns out to work perfectly. We briefly compare our findings to related molecular structures, the OPI n and OAE n wires, investigated by Frisbie and by Wandlowski and coworkers, respectively, 16,19 for which the protocol also works well. Second, we present our computational results on molecular wires where the degree of charge delocalization predicted from theory deviates to some extent from the experiments, the ONI n and OPE n wires, investigated again by Frisbie and coworkers and by the group of Wang, respectively. 17,18 Finally, we discuss the results obtained from our DFT calculations in the light of capabilities and possible limitations of our approach in a conclusive summary. ## OPTI n Wires In conductive-probe atomic force microscopy (CP-AFM) experiments by Frisbie and coworkers at room temperature, a length-dependent transition from tunneling to hopping was observed in junctions based on SAMs of OPTI n wires up to 6 nm in length (see Figure 3), at a molecular length of approximately 4-5 nm (OPTI 6 to OPTI 7 ), 15 see Table 1. Our DFT calculations are consistent with the experimental crossover length: Distinctly higher localization is observed for the longer wires OPTI 7 and OPTI 8 when compared to the shorter wires. In Figure 3, the local spin density per atom computed with the BLYP35 functional is depicted for each molecular wire next to the corresponding subregions of highest spin density, as defined in Section 2, marked in yellow. Since the experiments were carried out on SAMs in vacuum, the PCM for thiophene was employed during optimizations to model environmental effects arising from adjacent wires on charge localization. As can be seen from Figure 3, the spin density is rather delocalized for the first three members of the series, therefore suggesting OPTI 4 to OPTI 6 belonging to the tunneling regime. This assumption is confirmed by the calculated delocalization measure according to Equation (1) ranging from 0.64 to 0.51 from OPTI 4 to OPTI 6 . More importantly, a sudden drop of r deloc down to 0.44 indicates a distinctly increasing degree of charge localization for OPTI 7 , where most of the spin density is localized on a subunit comprising roughly three monomer units. The same delocalization length is observed for the longer OPTI 8 , characterized by an even lower r deloc of 0.42, although differently located three-ring registers are suggested to be involved in the transport for the two species. The fact that the length-dependent transition from tunneling to hopping can be predicted based on charge localization properties from static DFT calculations, where environmental effects are entering the model but conformational sampling and dynamics are lacking, is quite remarkable since the number of structures with different spin localization patterns likely rises with growing molecular length due to an increasing number of possible conformations. Hence, although other transport pathways may exist involving not only three-but also one-and tworing registers, as pointed out by Frisbie and coworkers, 15 the static picture considered in this approach is sufficient for reasonably describing the crossover length for these molecular wires and therefore has predictive character. While the inclusion of environmental modelling is found to be highly important for ). Consequently, a more realistic description closer to the experiment, such as the electrode contact and environment, does not necessarily provide a better prediction of the crossover length, as previously pointed out by the groups of Elstner and Kleinekathöfer. 27,28 In particular the accurate first-principles description of molecule-metal interfaces in general is not trivial due to the formation of dipole layers in this region, which is particularly important in case of SAMs. 62-64,67-74 ## OPI n and OAE n Wires Since the BLYP35+PCM protocol worked well to predict the crossover length in case of the OPTI n wires, we applied it to structurally related conjugated wires. A length-dependent crossover from tunneling to hopping was equally identified on SAMs of oligophenyleneimine (OPI n ) 19 wires up to 7 nm long at a molecular length of approximately 4 nm (OPI 5 to OPI 6 ) by the group of Frisbie. 19 Similarly, a crossover to hopping was observed for oligoaryleneethynylene (OAE n ) wires longer than 3 nm (OAE 5 to OAE 6 ) by Wandlowski and coworkers, where experiments were performed on single molecules in solution up to 6 nm long and functionalized on both termini with pyridyl-groups for the attachment to gold leads 16 (see Figure 3 for chemical structures and Table 1 for comparisons of crossover lengths). The spin density distributions of OPI n and OAE n wires are provided in the Supporting Information (Section S1, Figure S1). In Figure 4, the calculated delocalization measure as defined by Equation ( 1 For the OAE n wires, a distinctly higher degree of charge delocalization is observed for the shorter species OAE 4 and OAE 5 when compared to longer ones. Here, increasing charge localization is indicated between OAE 5 and OAE 6 by the highest change in r deloc from 0.49 to 0.40, being the same for the longer OAE 7 . Concluding these results, the BLYP35+PCM protocol is capable of predicting the lengthdependent crossover from tunneling to hopping as observed in conductance experiments on the molecules under study. Moreover, in any case of these wires the crossover is predicted to occur around a value of 0.5 for the delocalization measure. Consequently, the degree of charge localization as deduced from the calculated delocalization measure serves as a valuable tool for evaluating the underlying transport mechanisms in these molecular wires. ## OPE n and ONI n Wires -Borderline Cases To put our DFT approach on more solid ground, we applied it to molecular wires similar in structure to the former species, the oligonaphthalene-fluorene-imine (ONI n ) and oligoparaphenylene-ethynylene (OPE n ) wires (see Figures 5 and 6), investigated by the groups of Frisbie and Wang. Interestingly, these molecular wires provide an example of situations where the calculated localization properties point less clearly to the experimental crossover lengths. Our DFT results for these species are therefore discussed in more detail in the light of possible limitations of the BLYP35+PCM protocol, with the aim of identifying situations where caution needs to be exercised in applying it for the prediction of transport regimes. On SAMs of ONI n wires up to 10 nm long and consisting of alternating fluorene and naphthalene units (see Figure 5), the tunneling-to-hopping crossover was observed at a molecular length of around 4 nm in the experiments (ONI 3 to ONI 4 ). 18,19 Increasing charge localization is indicated between ONI 3 and ONI 4 by a strongly decreasing r deloc from 0.42 to 0.35, matching the experimental crossover (see Figure 6). In contrast, values of r deloc ranging from 0.39 to 0.42 for the first two members of the series indicate charge delocalization to a similar extent and therefore the same transport regime. The majority of the spin density is located on one fluorene subunit in any case for n ≥ 3, which was also found previously in computational studies by the group of Frisbie. 19 Although the change in r deloc is in line with the experimental crossover, an overall rather localized description of the ONI n wires is provided by our calculations, as indicated by r deloc never exceeding a value of 0.5 and as also illustrated by the spin densities in Figure 5. the relevant conformations in the experiment in such a way that the degree of charge localization is affected. For example, the electronic properties of conducting polymers strongly depend on their conformation, since torsion angles between adjacent rings determine the magnitude of the overlap between participating molecular orbitals, as pointed out by André and Brédas. 75 However, unless torsion angles between adjacent ring units do not exceed a value of 40 degrees, the electronic properties are not expected to be substantially different in comparison to the coplanar situation. As can be seen from Figure 5, the molecular structures of the optimized ONI n molecules are considerably twisted. Still, torsional angles around the C-N bond between adjacent ring units of more than 40 degrees are exclusively present in the longer wires from ONI 3 to ONI 6 , whereas for the shorter ONI 2 species they do not exceed 35 degrees (see Supporting Information, Section S2, Table S1). The twisting of the single-molecule structures in our calculations is likely not occurring to the same degree in the experiments, as their flexibility is limited by the presence of adjacent wires in SAMs, therefore possibly leading to a higher degree of planarization. Moreover, πinteractions between the relatively large fluorene and napthalene building blocks of adjacent wires may lead to an overall more flattened structure in the latter scenario, resulting in more efficient charge and spin delocalization. 75 In order to reveal steric effects in a densely packed environment on the molecular structures, the structural optimization or MD simulation of dimers or trimers comprising a small number of wires may provide valuable insight into the mutual impact of molecular wires on their structure and therefore their localization properties. Still, despite a rather localized description possibly due to the lack of more detailed environmental modelling, the experimental crossover is correctly described for the ONI n wires by our DFT approach. The charge transport characteristics of amine-terminated OPE n wires up to 5 nm long were investigated at the single-molecule level using the scanning tunneling microscopy breakjunction (STM-BJ) technique (see Figure 5), 17 where the crossover from tunneling to hopping was observed at a molecular length of around 3 nm (OPE 3 to OPE 4 ), see Table 1. For the longer wires, dimethoxyparaphenylene (DMP) units were incorporated into the molecular backbone to increase the solubility without affecting the conductance properties. For the first three species, OPE 2 to OPE 4 , a relatively high degree of charge delocalization is indicated on the basis of the calculated r deloc as it ranges from 0.90 to 0.88. A distinct increase of charge localization is observed with growing molecular length when going from OPE 4 to OPE 5 , as r deloc decreases from 0.96 to 0.56. The increase of charge localization setting in for OPE 5 in our calculations is not fully in line with the experiments, where the crossover is observed already one monomer unit earlier for OPE 4 . Consequently, a slightly overdelocalized description of the OPE n wires is provided by our DFT calculations when compared to the experiment, which is also indicated by r deloc of the longest member OPE 5 not falling below 0.5. For the latter species, the hopping site comprises roughly three ring units, hosting the majority of the spin density. The rather delocalized description of the OPE n wires compared to the experiments might be attributed to discrepancies between the bonding situation of the molecular wire to the electrode surface modeled in our calculations and in the experiments. For the first three members of the series, a shortening of the lengths of the C-N bond, connecting the molecular backbone to the amine linker unit, is observed from 1.37 to 1.34-1.35 on both termini when compared to the molecular structure optimized in the neutral state (see Section S2, Table S2 in Supporting Information). In contrast, for the longer OPE 5 species, shortening of the C-N bond length to 1.34 is observed only on one side of the molecule where the charge is predominantly localized. The bond length shortening in our calculations might be caused by π-backbonding of the free electron pair of the nitrogen atom to the adjacent carbon atom, leading to stronger planarization and therefore facilitating charge delocalization. However, a different situation may prevail in the actual experiments: 76 One can assume that the free electron pair is less strongly donated to the molecular backbone than suggested from the calculations, but is rather binding to the gold electrode surface by the formation of a donor-acceptor bond. 77 Still, as pointed out by Venkataraman and coworkers, this bond is weakened due to partial delocalization of the lone pair into the molecular π-system when the amine group is connected to an aromatic system, such as in case of OPE n . 77 In our calculations, however, not only partial but full delocalization into the molecule can be assumed due to the lack of the electrode contact. In case of amines, the molecule-electrode bond is relatively weak when compared to previously discussed molecular wires with different anchoring groups, such as thiols and pyridines, therefore leading to weaker coupling to the gold electrode. As pointed out by van der Zant and coworkers, 78 the formation of so-called image charges is particularly apparent in these weakly coupled molecules, where charges residing on the molecular backbone are screened by the electrodes, consequently leading to long-range polarization effects. 79 Accordingly, in a theoretical study of Thygesen and Mavrikakis it was found that the dipole moment at the interface of junctions with a diamine anchoring group is larger than for a dithiol group. 80 As these interface effects, caused by different anchoring groups in the experiments, are not covered in our calculations, the degree of charge localization present in the experiments might be underestimated in our DFT simulations. Interestingly, as discussed in Section 3.1, our DFT results are in good agreement with the experiments in case of the structurally similar OAE n wires, where the anchoring units are represented by pyridyl linkers instead of amines. In both cases, the crossover is indicated around the same length of 3 nm by our DFT cal-culations, irrespective of the anchoring unit, suggesting that effects on charge localization resulting from the anchoring groups are neglected in our single-molecule calculations. ## Conclusions In this study, we used a DFT protocol suggested by Renz and Kaupp based on the BLYP35 hybrid functional with a continuum solvent model to evaluate charge localization properties of molecular wires. Our aim was to predict the length-dependent crossover from tunneling to hopping obtained from molecular conductance experiments. 15, In Table 1, the results of our DFT calculations regarding the crossover from tunneling to hopping are compared to the experiments for the five different series of molecular wires investigated in this study. The theoretical crossover was derived from the change in the delocalization measure r deloc of consecutive wires with growing bridge length, as defined in Section 2. As a primary criterion, we used the largest relative change, i.e. a sudden drop in r deloc for consecutive molecular wires as an indicator for increasing charge localization and therefore the crossover from tunneling to hopping. As a secondary criterion, a value of r deloc = 0.5 was used as guideline for identifying the crossover length, as the crossover likely occurs in this transition zone from delocalization to localization. Depending on the experimental conditions, the optimization of the molecular structures was performed with linker units either attached to one terminus (SAMs) or on both termini (single-molecule experiments). Our computational results suggest that the BLYP35+PCM protocol works well not only in describing the general trend of increasing charge localization with growing bridge length, but also correctly predicts the crossover length in most cases. For all wires except for OPE n , we found a good agreement between theory and experiment regarding the transition length when the protocol is used for calculations on single molecular wires When applying the DFT protocol with PCM to the ONI n and OPE n wires, a slightly overlocalized or overdelocalized description is obtained, respectively. Still, for ONI n , the crossover between tunneling and hopping is predicted correctly based on a sudden drop in the delocalization measure, whereas for OPE n the crossover is delayed by one monomer unit in our calculations compared to the experiment. The approximate DFT protocol used in this study therefore works well in predicting the experimental crossover in most cases, but needs to be used carefully where effects arising from the electrode-molecule interface are expected to have an influence on charge localization, as some aspects are neglected: First, our calculations lack the description of the gold electrode. In case of amine anchoring groups such as in OPE n , π-backbonding of the free electron pair from the nitrogen atom to the molecular backbone rather than binding to the electrode may lead to a more planarized molecular structure in the calculations, suggesting stronger charge delocalization compared to the experiments. Second, amine linkers might induce the formation of larger dipole moments at the molecule-electrode interface when compared to thiols due to weaker coupling to the substrate, as pointed out by Thygesen and Mavrikakis. 80 The formation of image charges might therefore be facilitated, leading to long-range polarization effects that are not covered in the calculations. The neglect of interface effects possibly results in the underestimation of charge localization in the calculations compared to the experiments. For example, for both wires, OPE n and OAE n , which share a similar molecular backbone but are terminated by different anchoring groups, our DFT protocol predicts the crossover at the same molecular length of around It has to be noted that the incorporation of a particular correction to a model might not necessarily lead to a more reasonable description. On the example of OPTI n wires terminated by small gold clusters we showed that the approximate inclusion of both effects, the gold electrode and the environment via PCM, at the same time results in a completely overlocalized description in the calculations and therefore dramatically impairs its performance (see discussion in Section S4.4 in Supporting Information). Therefore, caution needs to be exercised when choosing the parameters that are entering in the calculations, as pointed out by Elstner and Kleinekathöfer. 27,28 To conclude, although the transition length is not predicted perfectly in case of OPE n , the DFT protocol still describes not only the increasing trend of charge localization as a function of length qualitatively well, but also correctly predicts the crossover from tunneling to hopping with an error of at most one monomer unit, possibly through error compensation. Further improvement might result from employing local hybrid functionals. 47 The approach presented in this study is easily applicable to a variety of conjugated organic molecular wires and has predictive character as it is based on first-principles approaches, at least when applied to molecules studied at room temperature. Since the transport mechanism is not only dependent on the length but also on the temperature, 1 our approach may be limited to experiments performed in this temperature range. Our protocol may prove useful not only to gain insight into the charge transport characteristics of a particular molecular system to identify the underlying transport mechanism, but also to reveal the nature and extent of the hopping sites that are involved in the charge transport. The latter aspect is of vital importance, especially in the context of charge transport through biomolecules, e.g. proteins and DNA, which not only strongly depends on the molecular length but also on the architecture of the molecular backbone and thus on the number and sequence of amino acids and base pairs, respectively. 13,81 An interesting objective of future studies is therefore to verify the validity of the present DFT protocol in predicting transport mechanisms not only as a function of molecular length, but also of the molecular structure, e.g. the base sequence in DNA. 13,81 Beyond that, it appears worthwile to evaluate semi-empirical methods as a more efficient and therefore promising alternative to the present DFT protocol, as they were previously applied to the investigation of charge transfer in organic mixed-valence systems. This could be valuable for screening of larger data sets, and would also help to identify efficient electronic structure protocols for use in comprehensive multiscale approaches. 27,28 For a more detailed analysis of potential error compensation in our protocol, it may be helpful to explicity compare coherent and hopping charge transport rates and to identify the length-dependent crossover point for a given electronic structure description, as suggested in Ref. 86 It would also be worthwhile checking whether machine learning methods, as recently successfully applied for predicting transport mechanisms in DNA, 87 can be transferred to the types of wires under study here. Altogether, such approaches may contribute to a more predictive theoretical framework for molecular and nanoscale electronics. performed for both the isolated molecules and molecules in solution, since the importance of environmental effects on charge localization properties was pointed out by Kaupp and coworkers and was shown in our previous study on organic mixed-valence systems. 46,48 For the inclusion of solvent effects, the polarizable continuum model with the integral equation formalism model (IEFPCM) 95,96 was employed as implemented in the Gaussian 09 program package by using the SCRF keyword with the available dielectric constants for thiophene ( = 2.7270), benzene ( = 2.2706) and tetrahydrofuran ( = 7.4257). Natural population analyses 97 were performed with the Gaussian 09 program package to gain information about the distribution of local spin densities and charges, which were summed over specific subregions of the molecule. For the local spin density, an absolute value of 1 refers to one unpaired electron (i.e., a more precise yet more cumbersome expression would be "local unpaired electron density"). Molecular structures and spin densities were visualized with the Avogadro editor, 98 applying an isosurface value of 0.001 for plotting the spin density distributions.

chemsum

{"title": "Towards a First-Principles Evaluation of Transport Mechanisms in Molecular Wires", "journal": "ChemRxiv"}

modeling_of_hydrogen_atom_diffusion_and_response_behavior_of_hydrogen_sensors_in_pd–y_alloy_nanofilm

## Abstract: To detect hydrogen gas leakage rapidly, many types of hydrogen sensors containing palladium alloy film have been proposed and fabricated to date. However, the mechanisms and factors that determine the response rate of such hydrogen sensor have not been established theoretically. The manners in which response time is forecasted and sensitive film is designed are key issues in developing hydrogen sensors with nanometer film. In this paper, a unilateral diffusion model of hydrogen atoms in Pd alloy based on Fick's second law is proposed to describe the Pd-H reaction process. Model simulation shows that the hydrogen sensor response time with Pd alloy film is dominated by two factors (film thickness and hydrogen diffusion coefficient). Finally, a series of response rate experiments with varying thicknesses of Pd-Y (yttrium) alloy film are implemented to verify model validity. Our proposed model can help researchers in the precise optimization of film thickness to realize a simultaneously speedy and sensitive hydrogen sensor. This study also aids in evaluating the influence of manufacturing errors on performances and comparing the performances of sensors with different thicknesses. Hydrogen gas is a widely used raw material and product (or byproduct) of modern industries 1 . However, the gas is flammable and easily explodes in the air under normal conditions at concentrations ranging between 4.65% and 74.5% 2 . Hence, almost all production activities involving hydrogen gas entail extremely safe monitoring systems to prevent any leakage. In general, the main research interest in hydrogen leakage detection is how to reduce the time between the hydrogen leak and the corresponding alarm and to increase the chance of escaping the dangerous site or shutting down the device . Optical fiber hydrogen sensors play a highly important role in monitoring hydrogen leakage because of their anti-electromagnetic interference and intrinsic safety ability. Most optical fiber hydrogen sensors employ palladium as the transducer element because of the intrinsically high sensitivity and selectivity toward hydrogen . To improve the sensor response rate, numerous experimental prototypes for hydrogen gas detection have been developed . Zhao et al. found that response time is strongly dependent on the α , mixed α /β, and β Pd-hydride phases formed in the films. The longest response time (about thousands of seconds) occurred at the hydrogen concentration corresponding to the α → β phase transition region. The phase transition region can be adjusted by changing the content of the alloy element 13 . Song et al. proposed a kind of annealing-stimulated method to retard the aging behavior of thin film and to improve the response speed 14 . Kay et al. studied the kinetics of hydrogen absorption by bulk Pd (110); they further demonstrated that in the α -phase region, where concentration is low, the hydrogen absorption rate of Pd is limited by the diffusion process of hydrogen atoms in the bulk rather than the chemisorption of H 2 molecules on the Pd surface. The research results well demonstrated the dominant factors of penetration time in bulk Pd. However, such results are difficult to be used in forecasting the response time of Pd nanofilm, particularly, the single surface of the film exposed to hydrogen 15 . These experiments indicate that the dimension and content of Pd nanometer film play the most important roles in rapid hydrogen detection. However, few theoretical analyses have been performed on the diffusion behavior and response time of Pd or Pd alloy thin films exposed to hydrogen. For a different application, precise designing the Pd or Pd alloy thin film to satisfy the measuring requirements is also a crucial issue. In this paper, a hydrogen single-side diffusion model of the Pd-based metal thin film was established, and a function was derived to describe the influences of the dimension and content of the thin film on the response time. Specifically, we have performed experiments on the optical reflectance response characteristics of Pd-Y films to validate the diffusion model. We theoretically and experimentally confirmed that the response time holds a direct ratio to the square of the film thickness. This conclusion benefits the precise designing of nanofilm thickness to satisfy the different requirements for response rate. ## Hydrogen Diffusion Model in Pd-based Metal Thin Films Hydrogen diffusion model based on Fick's second law. When hydrogen molecules meet Pd, the former dissociate into hydrogen atoms on the Pd film surface. The hydrogen atoms then dissolve and diffuse in Pd and form an interstitial solid solution PdH x (x is the atomic ratio of H/Pd). When the hydrogen concentration decreases, the hydrogen atom desorbs from the Pd film surface. This reaction is reversible. Given that the hydrogen concentration in the film evolves, the diffusion of hydrogen atoms is a typical non-steady state diffusion process that can be described by Fick's second law 16 . The Pd-based thin film in the optical fiber hydrogen sensor is usually coated at the end of the optical fiber (or on one side of a substrate) and exposed to the hydrogen gas. The Pd-H 2 interaction model is shown in Fig. 1(a). This model consists of hydrogen dissociating on the outside surface, diffusing in the inner Pd film, and restraining on the Pd-substrate interface. Generally speaking, the substrate that supports the Pd-based thin film cannot be penetrated by hydrogen. Therefore, it is a diffusion-limited reaction in one dimension with a constant source in solid. A coordinate system O is established at the center of the interface between the film and the hydrogen gas [Fig. 1(b)]. The x-axis is perpendicular to the film surface. We denote the thickness of the thin film as L, the diffusion coefficient of the hydrogen atom in the Pd-based metal nanofilm as D, and the hydrogen atom concentration as C(x, t), which is a function of the time t and the position x in the Pd-based metal nanofilm. When the film comes in contact with the hydrogen gas (H 2 partial pressure is P H2 ), the hydrogen molecule is dissociated continuously into hydrogen atoms on the solid-gas interface, and a constant concentration of hydrogen atoms C s is formed on the interface. According to Fick's second law, the hydrogen atom concentration C(x, t) satisfies a second-order partial differential equation (PDE) as follows: The hydrogen concentration in the film is zero before the Pd-H 2 reaction; hence, the initial conditions of the equation are given as follows: Furthermore, the hydrogen concentration on the solid-gas interface is a constant C s determined by the hydrogen concentration in the surrounding environment. Thus, the first boundary condition is given as follows: s The hydrogen atoms diffuse in the opposite direction when they reach the film-substrate interface. This process is similar to the reflection of a flat mirror. The concentration gradient of the incident hydrogen atoms is equal to that of the reflected hydrogen. Therefore, the second boundary condition is given as follows: x L The PDE (1) is solved by using separation of variables method (also known as the Fourier method). The hydrogen atom concentration C(x, t) in the Pd-based metal film is obtained as follows: Equation (5) shows the concentration distribution of hydrogen atoms at any time. When we design an optical fiber hydrogen sensor, whether the film is coated on the end or on the side of optical fiber, we detect the average effect of change on refraction index in the film. Therefore, the average concentration of hydrogen atoms in the film along the x-axis direction is given as follows: The average concentration of hydrogen atoms in the film C t ( ) only depends on the time and film thickness and not on the film position. If we define the percentage of response as η to represent the completion level of reaction, the expression is given by: Model discussion and simplification. (A) Transient concentration analysis and simplification. Equation ( 5) shows the transient concentration distribution of hydrogen atoms in the film at any time. The right-hand side of equation ( 5) is expanded as follows: When time t is equal to zero in equation ( 8), the exponent term Equation ( 8) is simplified as follows: m s 1 The infinite series in equation ( 9) is summed, and the details of derivation are included in Appendix I. Equation ( 9) can be rewritten as follows: s Equation (10) indicates that the hydrogen concentration is C s at the gas-solid interface and 0 in the Pd-based metal film when time t = 0. This representation means that the hydrogen atom diffusion in the film has not yet begun. For t > 0, the exponent term attenuates rapidly in equation ( 8), particularly, in the high-order harmonic wave. If we choose the first two terms of the infinite series as the approximation of hydrogen concentrate distribution, the result is given by Moreover, the error of truncation is given by A series of time, t = 0, 2, 4, 6, 8, 10, 20, 40 s, is proposed to study the concentration distribution of transient hydrogen in Pd-Y alloy film. We then suppose the thickness L of Pd-Y alloy film as 30 nm, diffusion coefficient D as 10 nm 2 /s, and the hydrogen concentration on the solid-gas interface C s as 10 −5 mol/mm 3 . The concentration distribution of hydrogen based on equations ( 10) and ( 11) is shown in Fig. 2. The hydrogen concentration in the film on the solid-gas interface is higher than that in the film. As time progresses, the hydrogen concentration increases. For t = 2, the curve is under y-axis, which is caused by the error of truncation. (B) Average concentration analysis and simplification. With the same consideration, the first two terms of the infinite series in equation ( 6) are reserved as the approximation of the average hydrogen atom concentration in the film. The result is given by Meanwhile, the error of truncation is as follows: (C) Percentage of response simplification and analysis. On the basis of equation ( 13), the percentage of response η is rewritten correspondingly as Varying thicknesses, L = 10, 20, 30, 40, 50 nm, is proposed to study the influence of thickness on the response. We suppose the diffusion coefficient D as 10 nm 2 /s. The percentage of response based on equation ( 15) is displayed in Fig. 3. The figure illustrates that the percentage of response rises exponentially and that the thinner film can complete the reaction earlier. (D) Response rate analysis. To determine the change rate in the percentage of response, we denote the response rate V r as Equation ( 16) indicates that the sensor response rate is not a constant. The response rate is the highest at the initial status of the reaction and attenuates rapidly afterward. When time t approaches infinity, the response rate approaches 0, and the reaction proceeds toward the equilibrium status. (E) Response time analysis. The percentage of response η is an exponent function of time t. Only when the time t approaches infinity can the percentage of response reach 100% and the hydrogen sensor reading can be stable. In practice measurement, we often regard the 90% of reaction as the stable value of reading. The response time T response is defined as the time spent from reaction initiation to 90% steady state. Therefore, the response time must satisfy equation (17). The typical value of the response time T response changes within several seconds to tens of minutes. Thus, the second-order term in equation ( 17) is much smaller than the first-order term. Equation ( 17) can be approximated with the first-order term as follows: ## Experiments Experimental setup and principle. In this work, a reflective optical fiber bundle sensor structure with compensation is used to record the reaction process of several Pd alloy films with different thicknesses when they load hydrogen. Figure 4 presents the diagram of the gas sensor system. Broadband light generated by a highpower LED is coupled into an optical fiber bundle. The light is split in two beams, namely, testing and reference signals, by the fiber bundle and then propagated to two chambers separately. The reflected signals from the two chambers are collected by their respective silicon photodiode. The signal originating from the reference chamber is only affected by the light source fluctuation. The final output S is only determined by the refractive index on the Pd alloy film surface after the errors are compensated by the reference signal. Hydrogen exposure reduces both the real and imaginary parts of the Pd complex refractive index, which then increases output S 17 . S is a function of the average hydrogen concentration in the film. If we denote the initial value of S before the film is exposed to the hydrogen gas as S 0 and the stable status value of S as S ∞ , the percentage of optical response of thin film η e can be defined as The percentage of optical response of thin film η e increases gradually from 0 to 1 during the Pd-H reaction. If we record time T e spent during the ascent processing of parameter η e , the time T e corresponds to the response time of the film T response in equation ( 19). Film preparation and characterization. At one time, sputtering all films guarantees the uniformity of material composition and reduces the influence of the non-uniform diffusion coefficients on response time. Several similar substrates were placed in the sputtering chamber. These substrates were placed at the center of the sample platform and with different layer heights in the BESTECH sputtering system. Three-inch Pd and Y targets were installed in the DC and RF sources of the sputtering system, respectively. Under a sputtering pressure of Ar at 0.5 Pa, the deposition power was controlled to 100 W for Pd and 150 W for Y, which corresponded to the deposition rates of 1.3 and 0.2 /s, respectively, for the center substrate. The sputtering time was approximately 200 s. The sputtering parameters of the film deposited on the central fused quartz substrate (ID: #1) was monitored by a quartz crystal thickness monitor (Fig. 5). Ellipsometry is a noncontact optical measurement technique that can obtain the thickness and dielectric properties of thin films by analyzing the change in polarization state between the incident and the reflect light on the thin film. An ellipsometry apparatus (M-2000 type produced by J. A. Woollam) was used to measure the thicknesses of the films prepared by the sputtering system. The thicknesses were 11.96 (ID: #4), 15.04 (ID: #3), 21.01 (ID: #5), and 29.84 nm (ID: #2). The #1 film, which was a calibration sample, was not used in the follow-up experiments. Meanwhile, the refractive index of the metal material is a complex number  n, which consists of the real (index n) and imaginary parts (extinction coefficient k). The test results indicate the extremely close complex refracted indices of the five film pieces and the high consistency of the material composition of these films. The refraction index of the Pd 0.92 − Y 0.08 thin film is presented in Fig. 6. ## Results and Discussions The four pieces of thin films are exposed to a mixture of 4% H 2 in N 2 , respectively, to observe the response processing. Figure 7 shows the percentage of optical response of the four thin films. The curves shown in Fig. 7 are identical to those analyzed in Fig. 3, which all exhibit exponential growth characteristics. No overshoot is produced in the whole growth. When the time trends to infinity, the optical response of the thin film trends to 1. The thick film needs a much longer response time than that required by the thin film. The four pieces of thin films are exposed to a mixture of 4%, 2%, and 0.5% H 2 in N 2 , respectively. The response time is shown in Table 1. To clearly show the relationship between the response time T and the thickness of thin films L, a group of curves are displayed in Fig. 8. The horizontal coordinate represents the square of the film thickness L 2 . The vertical coordinate represents the response time T. At different hydrogen gas concentrations, the response time is in direct ratio to the square of film thickness. This aspect is identical to the theoretical analysis results (equation 19). The slope of the line is determined by the diffusion coefficient of the hydrogen atoms in Pd. Different reaction productions (α -PdH and β -PdH) are known at different hydrogen concentrations, leading to different paths of diffusion. In the experiment, three lines are very close to being parallel. This similarity demonstrates the similar diffusion coefficients for different diffusion patterns in the Pd-Y alloy film. We also found that the response time at 2% H 2 is much longer than those at higher or lower concentrations of hydrogen gas (Fig. 8), which may be caused by the phase transition of PdH. In the phase transition concentration, the Pd material undergoes phase transition from α -PdH to β -PdH, which consumes a specific time. Hence, the response time at 2% H 2 of each thin film is a specific time longer than those at other concentrations. The intercept on the y-axis of the line represents the specific time of phase transition. The phase transition time is about 50 s (Fig. 8). The slow response phenomena of sensors with Pd-Y alloy films induced by phase transition are also observed in Pd-Au alloy 13 . ## Conclusion The mechanism and factors that determine the response rate of hydrogen sensors are key issues in the design and development of hydrogen sensors for the rapid leak detection. The response process of the sensor to the hydrogen gas is a chemical reaction, which determines the response rate. To describe this process theoretically, a unilateral diffusion model of hydrogen atoms in Pd alloy based on Fick's second law is proposed. By resolving the diffusion model, we found that the sensor response is an exponential function as time changes. The response time of hydrogen sensor with the Pd alloy film is dominated by two factors (film thickness and hydrogen diffusion coefficient in Pd). The response time of the hydrogen sensor depends on the square of alloy film thickness. Experimental results not only validate this point, they also show the presence of close diffusion coefficients in the α -PdH and β -PdH. Our proposed model can be used to design the film thickness in hydrogen sensors for various rate requirements. The experiments also reveal the phase transition process and corresponding duration excluded in the diffusion model. Among the important issues for future investigation include which factors determine the phase transition time and how to reduce such duration.

chemsum

{"title": "Modeling of hydrogen atom diffusion and response behavior of hydrogen sensors in Pd\u2013Y alloy nanofilm", "journal": "Scientific Reports - Nature"}

microgel_paint_–_nanoscopic_polarity_imaging_of_adaptive_microgels_without_covalent_labelling

## Abstract: Polymer nanostructures have enormous potential for various applications in materials and life sciences. In order to exploit and understand their full capabilities, a detailed analysis of their structures and the environmental conditions in them is essential on the nanoscopic scale. With a super-resolution fluorescence microscopy technique known as PAINT (Points Accumulation for Imaging in Nanoscale Topography), we imaged colloidal hydrogel networks, so-called microgels, having a hydrodynamic radius smaller than the diffraction limit, gaining unprecedented insight into their full 3D structure which is not accessible in this much detail with any other experimental method. In addition to imaging of the microgel structure, the use of Nile Red as the solvatochromic fluorophore allowed us to resolve the polarity conditions within the investigated microgels, thus providing nanoscopic information on the x,y,z-position of labels including their polarity without the need of covalent labelling. With this imaging approach, we give a detailed insight into adapting structural and polarity properties of temperatureresponsive microgels when changing the temperature beyond the volume phase transition. ## Introduction Microgels have received enormous attention in the last two decades and currently count among the most studied colloidal and polymer systems. 1,2 Especially microgels that can change their shape or properties in response to external stimuli such as changes in temperature 3,4 and pH, 5,6 electrochemical triggers 7,8 or light 9 and, for multi-responsive microgels, 10,11 combinations thereof 12 bear huge potential for applications. 10, For a deeper understanding of this switching behaviour and its consequences for applications, a thorough in situ investigation of concomitant changes in size, shape and polarity is essential. Cryogenic transmission electron microscopy (cryoTEM), 16 in situ TEM 17 and atomic force microscopy (AFM) 18 have proven to be suitable imaging techniques for soft matter. The low contrast without heavy atom staining, however, prevents exploitation of the full resolving power of electron microscopy, and AFM, despite its ability to probe rheological properties, is limited to information obtained by probing the microgel mechanically. The development of super-resolved fluorescence microscopy methods made also fluorescence microscopy suitable for visualizing microgels with diameters in the size range of hundreds of nanometers. 23 Super-resolved fluorescence imaging of microgels has been restricted so far to a method called direct stochastic optical reconstruction microscopy (dSTORM) 24 utilizing the chemical on-and off-switching of fluorophores induced by redox reactions with additives. With this technique, size changes of microgels due to co-nonsolvency were visualized 25 and the selective labelling of different functionalizations in the core and shell, respectively, could be investigated. 17 Furthermore, Conley et al. recently reported the compression and deformation of single microgels with increasing microgel concentration. 26 Additives such as the switching buffer can be avoided when the structures are covalently labelled with photoswitches. 27 However, in many cases covalent labelling requires signifcant efforts. Strategies to circumvent covalent labelling are of signifcant practical advantage since it allows super-resolution imaging to access a plenitude of polymer systems. Bergmann et al. found that non-covalent labelling of microgels with rhodamine 6G and subsequent imaging under typical dSTORM buffer conditions resulted in superresolved microgel images of high quality. 28 An alternative to dSTORM, also without the need of covalent labelling, is superresolved Points Accumulation for Imaging in Nanoscale Topography (PAINT). 29,30 Additionally, in contrast to dSTORM, the PAINT approach has the advantage that no switching buffer has to be added which might disturb the (polymer) system under investigation signifcantly. It has, for example, been used to study the structure of Large Unilamellar Vesicles (LUVs) where fluorescent probes continuously target the object of interest in a flux 29 and been combined with other super-resolution methods. 31 Additionally the technique is used for super-resolved imaging of DNA (DNA-PAINT). 32 Furthermore, as an additional readout parameter, spectral information can be accessed. Bongiovanni et al., for example, investigated the hydrophobicity of biological structures exploiting the solvatochromic behaviour 36 of Nile Red in addition to the super-resolution imaging via PAINT. 37 Local changes in the membrane polarity in live mammalian cells were investigated by employing Nile Red based SR-STORM and SR-PAINT by Moon et al. 38 In general, super-resolution transient binding (STB) methods are emerging imaging techniques with high resolution, which avoid the limitation of photobleaching. 39 Herein, we report on the PAINT approach to study the 3D structure and the point-wise polarity of thermo-sensitive coreshell microgel systems with super-resolution. In the model system investigated here, the core consists of poly(N-isopropylacrylamide) (PNIPAM) and the shell of poly(N-isopropylmethacrylamide) (PNIPMAM) with volume phase transition temperatures (VPTTs) of 32 C and 42 C, respectively. The polarity information is obtained by the use of the solvatochromic dye Nile Red, 40 thus demonstrating an approach to gain polarity information along with the 3D structure. ## Sample preparation The synthesis of these core-shell microgels with the PNIPAM core and PNIPMAM shell has been previously published. 17 10 mL of 0.5 mg mL 1 of core-shell microgel solution was spin-coated onto a freshly plasma cleaned coverslip and placed inside a temperature cell (see the ESI ‡ for more details). 100 mL of bidistilled water was added to the sample in vitro. In order to label the microgels non-covalently via the PAINT method, 2 mL of 10 11 M Nile Red solution in methanol was added. ## Experimental setup For 3D PAINT measurements, the sample was excited by focusing a 488 nm laser beam onto the back focal plane of a 100 1.3NA oil immersion objective lens (UPLFLN 100XO2, Olympus) in an inverted microscope (IX83, Olympus) using a plano convex lens of focal length 500 mm to obtain Köhler illumination. A single line dichroic mirror (zt 488 RDC, AHF analysentechnik, Tübingen) reflects the excitation beam and allows the emission from the sample to subsequently pass through an emission flter 510LP (AHF analysentechnik, Tübingen). The detection path consists of an imaging system made up of two plano convex lenses of focal lengths 100 and 200 mm for a two-fold magnifcation. In this study, we also make use of the solvatochromic behaviour of Nile Red in order to resolve the polarity of microgels at the nanoscale. In contrast to the studies by Ke Xu and coworkers, 35,38 we did not refract the emission light through a prism, because we wanted to avoid additional spreading of the point spread functions that were already distorted by astigmatic imaging (see below). Instead, the emission was split into two channels using an Optosplit 2 bypass (CAIRN Research, UK). A dichroic flter (zt 594RDC, AHF analysentechnik, Tübingen) in the Optosplit reflects shorter wavelengths (<594 nm) and transmits longer wavelengths (>594 nm). Additionally, a 617/73 Brightline HC Bandpass emission flter (AHF analysentechnik, Tübingen) and a 514LP Razor edge emission flter (AHF analysentechnik) were introduced in the transmitted and reflected path of the Optosplit, respectively, to reduce the background. The different z-offsets in both the reflected and the transmitted channels were minimized with a corrector lens in each channel. For 3D imaging, a cylindrical lens of focal length 500 mm was introduced approximately 5 cm in front of the EMCCD chip of an Andor Ixon Ultra 897 camera. For all the PAINT measurements, a laser power density of 5.3 kW cm 2 , an exposure time of 5 ms and electron multiplying mode setting with a gain of 200 were used. The number of frames recorded per area of the sample varied between 60 000 and 120 000. The images were recorded using a home-made temperature cell at six different temperatures: 21, 33, 35, 38, 43 and 53 C. The temperature range was selected such that the VPTT of both, the PNIPAM core and PNIPMAM shell, could be covered. The recorded movies were separately cropped for the shorter and the longer wavelength channels of the Optosplit and analysed with ThunderSTORM. 41 The cropped movies and the super-resolved localization fles were subsequently fed as inputs to custom-made MATLAB Optosplit routines (see the ESI ‡ for more details). The output obtained from the Optosplit routines contains the x, y, z positions of labels and the ratio of the emission intensity between the longer and the shorter wavelength channels. Details on this intensity ratio and its calibration to compare it to the conditions in different solvents can be found in the ESI. ‡ The localization fles are subsequently fed into VISP software 42 where the localizations of individual microgels are cropped and the 3D distribution of localizations is plotted using a custom-made MATLAB routine (see the ESI ‡). ## Results and discussion The average localization densities of 20 microgels at six different temperatures are shown in Fig. 1. For better clarity, we present the axially symmetrical data in a graph where the averaged densities are plotted versus a relative z-position and the distance from the symmetry axis (see Fig. 1a). At room temperature, the density of localizations is rather constant throughout the microgel and the sparse pixels from the outer regions represent localizations close to dangling chains of the swollen microgels. Additionally, some localizations in this region can be due to localization inaccuracies. At 33 C and 35 C, respectively, it can be observed that the PNIPAM core starts to collapse since the density of localizations in the centre starts to increase gradually with the increase in temperature. Also the overall size of the microgels decreases with respect to room temperature. At higher temperatures, which reach beyond the VPTT of the shell, the entire microgel becomes more compact. Still, the labelling density in the core remains significantly higher than that in the shell. The 3D distribution of Nile Red labels of a single microgel along with xy, yz and xz projections for the lowest (21 C) and the highest temperatures (53 C), respectively, is shown in Fig. 2. The sizes of the microgels obtained via PAINT are in agreement with those of the DLS measurements at the respective temperature denoted by the black circle in the xy projection (see the ESI of Gelissen et al. 17 ). In the z-direction, a slight elongation can be observed which originates from the lower localization accuracy of approx. 60 nm in the z-direction in contrast to approx. 20 nm in the x-and y-directions. Since the coverslip surface cannot be detected with our method, we set the z-axis to zero at the centre of the microgel. Conclusively, the position of the coverslip is always at negative z-values. In addition to the localizations, the polarity of each position is represented using its colour. The colour of the points indicates the intensity ratio between the long wavelength (>594 nm) and the short wavelength (<594 nm) channel which, in Fig. 2, is scaled from 0.4 to 0.8. Throughout our paper, this intensity ratio representing the polarity of the corresponding positions will be called their solvatochromic value. As shown in Fig. 2 and Fig. S2 of the ESI, ‡ the environment in the microgels is in general more polar in the swollen state below the VPTT. In the collapsed state, the fluorescence emission of almost all Nile Red labels is hypsochromically shifted as expressed using the lower solvatochromic value. The averaged super-resolved solvatochromic image at different temperatures is shown in Fig. 3. Compared to Fig. 1, in which the localization density is shown, Fig. 3 to the inner part of the microgel account to values ranging from 0.4 to 0.6 (blue to green). In the outer part of the microgel, values of around 0.8 are observed (red), due to the more polar surroundings or even to the so-called dangling polymer chains. At 33 C (Fig. 3e), a temperature just slightly above the VPTT of the core, the size of the point cloud decreases and the solvatochromic values close to the centre are shifted towards lower values (approximately between 0.4 and 0.55), indicating that the core is more apolar after the collapse. Interestingly, the collapse of the core induces increased solvatochromic values in the shell. It seems that the core collapse induced structural changes in the surroundings of Nile Red labels adsorbed to the shell, a fact that is also highly important for microgels as encapsulating agents, for example, for drug delivery and for their application in catalysis. 43,44 Especially close to the coverslip, we observe a signifcantly higher polarity. This increased polarity is reasonable since the coverslips are slightly negatively charged due to our surface cleaning procedure (see the ESI ‡). The contrast between the core and shell is consistently observed for the measured temperatures between the VPTT of the core and that of the shell. Also, the size of the point clouds, and thus of the microgels, decreases gradually with increasing temperature (see Fig. 3d-i). Increasing the temperature above the VPTT of the shell at ca. 42 C also causes collapse of the PNIPMAM polymer chain shell. Consequently, at this point the shell becomes rather hydrophobic and the solvatochromic values within the shell decrease to values below 0.55, similar to the values in the core. Only some of the solvatochromic values in the most outer regions of the microgels remain rather high. At 53 C, the highest measured temperature, we observe the smallest size and the highest overall hydrophobicity of the microgels. In general, it should be noted that the Nile Red molecules observed in the microgels presented here are in surprisingly apolar surroundings as is obvious from their solvatochromic values. The local environment of the majority of Nile Red labels can be compared to that ranging between n-decane and ethylacetate (for calibration see the ESI ‡). The reason for this could White pixels mean that no localizations were detected in the corresponding toroid. For pixels close to the symmetry axis, this is sometimes observed in particular in the pole regions. It should be noted that the sampling density of the localizations at the periphery is significantly lower than that in the center (see also Fig. 4). be that the Nile Red molecules cause local changes of the polymer conformation around them. These changes, however, do not affect the microgel properties due to the sub-nanomolar concentration of the dye. For further discussion, we reduced the full 3D information to radial distributions assuming the spherical symmetry of the microgels which is a reasonable assumption as shown in the images in the fgures above. Based on this simplifcation, in Fig. 4 we plotted the median solvatochromic values and the localization densities versus the radial distance from the centre of microgels at different temperatures for 20 to 30 microgels. Fig. 4 emphasizes the trends discussed above that the median solvatochromic values are rather constant at 21 C when the polymers, both in the core and in the shell, are in the swollen state. Collapse of the core results in lower polarity in this region, which surprisingly causes a shift to higher solvatochromic values and thus an apparent higher polarity in the shell. We attribute this to a change in polymer conformation in the shell. 45 At 53 C, where the core and shell are collapsed, the lowest solvatochromic values are observed throughout the microgel. Only in the most outer parts of the shell, the values increase even beyond the ones at lower temperatures. Presumably, the Nile Red labels are in very low polymer density regions mainly surrounded by water. It is also interesting to compare these results with the solvatochromic values of a microgel which consists only of the PNIPAM core before polymerizing the shell onto it (for detailed analyses see also the ESI ‡). As shown in Fig. 5, the solvatochromic values inside the microgels at r < 100 nm are signifcantly lower in the core-shell microgel and equalize at larger distances from the centre. After increasing the temperature to 33 C, all solvatochromic values decrease. However, at this temperature, the polarity inside the microgel is lower throughout the core-only microgel. At 53 C, where all polymers are collapsed, the solvatochromic values of the core-shell microgel are signifcantly lower. At this temperature, both the core-shell and the core-only microgels show a similar increase in polarity in vicinity of their hydrodynamic radii. The comparison of the temperature-dependent behaviour of the core-shell and core-only microgels shows that the shell has a signifcant influence on the core when considering polarity. At 21 C, the shell makes the core more apolar than the core alone. Since the shell was synthesized onto the core in its collapsed state at elevated temperature, the conformation at room temperature presents additional stress, especially in the core which is more apolar than the same core without the shell. When the core collapses at ca. 32 C, it cannot reach the fully collapsed state due to the conformational restrictions opposed by the shell. At 53 C, when both the core and shell are collapsed, the polarity in the core-shell microgel is below the values of the core-only microgel. In both cases, however, the polarity increases at the periphery of the microgels. In order to correlate the solvatochromic values with polymer densities, we compare the localizations and their solvatochromic values from PAINT with the radial distribution of polymers within the microgels obtained by ftting the static light scattering (SLS) intensities at different temperatures, Fig. 6. It is worth noting that while SLS probes the polymer density within the microgels, PAINT probes the distribution of localizations of the dye diffusing within the microgels. These two quantities are not the same but offer complementary information on the structure of the microgels. The SLS intensities in Fig. 6a were ftted with a fuzzy-core-shell-model (see the ESI ‡ for more details on the ftting parameters). 46 This is the simplest model that allows obtention of good fts and values of the microgel radius consistent with the hydrodynamic radius obtained by DLS. Due to the high quality of the fts and the consistency of the obtained radial distribution of polymers shown with black lines in Fig. 6b-d with DLS data, more complex models with a higher number of parameters are not needed. 28,47 The radial distribution of polymers within the microgel is compared to (i) the respective radial distributions of localizations obtained by PAINT measurements averaged over 20 microgels (black dotted line) as also outlined in the study by Siemes et al. 48 and (ii) to the median solvatochromic values (red dotted line). At 21 C, the localization density distribution is similar to the polymer distribution within the core but gradually decreases towards the outer microgel regions whereas the polymer density in the shell shows only a small drop. The median solvatochromic values remain rather constant. It seems that neither the polymer density nor the polarity as probed by Nile Red changes signifcantly. Still Nile Red fluorescence appears more often in the inner microgel region. The reason for this is probably the higher cross-linker density in the core which increases the probability of Nile Red to become immobile. As shown in Fig. 6c, at 38 C, SLS data show that the core is collapsed and hence has higher polymer density when compared to the still swollen shell in the radial distance range of 125 nm to 170 nm. The radial distribution of localizations is the highest close to the centre and then decreases rapidly. Only a few localizations appear in the low-density shell region. These localizations have signifcantly higher solvatochromic values. As already discussed above with respect to Fig. 3, the collapse of the core seems to change the network structure in the shell in a way that it becomes more polar, thus reducing the probability of Nile Red adsorption and increasing the local solvatochromic values. Fig. 6d presents the data at 53 C when the core and shell are collapsed. Surprisingly, even here, the localizations drop rapidly from the core towards the periphery and the solvatochromic values in the now collapsed shell remain increased. Our fndings point to a situation in which the shell collapses, as indicated by SLS, but despite the collapse, heterogeneities in the polymer network structure exist which result in higher probability of Nile Red molecule adsorption in the centre. Additionally, the polarity inside the collapsed shell is higher than that in the core. Presumably, the shell cannot fnd its optimal collapse conformation due to the restrictions of the previously collapsed core. It is worth noting that combining the complementary information obtained from SLS (polymer radial density profle) with PAINT (radial density of localizations) provides more insight into the structural changes within the microgels. The fact that scattering and PAINT probe different quantities allows for a more complete characterization of the microgels. All of our fndings also show that small molecules can diffuse through hydrogel networks in the swollen and even in the collapsed state. This is demonstrated with Nile Red which is localized in the microgel core even at 53 C where the core and shell are both collapsed. This result has to be taken into account, for example, for the design of microgels as drug delivery systems and for catalysis. 43,44 ## Conclusions We analysed the nanoscopic 3D structure and local polarity conditions in thermo-responsive core-shell microgels with a localization-based super-resolution fluorescence microscopy approach which does not require covalent labelling. The polarity was accessed using the solvatochromic dye Nile Red. The temperature-dependent change of the structure of the microgels and the polarity within them were obtained. The size changes of the microgels are in good agreement with static and dynamic light scattering data. The polarity decreases with the collapse of the respective microgel compartment. We are convinced that, beyond its power to resolve nanoscopic polymer structures, the presented PAINT approach can give unprecedented insights into the environmental conditions within such polymer structures on the nanoscale. In our study, we found that the majority of Nile Red labels reported surprisingly apolar local surroundings. Such information can currently not be obtained by any other experimental technique. Not only does it present a novel method to analyse compartments of synthesized nanostructures, but it is essential, for example, for the design of drug delivery systems, for understanding the performance of microgels in extraction and separation processes and for their application in multi-step catalysis.

chemsum

{"title": "Microgel PAINT \u2013 nanoscopic polarity imaging of adaptive microgels without covalent labelling", "journal": "Royal Society of Chemistry (RSC)"}

transmembrane_anion_transport_mediated_by_halogen_bonding_and_hydrogen_bonding_triazole_anionophores

## Abstract: Transmembrane ion transport by synthetic anionophores is typically achieved using polar hydrogen bonding anion receptors. Here we show that readily accessible halogen and hydrogen bonding 1,2,3triazole derivatives can efficiently mediate anion transport across lipid bilayer membranes with unusual anti-Hofmeister selectivity. Importantly, the results demonstrate that the iodo-triazole systems exhibit the highest reported activity to date for halogen bonding anionophores, and enhanced transport efficiency relative to the hydrogen bonding analogues. In contrast, the analogous fluoro-triazole systems, which are unable to form intermolecular interactions with anions, are inactive. The halogen bonding anionophores also exhibit a remarkable intrinsic chloride over hydroxide selectivity, which is usually observed only in more complex anionophore designs, in contrast to the readily accessible acyclic systems reported here. This highlights the potential of iodo-triazoles as synthetically accessible and versatile motifs for developing more efficient anion transport systems. Computational studies provide further insight into the nature of the anion-triazole intermolecular interactions, examining the origins of the observed transport activity and selectivity of the systems, and revealing the role of enhanced charge delocalisation in the halogen bonding anion complexes. ## Introduction The development of supramolecular anionophores for transmembrane ion transport is driven by their potential utility as tools for studying ion transport processes and as therapeutics for diseases arising from mis-regulation of protein ion channels. Signifcant effort has been devoted to designing mobile carrier systems with high anion transport activity in vesicles (particularly for chloride), and more recently, in cells. As with naturally occurring ion transporters, anion selectivity is crucial, and depends on the delicate balance between transporter anion binding selectivity and anion desolvation. Selectivity for chloride over proton transport (or the functionally equivalent hydroxide) is particularly necessary for applications in which dissipation of transmembrane pH gradients must be avoided. The naturally occurring anionophore, prodigiosin, 11 and its synthetic analogues, 12 are known to uncouple H + -ATPases or neutralise organelles by facilitated H + /Cl symport, and for this reason are promising candidates for anti-cancer agents. Conversely, anionophores designed for treating diseases arising from misregulated chloride channels, including cystic fbrosis and Bartter syndrome, must have high Cl > H + /OH selectivity to avoid toxicity arising from disrupting transmembrane pH gradients. However, designing transporters with such selectivity remains a signifcant challenge. Polar NH hydrogen bond (HB) donors (such as ureas or squaramides) which are typically used in supramolecular anionophores exhibit no intrinsic Cl > H + / OH selectivity. 10 Improved chloride selectivity has been achieved by designing anionophores which encapsulate the anion, such as tripodal or cholapod-based receptors, but in general, more acidic NH HB-donors that are required for anion binding and efficient transport lead to decreased Cl > H + /OH selectivity. 10 C-H hydrogen bonding or C-I halogen bonding (XB) interactions have recently emerged as potential alternatives to the classical N-H or O-H HB donors for anion transport. Compared to classical HB, these interactions are less hydrophilic, encounter diminished dehydration penalties, and are less inclined to promote detrimental transporter aggregation. 16 XB in particular can exhibit superior anion binding affinity in competitive polar organic 17 or aqueous media to hydrogen bonding, pointing to its potential utility in transmembrane anion transport systems. However, to date only a handful of XBmediated anion transport systems have been reported, exploiting iodo-perfluoro alkane and arene derivatives for anion recognition and transport, and examples of transporters mediating anion transport solely through C-H HB interactions are also rare. 16, The potential power of XB for anion transport was exemplifed by Matile and co-workers who reported that gaseous iodotrifluoromethane facilitates anion transport when bubbled through the vesicle solution. 25 Herein we show that simple acyclic XB or HB triazoles (Fig. 1) can efficiently mediate anion transport across lipid bilayer membranes with an unusual anti-Hofmeister selectivity. Importantly, we demonstrate unprecedented activity for XB anionophores with up to two orders of magnitude improvement over the previously reported iodo-perfluorobenzene systems, and reveal their remarkable intrinsic Cl > OH selectivity. Iodo-and proto-triazoles are versatile motifs readily accessible via Cu-catalysed azide-alkyne click chemistry, 32 and these results demonstrate their potential for applications in synthetic anion transport systems. ## Synthesis and anion recognition properties Acyclic 5-iodo/proto/fluoro-1,4-disubstituted-1,2,3-triazoles, of the general form shown in Fig. 1, were prepared as minimalistic scaffolds with which to explore the intrinsic activity and selectivity of the hydrophobic C-I XB-and C-H HB-mediated anion transport processes. The family of compounds spans a range of iodo-triazole derivatives, and their proto-triazole analogues, with varied electron-withdrawing/donating substituents on the aryl group and varying alkyl chain lengths. The library of 5-iodo-1,4-disubstituted-1,2,3-triazoles (compounds 1b-9b) were prepared by Cu-catalysed azide-alkyne cycloaddition of the respective alkyl azide and iodo-aryl-alkyne, in the presence of tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA) (Scheme 1). The analogous proto-triazoles (compounds 1a-9a) were prepared from the corresponding alkyne. As control compounds, the fluoro-triazole analogues of 1 and 9, namely 1c and 9c, which are unable to form HB or XB interactions, were also prepared from the corresponding iodo-triazole by heating with potassium fluoride. Full synthetic procedures and characterisation are available in the ESI. † The anion recognition capability of representative prototriazole 1a and iodo-triazole 1b were frst investigated by 1 H NMR binding titrations with Bu 4 N + X in d 6 -acetone, monitoring the binding induced chemical shift perturbations of the proto-triazole H and adjacent aryl protons. For each anion, the data could be ftted to a 1 : 1 binding isotherm (see ESI †) and the 1 : 1 stoichiometric association constants determined using the Bindft program (Table 1). 33,34 The data revealed an overall selectivity trend of Cl > Br > I for both 1a and 1b, and approximately one order of magnitude enhancement of halide binding affinity for the XB iodo-triazole receptor 1b compared to HB proto-triazole 1a. For both compounds, no measurable binding of nitrate was observed under these conditions. ## Transmembrane anion transport activity The ability of anionophores 1-9a/b to mediate OH /Cl transmembrane anion transport was frst determined in 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine large unilamellar vesicles (POPC LUVs), loaded with 8-hydroxypyrene-1,3,6-trisulfonate (HPTS) and buffered to pH 7.0 in NaCl solution. HPTS is a pH-sensitive fluorophore, which allows for ratiometric determination of the internal pH of the LUVs. A pH gradient was applied across the membrane by addition of a base pulse, followed by addition of the carrier as a DMSO solution (<0.5% v/v). The ability of the anionophore to dissipate the pH gradient by transmembrane OH /Cl exchange was determined by recording the change in the HPTS emission, I rel (l em ¼ 510 nm), with time following excitation at l ex ¼ 405/465 nm (Fig. 2a). At the end of each experiment, excess detergent (Triton X-100) was added to lyse the vesicles for calibration of the emission intensity. The HPTS assay was used to determine the concentration dependence of the activity of each anion carrier. Fig. 2a shows representative data for the activity of compound 1b at a range of concentrations. The fractional activities (y, the relative intensity at 288 s, immediately prior to lysis) were plotted as a function of concentration (Fig. 2b) and ftted to the Hill equation (eqn (1)). Original data for all other compounds is available in the ESI. † The Hill equation here is used to describe the dependence of the fractional activities y on the n th power of the anionophore concentration x, and facilitates comparison of the relative activities of transporters through an effective concentration value (EC 50 ) required to reach 50% activity in the given assay (eqn (1)). Compatibility with the Hill equation reveals endergonic self-assembly of the active supramolecule, and Hill coef-fcients n > 1 are indicative of the stoichiometry of the unstable supramolecular assembly that exists in the presence of an excess of uncomplexed anionophores. 35 EC 50 values, hill coefficients and calculated partition coefficients (clog P) values for the proto-and iodo-triazoles are shown in Table 2. clog P values were calculated using the VCCLab software. 36 In general, the electron withdrawing 3,5-bis(trifluoromethyl) benzene and 4-nitrobenzene substituents led to excellent activity despite the simplicity of the anionophores (compounds 1-3, and 7-9, respectively). The electron defcient halogen bonding transporters 1b and 9b are approximately 3 more active than the most active XB anionophore reported to date, iodoperfluorohexane (EC 50 ¼ 3.1 mM); 25 and notably over two orders of magnitude more active than the archetypal XB donor iodoperfluorobenzene under comparable assay conditions (EC 50 ¼ 260 mM). 25,28 Electron donating 4-n-butylbenzene substituents (compounds 4-6) conversely decreased activity, consistent with the expected decreased XB and HB donor strength. The correlation between activity and log P is complex, because transport efficiency is also strongly dependent on other factors including anion binding affinity and molecular size, which are not easily decoupled. 37 Nevertheless, the observed trend of maximum activity of compounds with log P of 4-6, and reduced activity observed for compounds with higher and lower lipophilicities, is broadly consistent with previous reports on Table 2 Characteristics of the HB and XB anion transporters the role of lipophilicity on HB-mediated anion carriers. 38 For compounds in which both XB and HB derivatives fall within these optimum parameters (compounds 1, 4, 7 and 8), a general enhancement of activity of around 2, and up to 8, is observed for XB systems over their HB analogues. The most lipophilic transporters (3, 5, 6) are inactive, likely arising from reduced mobility and reorganization of longer chain derivatives within the lipid bilayer. 39 Calculation of the electrostatic potential (ESP) surface for methyl-truncated analogues ‡ of 1a/b, 4a/b and 7a/b (denoted 1a 0 /b 0 , 4a 0 /b 0 and 7a 0 /b 0 respectively) reveals the characteristic iodine-centred sigma-hole 40,41 associated with halogen bonding donors, and the accumulation of d+ surrounding the triazole protons active in hydrogen bonding (Fig. 3). Comparison of the Hammett s values for each substituent demonstrate that the more electron-withdrawing substituents result in a more positive value for the ESP maximum, 42 leading to a stronger XB-/HBintermolecular interaction. This correlates with the observed greater activity of 1a/b and 7a/b over 4a/b (Table 2). ## Evidence for XB/HB-anion interactions and mobile carrier mechanism To confrm the role of the C-I XB and C-H HB interactions in the anion transport process, we investigated the anion transport capability of fluoro-triazole derivative 1c, which is unable to form XB or HB interactions through the triazole motif. Inactivity of 1c in the HPTS assay confrmed the role of the XB and HB intermolecular interactions in the transport processes (Fig. 4a). Similar behaviour was observed for nitro-phenyl fluoro-triazole derivative 9c. Inactivity of the transport systems 1-9 reported here in the carboxyfluorescein dye leakage assay rules out non-specifc leakage by these systems (ESI †). Replacement of the zwitterionic POPC lipids in the HPTS assay with anionic egg yolk phosphatidylglycerol (EYPG) lipids led to a signifcant decrease in observed activity, consistent with the requirement for formation of an anionic complex in the rate limiting transport process. Transport activity at 25 C with XB transporter 1b and dipalmitoylphosphatidylcholine (DPPC) lipids, under otherwise identical conditions, was negligible (Fig. 4b). The lipid gel phase inhibits translation of mobile carriers through the membrane which are otherwise mobile in a fluid lipid phase. The gel to fluid phase transition for DPPC lipids is 41 C, and repeating the assay above this temperature (45 C) restored anion transport activity. The observed temperature dependence is indicative of a mobile carrier mechanism, and rules out formation of a membranespanning supramolecular channel structure whose activity would be independent of the lipid phase. ## Anion selectivity Dissipation of the transmembrane pH gradient measured in the HPTS assay by the carrier species can in principle occur through either cation (H + /M + ) or anion (OH /A ) antiport (exchange), or H + /A symport (co-transport) mechanisms. The activities of HB transporter 1a and XB transporter 1b were not affected by isoosmolar replacement of the external Na + cation with Li + , K + , Rb + or Cs + (see ESI †), indicative of selective anion transport (OH /A antiport or H + /A symport) rather than H + /M + cation antiport. Further evidence to support the cation-independent transport mechanism was provided by conducting analogous experiments in the presence of sodium gluconate, a large hydrophilic anion (ESI †). 10 The absence of detectable transport indicates that, as expected, neither OH /gluconate antiport or H + /gluconate symport mechanisms are active because of the insurmountable dehydration penalty of gluconate, and also that the alternative H + /Na + cation antiport process is negligible. OH /A antiport and H + /A symport mechanisms are functionally equivalent and cannot be distinguished through these transport assays. However, the low basicity of triazole derivatives (pK a H $ 0-1) 43 suggests that H + /A symport (achieved via triazole protonation and XB-/HB-mediated anion recognition) is unlikely to contribute to any signifcant extent to the ion transport process at neutral pH. This is consistent with the activity of previously reported XB iodofluoroalkene/arene transporters in same assay, which do not possess any basic atoms and are therefore most likely operate through an anion antiport mechanism. 25 For simplicity, we refer to the transport process from here on in as OH /A antiport. To examine the relative rates of Cl vs. OH transport, we repeated the HPTS assay with the addition of carbonyl cyanidep-trifluoromethoxyphenylhydrazone (FCCP), a weak acid protonophore at a low concentration (0.25 mM) insufficient to cause activity alone. FCCP transports protons via transmembrane shuttling of both protonated and deprotonated forms of the molecule. Enhancement of activity by FCCP is indicative of a rate limiting electrogenic OH transport process, because FCCP decouples the anionophore-mediated OH /Cl antiport process (Fig. 5a) by facilitating rapid electrogenic H + transport in an overall coupled Cl /H + symport process (Fig. 5b). In this scenario, the observed rate of pH gradient dissipation reports on the rate-limiting electrogenic Cl transport process. As such, if the activity of a given carrier is invariant to FCCP addition, Cl transport must therefore be rate limiting in the antiport mechanism (i.e. slower than electrogenic OH transport). The EC 50 values for compounds 1a and 1b with FCCP are shown in Table 3. The activity of HB anionophore is invariant to FCCP addition, demonstrating rapid OH transport and rate limiting Cl transport by 1a. In contrast, a four-fold increase in activity for XB transporter 1b is observed in the presence of FCCP, revealing rate limiting OH transport in its absence. The ratio of EC 50 values in the absence and presence of FCCP (Table 3, column 4) therefore reports on the relative Cl > OH selectivity of the system (for the given conditions and ion concentration gradient), and reveals the unusual Cl > OH selectivity of the XB transporter 1b. Notably, transporter 1b in the presence of FCCP achieved nanomolar activity (EC 50 ¼ 300 nM) in the HPTS assay, which to the best of our knowledge is the highest activity reported to date for an XB anion transporter under comparable experimental conditions. This highlights the potential of iodotriazoles as synthetically accessible and versatile motifs for developing more efficient anion transport systems. In the presence of FCCP, the observed transport kinetics report on the rate limiting A transport for a given anionophore. This allows the relative selectivity of both XB and HB transporters for each anion to be investigated, enabling direct comparison of the relative rates of electrogenic A transport (Fig. 6). HB transporter 1a exhibits an overall selectivity profle of Cl > Br > I > NO 3 (Fig. 6, black data). This selectivity is remarkable because in general hydrogen bonding anion transporters exhibit Hofmeister selectivity, whereby more weakly hydrated anions are selectively transported across the membrane due to ease of desolvation. The analogous XB anionophore 1b exhibits a selectivity profle of NO 3 > Cl > Br > I . Comparison of the relative activities for both transporters in the absence of FCCP allows placement of OH into the overall selectivity trends, revealing OH > Cl > Br > I > NO 3 behaviour for 1a and NO 3 > Cl > Br > OH > I behaviour for 1b. Similar selectivity trends for analogous compounds 9a and 9b were also observed in these assays (see ESI †). Overall these transport experiments reveal (i) enhanced activity of halogen bonding anionophores vs. hydrogen bonding analogues across all anions, (ii) unusual Cl > OH selectivity for the halogen bonding anionophores and (iii) anti-Hofmeister bias for halide transport, demonstrating the dominant role of ion binding to the anionophore and correlating with NMR determined anion binding affinities (see Table 1). § ## Computational studies into anionophore-anion interactions Computational studies were used to probe the binding energies and structures of the anionophore-anion complexes. Calculation of the relative binding enthalpies for both 1 : 1 and 2 : 1 model anionophore-anion complexes in CHCl 3 implicit solvent as a membrane-mimetic environment reveals that DH bind (2 : 1) > 2 DH bind (1 : 1) (Fig. 7), and that 2 : 1 binding is enhanced by p-stacking between the arenes (see ESI †).{ Bidentate binding was found to be preferred in the model 1 : 1 HB system 1a 0 involving hydrogen bonds from both the alkyl 3 )-H1 and triazole C(sp 2 )-H2 (Fig. 7a). This fnding is in line with the result of the ESP calculations in Fig. 3. Monodentate XB interactions are observed with 1b 0 (Fig. 7b). Interestingly, our results indicate that in the 2 : 1 complex, four HB C-H/anion interactions per anion exists for 1a 0 (Fig. 7a), while only one XB C-I/anion interaction is present for 1b 0 (Fig. 7b). Additionally, bond order and second-order perturbation theory analysis, which allows the quantifcation of bond energies (E (2) ) 54 indicates that addition of the second transporter ligand weakens the total XB interactions compared with the 1 : 1 complex (E (2) XB ¼ 40.6 vs. 30.7 kcal mol 1 ). However, this is partially compensated by dispersion interactions between the two transporter ligands (E T1-T2 ¼ 16.0 kcal mol 1 ) and a C-H interaction (10.1 kcal mol 1 ), which leads to an overall larger stabilisation energy (E (2) total ¼ 56.8 kcal mol 1 ). Analysis of the partial charges on each anion in the 1 : 1 and 2 : 1 binding modes suggest a correlation between the rate of transport and the ability of the transporter to stabilise the anion charge (Fig. 8). For example, for Cl , Br and I , the negative charge is more effectively delocalised over the XB anionophore than the HB analogue. This is consistent with previous results from Donor K-edge X-ray Absorption Spectroscopy experiments which revealed signifcant covalency in the XB interaction between related iodo-triazole XB donors and chloride anions (comparable to that in transition metal chloride complexes), in contrast to proto-triazole analogues with negligible charge transfer contribution to the HB interaction. 55 Improved charge Fig. 7 Characterisation of 1 : 1 and 2 : 1 binding of (a) 1a 0 and (b) 1b 0 to chloride. Enthalpies, entropies and free energies of binding (kcal mol 1 ) were calculated at the [SMD(CHCl 3 )-DLPNO-CCSD(T)/def2-TZVP (ma-def2-TZVP on Cl, I)//uB97X-D3/def2-SVP (ma-def2-SVP on Cl, I)] level of theory. Mayer bond orders (dashed lines), their associated HB/XB bond energies (E (2) HB/XB ), and interaction energy between ligands (E (2) T1-T2 ) in kcal mol 1 were calculated at the [NBO/SMD(CHCl 3 )-uB97X-D3/def2-SVP (ma-def2-TZVP on Cl, I)//uB97X-D3/def2-SVP (ma-def2-SVP on Cl, I)] level of theory. Free energies were calculated using a quasi-RRHO approximation 53 and corrected for a 1 M standard state at 298.15 K. delocalisation over the anion-triazole complex is expected to decrease the barrier to transport across the hydrophobic membrane (Fig. 8), and this is consistent with the experimentally observed selectivity trend of the XB anionophores enhancing rates of anion transport over the analogous HB system. A different behaviour is observed for nitrate, where the observed efficient nitrate transport does not correlate with the increased calculated partial charge. We suggest that this behaviour is due to the greater number of thermally-accessible conformations for the binding of the trigonal nitrate anion, each of them having a different charge distribution. In this case, the most stable conformer in solution is likely not the most active for membrane transport (see ESI †). ## Conclusions We have shown that acyclic halogen bonding iodo-triazoles can efficiently transport anions across lipid bilayer membranes, with enhanced activity in comparison to hydrogen bonding proto-triazole analogues. Despite the relative simplicity of the design, the iodo-triazole systems are the most active halogen bonding anionophores reported to date. Anion transport experiments reveal unusual anti-Hofmeister selectivities, and a remarkable intrinsic Cl > OH selectivity for the halogen bonding systems. Experimental and computational studies provide further insight into the binding modes of the anionanionophore complexes. Calculations demonstrate that the strength of the non-covalent interactions are sufficient to overcome the entropic cost of 2 : 1 complex formation, and reveal the remarkable ability of the halogen bonding anionophores to delocalise the anion charge over the complex, which correlates with the enhanced anion transport capability observed by experiment. This work demonstrates the utility of the synthetically accessible and highly versatile XB and HB triazole motif for anionophore design. The observed Cl > OH selectivity of the iodotriazole derivatives also suggest that such motifs may provide the starting point for designing novel ionophores with enhanced selectivity for potential future application in channelopathy therapeutics. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Transmembrane anion transport mediated by halogen bonding and hydrogen bonding triazole anionophores", "journal": "Royal Society of Chemistry (RSC)"}

modulation_of_inherent_dynamical_tendencies_of_the_bisabolyl_cation_via_preorganization_in_epi-isozi

## Abstract: The relative importance of preorganization, selective transition state stabilization and inherent reactivity are assessed through quantum chemical and docking calculations for a sesquiterpene synthase (epi-isozizaene synthase, EIZS). Inherent reactivity of the bisabolyl cation, both static and dynamic, appears to determine the pathway to product, although preorganization and selective binding of the final transition state structure in the multi-step carbocation cascade that forms epi-isozizaene appear to play important roles. ## Introduction Many factors have been proposed as contributors to selectivity control in terpene synthases, e.g., reactant preorganization, geometric constraints imposed by the enzyme active site (not only on the reactant but on reactive species generated from it), 3, selective oriented intermolecular interactions (primarily p/p, C-H/p and C-H/lone pair) with intermediates and transition state structures (TSSs), 16,17 and inherent reactivity of carbocations generated from the reactant-both in terms of underlying potential energy surfaces (PESs) for carbocation rearrangements and inherent dynamical tendencies. Herein we describe computations that bear directly on the relative importance of all of these factors for a sesquiterpene synthase-epi-isozizaene (7; Fig. 1) synthase (EIZS)-that has received considerable interest from organic chemists and mechanistic enzymologists over the past decade. 9, To our knowledge, this is the frst report that parses out the relative contributions of these factors for any enzyme (note that inherent dynamical tendencies of a substrate is by far the least studied factor of those described). EIZS catalyzes the polycyclization of an acyclic substrate (farnesyl diphosphate, FPP), into a complex polycycle possessing three stereogenic centers via the 11 steps shown in Fig. 1. These reaction steps are proposed on the basis of previous quantum chemical calculations on reactions of the bisabolyl cation (1). 3 Note that step VII is predicted to have a low barrier or no barrier, depending on the conformation of homobisabolyl cation 2, and step IX involves the merging, asynchronously, of two 1,2-alkyl shifts into a concerted process. While several conformations of species involved in this pathway were examined previously, 3 we have now performed an exhaustive conformational search on the TSS for step VI with an eye toward elucidating the consequences of conformational preorganization on mechanism and selectivity. Relevant conformations were then docked into EIZS to assess the importance of enzyme-substrate interactions. ## Conformational concerns First, the conformational landscape associated with step VIthe step after initial formation of the 6-membered ring that is the hallmark of this branch of sesquiterpene structural space- was examined. The previously reported global minimum of the bisabolyl cation (1) 3 was used as a starting point for conformational searching. Using Spartan10, 32 the Merck Molecular Force Field (MMFF94) 33 was employed for a systematic search in which torsions about all rotatable bonds were sampled, yielding 86 conformers of 1. These conformers were then optimized in the gas phase at the mPW1PW91/6-31+G(d,p) 23a,34,35 level of theory using the Gaussian09 software suite, 36 and the lowest energy minimum was designated 1a. Manual inspection of geometries and free energies led to the removal of duplicate structures, leaving 67 unique conformations. Next, scans of forming and breaking C-H bond distances were performed for each minimum to obtain approximate geometries for the 1 / 2 1,2-hydride shift TSS. These structures were then fully optimized with mPW1PW91/6-31+G(d,p). This procedure led to 67 unique TSSs covering a range of barriers of 5.6 to 12.9 kcal mol 1 (full details in ESI †). These results reveal the magnitude of the conformational problem faced by EIZS, a scenario not unique to this terpene synthase. 37 Each of the 67 TSSs was then docked into EIZS (PDB ID 3KB9) using the Fast Rigid Exhaustive Docking (FRED) program in the Openeye software suite. The pyrophosphate group lost in step IV was considered to be part of the enzyme active site and was held fxed in these docking simulations. Rankings and docking scores (here, reflecting primarily shape complementarity) for all TSSs can be found in the ESI, † but results for the TSSs that lead to the relative stereochemistry in epi-isozizaene if the reaction proceeds without signifcant poststep VI conformational changes (TS1b-TS1d, Fig. 2), along with several other representative structures, are shown in Table 1. TS1b-TS1d have some of the best docking scores (ranking 2 nd -4 th ), despite not having the lowest free energies of the 67 TSSs; they rank 62 nd , 32 nd , and 65 th , respectively. Note, however, that their predicted free energies are within 6 kcal mol 1 of that of the lowest energy TSS, TS1e. TS1e is not predicted to be among the best suited for docking to the EIZS active site, suggesting that the shape of the active site plays a role in selecting the TSSs for 1,2-hydride shift that have conformations productive for subsequent reactions in the epi-isozizaene-forming pathway. ## Product selectivity Each carbocation (1-6) in Fig. 1, if encountered as an intermediate with a signifcant lifetime, represents a potential source of byproducts for EIZS. epi-Isozizaene is the predominant natural product produced by EIZS from Streptomyces coelicolor; 9,24,27 at 20 C, a product ratio of 93 : 5 : 1 : 1 is reported for 7 : 8 : 9 : 10 (Fig. 3; Table 2; this table also shows product distributions at other temperatures). Although EIZS does not produce a sole product, its selectivity is impressive given the number of 1. The carbons between which a bond will form in the 2 / 3 reaction (C6 and C10) are labeled in each TSS. In docked structures, the substrate is blue, protein sidechains are grey, the diphosphate group is red/orange and Mg ions are purple. potential exit channels from the epi-isozizaene-forming pathway and the inherent conformational flexibility of the species encountered en route to epi-isozizaene. Dynamical tendencies. To explore whether or not the observed EIZS product distribution corresponded to the inherent dynamical preferences (here not necessarily implying non-transition state theory behavior) of the substrate, direct dynamics simulations were performed, using Progdyn, 48 on the fve TSS conformations shown in Fig. 2. TS1b, TS1c and TS1d correspond to productive conformations of the TSS with respect to subsequent events en route to epi-isozizaene. TS1a is a non-productive conformation, despite having the best docking score. TS1e is inherently the lowest energy TSS but was not predicted to dock well (vide supra). Classical trajectories were allowed to propagate using the Verlet algorithm for 2500 fs at 298.15 K with 1 fs time steps. Starting points for dynamics trajectories were generated from a Boltzmann sampling of vibrations. 58,59 Fifty trajectories were generated for each TSS and the resulting product distributions after 2500 fs are given in Table 2. Only trajectories evolving in the epi-isozizaene (7) direction were considered and only a small amount of recrossing was observed. Fig. 4 shows structures of resulting carbocations not covered in Fig. 1. The product distributions from our dynamics calculations reveal: (1) TSS conformations TS1a, TS1b, and TS1e are not predisposed to proceed along the reaction coordinate towards 7, since trajectories for none passed the homobisabolyl cation (2) or its non-productive cyclization product 11. The inability of TS1a and TS1e to proceed toward product was expected on conformational grounds, while the inability of TS1b to do so was not so certain at the outset. The position of the "free" isoprenyl group in TS1b is apparently far enough away from the 6membered ring (Fig. 2; note relative position of C6 and C10) that bringing these two groups close enough together for reaction requires intervention. (2) TSS conformations TS1c and TS1d, which differ from each other in the puckering of their cyclohexenyl rings and which display a more productive orientation of their isoprenyl groups than does TS1b (Fig. 2), proceeded readily to the acorenyl cation (3) region and some trajectories made it to the cedryl cation ( 4) region (a greater number for the more compact TS1e; molecular volumes in Table 1). ( 3) Dynamics trajectories for no TSSs examined reached the prezizyl (5) or zizyl (6) cation regions. These results suggest that conformationally preorganized (in this case, compact) TSSs can proceed directly along the pathway to the cedryl cation (4) region without spending any signifcant amount of time near cations 2 or 3, consistent with the absence of byproducts expected to be derived from these two cations in the experimental product distributions. The fact that cations 4 and 11 are observed in our dynamics simulations is consistent with the observation of sesquiterpenes derived from them in the 2010 report on EIZS. 9 Minor products 9 and 10 would not show up in our dynamics studies, since these would be formed from deprotonation of species preceding the bisabolyl cation on the epi-isozizaene-forming reaction coordinate. The observation of cations 12 and 13 for TS1d is consistent with: (a) products of their deprotonation be formed in very small amounts, (b) their conversion, given more time and/or direct enzymatic intervention, to experimentally detected products, or (c) prevention of their formation by EIZS. The remainder of our discussion is focused on TS1d, since this TSS appears to best suited for epi-isozizaene formation, being the TSS that produced the most trajectories that passed the cation 3 region. For TS1d, approximately half of the trajectories exited the cation 2 region within 500 fs (some proceeding towards 3, others towards 11). Approximately 80% of the trajectories that ultimately reached cation 4 exited the cation 2 region within this frst 500 fs, proceeding directly to the cation 4 region. The free energy surface for conversion of bisabolyl cation 1 to zizyl cation 6 via TS1d (Fig. 5, solid lines) is consistent with dynamics trajectories not passing cedryl cation 4. The following three factors are expected to correlate with whether or not a barrier will "block" a direct trajectory: 43,52 (1) the height of the barrier, (2) the height of the downslope preceding the barrier, and (3) whether or not the vibrations occurring along the Fig. 3 Natural products produced by EIZS. 27 Table 2 First three rows: product distributions (%) for EIZS reported in 2010 9 and 2014. 27 Remaining rows: product distributions (%) from dynamics calculations. For EIZS, products correspond to species shown in Fig. 1 and 3 or species derived from deprotonation of indicated carbocations (1, 4, 11). For dynamics calculations (rows with bold TSS labels in first column), products correspond to carbocations or carbocation precursors to neutral species (7-10) and are listed as a percent to allow a direct comparison to experiment; 50 trajectories were generated for each transition state (the result of calculations with two different levels of theory on the same reaction) increased the percentage of direct trajectories. 43 With regard to factor 3, Carpenter and co-workers showed, for ring-opening/1,5-hydride shift of [2.1.0]bicyclopentene, that kinetic energy associated with traversing an initial barrier was only directly accessible to vibrations of the same symmetry in subsequent steps. 52 It is difficult to determine how much dynamic matching of vibrational modes contributes to the viability of the 4 to 5 reaction (a dyotropic or "double-shift" reaction). 3,21,63,64 While formation of both involves stretching/ compressing of the same bond (the C2-C11 bond that is made in forming 4 breaks in forming 5), the imaginary frequency for TS3-4 corresponds primarily to a twisting of the molecule around the C6-C10 bond while that for TS4-5 corresponds primarily to formation of the C3-C11 bond (Fig. 1). Active site restrictions. To assess the effects of shape selection in enzyme-substrate binding on the energetics of epi-isozizaene formation, we performed automated docking for intermediates and TSSs from Fig. 5 using FRED. 38-40 Docking scores for stationary points involved in zizyl cation (6) formation via TS1d (TS1-3) are shown in Table 3. Two poses were considered for each stationary point: (1) the pose that yielded the best docking score and (2) the best pose that was productive for subsequent reaction without Fig. 5 Computed mPW1PW91/6-31+G(d,p) free energy profile for formation of 6 via TS1d (here, labeled TS1-3) without conformational changes of intermediates (energies are relative to the global minimum of 1; note that for this series of conformations, 2 is not a minimum), along with productive docking poses (substrate is blue, protein sidechains are grey, the diphosphate group is red/orange, Mg ions are purple, groups involved in bond making/breaking are highlighted in transition state structures). Dotted lines and arrows indicate the qualitative effects of binding to EIZS, estimated based on computed docking scores for productive poses. signifcant "tumbling" in the active site. The latter was chosen on the basis of similarity to the best pose of TS5-6 (Fig. 5). Although FRED docking scores are not predicted binding energies, if one assumes that they correlate qualitatively with binding energies then effects on the energetics of epi-isozizaene formation can be assessed. As shown in Fig. 5 (dotted lines; only effects corresponding to differences in docking scores of >0.5 are shown), the frst two TSSs following the bisabolyl cation, along with the minima directly following them, are selectively destabilized relative to the bisabolyl cation when considering the best productive pose (also true for the best docked pose). Conversely, the fnal TSS, leading to the zizyl cation, is selectively stabilized. The net effect of this modulation in relative energies is to promote passage from 4 to 5 to 6. Both the 4 / 5 and 5 / 6 barriers would be lowered by selective complexation, and the height of the downslope leading to 4 would be increased, thereby reducing the lifetime of 4 and perhaps allowing for the direct passage to 5 not observed in our dynamics calculations in the absence of EIZS. The docking approach used scores primarily on the basis of the complementarity of cation and active site shape, 38-40 consistent with the idea that terpene synthase active sites resemble structures occurring later along carbocation cyclization/rearrangement reaction coordinates. 65 Note also that the substrate volume decreases monotonically from 1 to 5, but then increases for the fnal two stationary points (Table 3). Although one can identify specifc C-H/ H-C, C-H/p and C-H/O contacts in docked structures (see ESI †), quantifcation of their effects on the reaction pathway will require more advanced methods (vide infra). ## Conclusions On the basis of our results, avoidance of the previously postulated secondary carbocation between 4 and 5 can be ascribed to the nature of the PES in the absence of the enzyme, 3,24,66 and avoidance of byproducts derived from putative intermediates between 1 and 4 can be ascribed to inherent dynamical tendencies. 67 Enzyme-enforced conformational restriction (of reactant and subsequent species) clearly also plays a role in directing the reactant toward TSSs that are productive for epiisozizaene formation and appears also to play a role in promoting the conversion of 4 to 5 to 6. That leaves only fnetuning to be ascribed to specifc intermolecular electrostatic interactions (with OPP and/or active site aromatics) and/or effects of enzyme dynamics (rather than inherent substrate dynamics; tunneling may also play a role 37 ); these issues will be addressed in future quantum mechanical molecular mechanics (QM/MM) and theozyme 18, studies on EIZS. Note how little is left to explain; while EIZS clearly plays keys roles in promoting pyrophosphate dissociation and preventing premature quenching of carbocations, we show here that its product distribution can be rationalized in large part on the basis of inherent carbocation reactivity and shape selection. 74 Table 3 Docking scores (unitless) for the stationary points described in Fig. 5. Included are the best docking score for each stationary point (column 2) and the docking scores of the poses shown in Fig.

chemsum

{"title": "Modulation of inherent dynamical tendencies of the bisabolyl cation via preorganization in epi-isozizaene synthase", "journal": "Royal Society of Chemistry (RSC)"}

identification_and_physical_characterization_of_a_spontaneous_mutation_of_the_tobacco_mosaic_virus_i

## Abstract: Virus-like particles are an emerging class of nano-biotechnology with the Tobacco Mosaic Virus (TMV) having found a wide range of applications in imaging, drug delivery, and vaccine development. TMV is typically produced in planta, and, as an RNA virus, is highly susceptible to natural mutation that may impact its properties. Over the course of 2 years, from 2018 until 2020, our laboratory followed a spontaneous point mutation in the TMV coat protein-first observed as a 30 Da difference in electrospray ionization mass spectrometry (ESI-MS). The mutation would have been difficult to notice by electrophoretic mobility in agarose or SDS-PAGE and does not alter viral morphology as assessed by transmission electron microscopy. The mutation responsible for the 30 Da difference between the wild-type (wTMV) and mutant (mTMV) coat proteins was identified by a bottom-up proteomic approach as a change from glycine to serine at position 155 based on collision-induced dissociation data. Since residue 155 is located on the outer surface of the TMV rod, it is feasible that the mutation alters TMV surface chemistry. However, enzyme-linked immunosorbent assays found no difference in binding between mTMV and wTMV. Functionalization of a nearby residue, tyrosine 139, with diazonium salt, also appears unaffected. Overall, this study highlights the necessity of standard workflows to quality-control viral stocks. We suggest that ESI-MS is a straightforward and low-cost way to identify emerging mutants in coat proteins.Biotechnology based on non-infectious viral nanoparticles has been an emerging topic of research for the last two decades, with considerable advancement toward biomedical translation occurring recently 1, 2 . Nature has provided ample source material-viruses that infect humans, bacteria, and plants all fair game when engineering a new and more efficacious drug delivery, vaccine, and imaging platform. Plant-based viruses have gained attention in the field of biotechnology for several reasons. In addition to being robust, at least insofar as proteinaceous materials go, they have the benefit of being literally farmable-the viruses can be produced in plant tissue in a scalable approach. For instance, Medicago Inc, a biotechnology company presently based in Quebec City, has begun phase II-III trials of a plant-produced vaccine against COVID-19 3, 4 .Likely the most studied virus in plant-harvested viral-nanotechnology is the Tobacco Mosaic Virus (TMV), a noninfectious, plant-based virus-like particle 2, 5-8 . It is rigid, monodisperse, thermostable, functionalizable 2, 6, 8 , and biocompatible [9][10][11] , giving it a wide range of applications including as a scaffold for biomolecules and small molecules, nanocontainers for drug delivery 5,9,[12][13][14][15][16][17][18][19] , and as a platform for vaccine development and testing 10,11,[20][21][22][23] .TMV is a positive-sense single-stranded RNA virus under the Tobamovirus genus. Its RNA encodes four proteins: two 126-and 183-kDa replicase proteins, a 30 kDa movement protein, and a 17.5 kDa coat protein 5,24 . The intact TMV contains 2130 copies of the coat protein assembled into a helical rod with a length of 300 nm, an outer diameter of 18 nm, and a pore diameter of 4 nm. Both exterior and interior surfaces of this hollow rod can be modified on solvent-exposed amino acids 6,24 . We, like most groups that produce plant-based viral www.nature.com/scientificreports/ nanoparticles, use an extremophile strain of the Australian plant Nicotiana benthamiana for large-scale production of TMV. It is the most widely used host for plant virus replication as it is very susceptible to infection. An RNA silencing gene in N. benthamiana (NbRdRP1m) is mutated and less active, contributing to the species' susceptibility of RNA-virus infection . Single-stranded RNA viruses, like TMV, are very susceptible to mutation 28 . Though the mechanism is not established yet, Sanjuán et al. (2016) suggested that the genetic material in viruses with single-stranded genomes are more exposed to oxidative deamination and other chemical stress 28 . In fact, mutant populations are ubiquitous in plant RNA viruses and their presence is rarely considered significant until the mutant phenotype dominates 29 . RNA viruses, except Coronavirus, contain RNA polymerases that lack 3′ → 5′ exonuclease proofreading mechanisms, which makes them more prone to error than DNA viruses 28,30 . One study showed that infecting tomato plants containing the Tm-2 Gene with TMV strain Ltbl, resulted in point mutations in the 30 kDa movement protein 31 . TMV is well-known to be susceptible to environmental and evolutionary stressors and the emergence of mutants should not be surprising; however, there is little discussion in the literature of how coat protein mutations might alter the viruses' physical properties vis-à-vis chemical functionalization and antibody binding. Best practices for monitoring TMV mutation and the impacts of TMV mutation on virus usability are not well established. Knowing what to look for, the relative time scale, and how to identify mutation in plant-sourced viruses would be helpful to the broader community. Here, we report the emergence of a mutant strain of TMV over the course of two years in the absence of any purposeful chemical or environmental stressor. The strain contains a single point mutation in the coat protein and was identified by electrospray ionization mass spectrometry (ESI-MS) of TMV stocks archived from January 2018 to January 2020. Over this period, the mutant strain largely displaced the wild-type strain, suggesting a competitive advantage. While we were following best practices for purifying and testing isolated TMV at the time, we have since identified additional practices that should be followed to identify emerging mutants. In many assays, the mutant was indistinguishable from wild-type. This work highlights the need for standard workflows to quality-control viral stocks in the chemical and bioengineering laboratory, particularly if they are being considered for translational purposes. ## Result and discussion Our TMV stock is grown in N. benthamiana eight weeks after germination while the plants are around 11.5 cm tall. N. benthamiana is grown in a purpose-built plant growth room with constant temperature (22 °C) and humidity (67%). TMV (15 mg) from a prior harvest is added to 1 g of silicon carbide as an abrasive and rubbed into the leaves of N. benthamiana. Once the leaves become discolored, two weeks after infection, the leaves are collected and stored at − 80 °C until the virus is extracted as previously described 12 . A small amount of TMV is always set aside for the next infection. In the literature, the TMV stock is typically characterized by electrophoretic mobility (e.g. SDS-PAGE), transmission electron microscopy (TEM), size exclusion chromatography, and intact protein mass spectrometry 6,9,12,14,24 . RNA sequencing has been done in some cases 32,33 but is not routinely reported. If mutant TMV strains have been detected by sequencing, they have not been widely reported in the literature to date. We monitor our TMV purifications using liquid chromatography mass spectrometry (LC-MS). We conducted LC-MS using ESI-MS on denatured viral samples. 20 µL of 10 mg/mL TMV was mixed with 40 µL of glacial acetic acid and the RNA removed via centrifugation. Over two years, from January 2018 till January 2020, we observed the appearance of a new mutant TMV strain (Fig. 1A). In Jan 2018, we observed the anticipated coat protein mass of 17,534 Da, the theoretical mass of wild-type TMV coat protein (wTMV) acetylated at its N-terminus 6,9,12,34 . A second peak was also observed at approximately 17,564 Da. This peak was a relatively minor peak but was consistent in every sample of extracted TMV. By Jan 2020, the peak at 17,564 Da, which we now know is a mutant TMV coat protein (mTMV), had become dominant. Aside from its mass, the mTMV is indistinguishable from wTMV. Characterization approaches that follow best practice per the literature, do not indicate the heterogeneity of the TMV preparations. When the wTMV and mTMV samples are analyzed by electromobility in SDS-PAGE and agarose gels (Fig. 1B), for example, there is no difference in protein migration. SDS-PAGE shows the typical single protein mass. Agarose, which measures the migration of intact viral nanoparticles, can be affected by both changes in mass, as well as charge, and it also showed no difference between wTMV and mTMV. The intact viral nanoparticles of both mTMV and wTMV (Fig. 1C) also show identical rod morphology in TEM. Had we not routinely conducted ESI-MS analyses of our isolates, the mutant would have gone undetected. Matrix assisted laser desorption ionization-time of flight mass spectrometry (MALDI-MS) is widely used to measure protein mass. From our assessment of the literature, most mass analyses for TMV coat proteins are conducted via MALDI-MS, including many from our group 6,14,32,35 . Nominal resolution MALDI is attractive because the instruments are low cost, sample preparation and analyses are straightforward, and the gentle ionization conveniently produces the + 1 m/z peak. However, nominal resolution MALDI-TOF typically lacks the resolving power to identify small changes in mass on a large protein 36 . To identify the precise mutation(s) responsible for the 30 Da difference between wTMV and mTMV coat protein, we adopted a bottom-up approach. We compared a sample containing 60% wTMV and 40% mTMV, to one almost entirely composed of mTMV (Fig. 2A). Tryptic digests of both samples produced seven shared peptides that covered 90% of the coat protein sequence. Of these seven peptides, only one showed over an order of magnitude difference in intensity between the two samples (Fig. 2B left). This peptide covered residues 142-158, suggesting the TMV mutation was located within these C-terminal residues of the protein. For further confirmation, peptic digests of both samples were also performed. These similarly identified a change in intensity for a peptide covering residues 151-158 (Fig. 2B right). These data show that the mutation is in one of the last 8 residues of the coat protein (Fig. 3B). We next generated a list of all possible single point mutations that would cause a mass increase of 30 ± 1 Da in the last 8 residues of the coat protein. The list contained nine possible sequences and each sequence was used to interrogate the tryptic and peptic digests of the mTMV sample. Only one sequence gave a passing result and identified the C-terminal peptide of the mTMV coat protein. This sequence contained a G155S mutation. In the two C-terminal peptides (142-158 and 151-158), collision-induced dissociation data show a serine at position 155 with consecutive fragmentation products recovered at the precise site of mutation (Fig. 2D). When using the sequence with G155S, the C-terminal peptides also had more intensity in the mTMV sample compared to the mixed sample (Fig. 2C). The difference between wTMV and mTMV is thus a G155S point mutation in the coat protein. At the nucleic acid level, the most likely mutation is a single guanine to adenine which changes the codon of residue 155 from GGU (glycine) to AGU (serine). The mutation of glycine to serine at position 155 would not be expected to drastically impact the folding or stability of either the coat protein or the intact TMV. Glycine-155 is in a loop that is exposed to solvent on the exterior-side of the TMV rod (Fig. 3A). It is not involved in contacts between the many copies of the coat protein. A glycine to serine change is also somewhat conservative with both having small side chains. Differences may occur, however, as glycine allows for greater backbone flexibility and serine brings an additional hydroxyl www.nature.com/scientificreports/ group for hydrogen bonding. Algorithms predict a slight reduction in protein stability with the G155S mutation (ΔΔG = -0.41) 37 . Given the surface localization of the G155S mutation, we sought to test its impact on TMV antibody binding and chemical modification. With identical concentrations of TMV, changes in the relative binding of polyclonal antibodies would be obvious by comparing a dilution series in an enzyme-linked immunosorbent assay (ELISA). ELISA run on both mTMV and wTMV bind TMV antibodies at the same concentrations to produce linear curves with a correlation coefficient of 0.998, indicating the binding is almost identical (Fig. 4A). Therefore, it seems unlikely this mutation would have impacted any immunological or cell studies. So far, the ELISA result shows no significant difference between the binding of the epitopes of mutant and wild-type TMVs to TMV antibodies, but if there is any change in the epitope, it might likely be missed by the method, and further study is needed. TMV is also widely used for its facile chemical modifiability at the external tyrosine residue via a diazonium coupling reaction (Fig. 4B). We were able to show that the tyrosine residue can be functionalized through diazonium coupling quantitatively on the mTMV. Figure 4C shows the deconvoluted ESI mass spectra of wTMV-Alk at 17,662 Da-which is a 128 Da increase from wTMV mass from the attached azo-alkyne, although very small wTMV peak can still be observed on the spectrum. A preceding peak can be also observed (17,645 Da), which has been previously reported 6,9 . For mTMV-Alk, a complete conjugation is observed with the major peak at 17,692 Da, a 30 Da increase from wTMV-Alk. The presence of wTMV in the m TMV sample, generated four peaks in total. Based on these data, we conclude that the mutation does not have an obvious effect on the ability to functionalize Y130 via diazonium coupling. The morphology between wTMV-Alk and mTMV-Alk remains the same (Fig. 4D). These data show that the G155S mutation had no obvious effects on the physical properties of the TMV. ## Conclusion In a span of 2 years, the wild-type TMV has spontaneously mutated in the laboratory environment as observed in the ESI-MS data collected in that period. It should be noted that similar data collected on intact proteins using nominal resolution MALDI-TOF, where the detected ion is the + 1 m/z, would likely have insufficient resolution to detect the presence of single point mutant strains. Further RNA sequencing methods that can detect mutants when they are not the dominant species are complicated and expensive. Bottom-up analysis and peptide sequencing with collision-induced dissociation identified the residue 155 as the site of a glycine to serine point mutation. Dramatic physical effects of this mutation are not predicted but given its surface localization there was the possibility of altered binding and surface chemistry. However, ELISA shows that both mutant and wild-type TMV have identical binding to TMV antibodies and can be equally functionalized at tyrosine 139. It is unclear precisely why the G155S mutant emerged and overtook the wild-type in our laboratory environment. www.nature.com/scientificreports/ Our discovery, however, highlights the necessity to monitor every batch of purified TMV as spontaneous point mutation can occur even in the absence of stress and mutagens. ## Methods Materials. Chemicals were purchased from Sigma-Aldrich (St. Louis, MO), ThermoFisher Scientific (Pittsburgh, PA), Agdia (Elkhart, IN), ChemImpex (Wood Dale, IL), Alfa Aesar (Ward Hill, MA), and TCI America (Portland, OR) and were used without further purification. Instrumentation. LC-MS were obtained using an Agilent 1100 HPLC with a PLRP-S column for separation and an AB Sciex 4000 QTRAP system for detection; also, a Waters SYNAPT G2-Si Q-TOF with an M-class UPLC. TEM was conducted using a JEOL JEM-1400Plus transmission electron microscope. Bio-rad ChemiDoc MP was used for gel imaging. ELISA data were obtained using a BioTek Synergy H4 Hybrid microplate reader. ## Expression of TMV. The tobacco (Nicotiana benthamiana) plants were grown for 8 weeks and infected with TMV solution with 2 weeks of incubation. The harvested infected leaves were stored at − 80 °C. About 100 g of leaves were blended with cold (4 °C) potassium phosphate (KP) buffer (0.1 M, 1000 mL, pH 7.4) and 2-mercaptoethanol (0.2% (v/v)) was added. It was subsequently ground to have an effective extraction. The slurry was filtered and centrifuged at 11,000 × g (4 °C, 20 min). The resulting supernatant was filtered again. The volume was measured and an equal amount of chloroform/1-butanol with 1-to-1 ratio was mixed (4 °C, 30 min). Another centrifugation was done at 4500 × g for 10 min. The collected aqueous phase was mixed with NaCl (final concentration of 0.2 M), PEG 8000 (8% (w/w)), and Triton X-100 surfactant (1% (w/w)). It was stirred on ice for 30 min followed by storing for 1 h at 4 °C. It was centrifuged again at 22,000 × g (4 °C, 15 min). The pellet was collected and resuspended in KP buffer (0.1 M, pH 7.4) and was stored at 4 °C overnight. The suspension was added carefully to 40% (w/v) sucrose gradient in KP buffer (0.01 M, pH 7.4) and centrifuged using swing bucket rotor at 96,000 × g for 2 h. The blue band was collected with the assistance of LED light shined upward from the bottom of the centrifuge tube. The colloidal suspension from blue band was centrifuged at 360,562 × g for 1.5 h. The pellet was resuspended in KP buffer (0.01 M, pH 7.4). ## Transmission electron microscopy (TEM). TEM imaging was performed on a JEOL JEM-1400 + transmission electron microscope. Samples were prepared by incubating 5 µL of ~ 0.1 mg/mL TMV in water on a 300 mesh formvar-coated copper grid for 30 s. The sample was then stained with 5 μL 2% uranyl acetate for an additional 30 s. The excess liquid was wicked away with a Whatman (#1) filter paper and the grids were left to air dry. Images were taken with an accelerating voltage of 120 kV. ## ESI-MS. TMV samples were prepared by denaturing 20 µL of 10 mg/mL in 40 µL of glacial acetic acid. Sample was then centrifuged at 4300 × g for 10 s to separate the precipitated RNA. The supernatant was collected and run on an Agilent 1100 series HPLC system with a PLRP-S column for separation followed by a 4000 QTRAP mass spectrometer. The flow rate is 0.250 mL/min, and the solvent system comprises of Milli Q water, 0.1% formic acid, and pH 7.0 in a 0.1 M sodium phosphate buffer. This system was used for running both TMV and TMV-Alk samples. Electrophoretic mobility assays. 1% (w/v) Agarose gels were used. The sample was prepared by mixing 3 µg TMV with 5 µL Thermo Scientific 6× DNA Loading Dye. From that mixture, 4 µL was added to each well. The gel was run at 100 eV for 45 min, stained with coomassie brilliant blue, and visualized using Bio-rad ChemiDoc MP gel imager. 10% SDS-PAGE gel was used. The sample was prepared by mixing 3 µg TMV with 5 µL of SDS loading dye (β-Mercaptoethanol (5%), Bromophenol blue (0.02%), Glycerol (30%), SDS (Sodium dodecyl sulfate 10%), Tris-Cl (250 mM, pH 6.8)) and 5 µL of 0.1 M dithiothreitol. The mixture was boiled for 10 min. 4 µL of sample was added to each well and the gel was run at 100 eV for 45 min, stained with coomassie brilliant blue, and visualized using Bio-rad ChemiDoc MP gel imager. Bottom-up proteomics. Digestion. Trypsin digestion was performed according to manufacturer's protocol (Thermo Scientific, Product No. 90057). Briefly, TMV samples were dialyzed into a solution of 8 M urea and 50 mM triethylammonium bicarbonate (TEAB) at pH 8.5. 500 mM DTT was added to the sample to a final concentration of 20 mM and incubated for 1 h at 60 °C. 1 M Iodoacetamide was prepared and added to the sample to a final concentration of 40 mM and incubated for 30 min at room temperature in darkness. The alkylation reaction was quenched by adding 500 mM DTT solution to a final concentration of 10 mM. The samples were then diluted to reduce the concentration of urea to 1 M by adding 50 mM TEAB, pH 8.5. Trypsin in 50 mM TEAB pH 8.5 was added to the samples to a final protease-to-protein ratio of 1:20 (w/w) and incubated for 24 h at 37 °C. Samples were then flash-frozen and stored until injection. Samples for pepsin digestion were diluted to 1 µM in TEAB pH 8.5. Samples were then mixed 1:1 (v/v) with a solution of 1.6 M GuHCl, 0.8% formic acid at pH 2.3 to prepare them for on-line pepsin digestion. Samples were then flash-frozen and stored until injection. LC-MS. LC-MS (data presented in Fig. 2) was performed using a Waters HDX manager and SYNAPT G2-Si Q-TOF. Two technical replicates of each sample were analyzed. Pepsin digest samples were digested on-line using Sus scrofa pepsin A (Waters Enzymate BEH) at 15 °C. Peptides were desalted on a C18 pre-column (ACQUITY UPLC BEH C18 VanGuard Pre-column) for 3 min at 100 μl min − www.nature.com/scientificreports/ was 0.1% formic acid. Peptides were separated over a C18 column (Waters Acquity UPLC BEH) and eluted with a linear 3-40% (v/v) acetonitrile gradient for 7 min at 40 μl min −1 and 1 °C. Mass-spectrometry data were acquired using positive-ion mode in HDMS E mode, collecting both low-energy (6 V) and high-energy (ramping 22-44 V) peptide-fragmentation data for peptide identification. All samples were acquired in resolution mode. Capillary voltage was set to 2.8 kV for the sample sprayer. Desolvation gas was set to 650 l h −1 at 175 °C. The source temperature was set to 80 °C. Cone and nebulizer gas was flowed at 90 l h −1 and 6.5 bar, respectively. The sampling cone and source offset were both set to 30 V. Data were acquired at a scan time of 0.4 s with a m/z range of 100-2000. Mass correction was done using [Glu1]-fibrinopeptide B as a reference mass. Data processing. Raw data were processed by PLGS (Waters Protein Lynx Global Server 3.0.2) using a database containing Sus scrofa pepsin A and native TMV coat protein. In PLGS, the minimum fragment ion matches per peptide was 3, and methionine oxidation and N-terminal acetylation were allowed. Trypsin samples included cysteine carbamidomethylation (CAM) as a fixed modifier. The low and elevated energy thresholds were 250 and 50 counts, respectively, and the overall intensity threshold was 750 counts. Peptides were curated in DynamX 3.0 with thresholds of 0.3 products per amino acid and one consecutive product. ELISA. 96-well Microtiter Plates High Bind-Solid (ACC 00948/0005 Agdia) were coated with 100 µl of diluted (1:200) capture antibody rabbit anti-TMV IgG (CAB 57400/1000 TMV Capture antibody Agdia) in coating buffer (0.015 M Na 2 CO 3 , 0.034 M NaHCO 3 , NaN 3 in dH 2 O, pH 9.6) and incubated overnight at 4 °C. The plate was washed 3 × with washing buffer (0.2% (v/v) Tween-20 in PBS, pH 7.4). After washing, wells were blocked with blocking buffer (1% (w/v) BSA in washing buffer, pH 7.4) at RT for 1 h, followed by 4 × washes with washing buffer. 100 μL of native TMV and mutant TMV-concentrations determined by Lowry assay (0.5-0.015 µg/ ml)-were serially diluted with sample extraction buffer (0.009 M Na 2 SO 4 , 2% (v/v) Polyvinylpyrrolidone (PVP) 40 k, 0.2% (w/v) Powdered egg (chicken) albumin, 0.003 M NaN 3 , 0.2% (v/v) Tween-20 in washing buffer) was then added to each well and incubated at RT for 2 h. Wells were then washed 8 × with washing buffer, followed by the addition of 100 μL alkaline phosphatase-conjugated rabbit anti-TMV IgG in conjugate buffer (0.5 mg/ mL BSA, 2% (v/v) PVP 40 k, and 0.003 M NaN 3 in washing buffer) and incubated at RT for 2 h. Wells were then washed 8 × with washing buffer, then developed by adding 100 μL one-step p-nitrophenylphosphate (PNPP) substrate for 45 min at RT. The plate was read at 405 nm, 420 nm, and 450 nm, and the absorbance values of buffer blank wells averaged and subtracted from the entire plate. Experiments were performed in triplicate. TMV functionalization. The diazonium salt is prepared by carefully mixing 200 μL of 0.30 M p-toluenesulfonic acid monohydrate, 75 μL of 0.68 M 3-ethynylaniline, and 25 μL 3.0 M sodium nitrite followed by incubation in ice for 1 h without light exposure. The resulting diazonium salt (50 μL) was added to a 2 mg/mL wTMV or mTMV solution in 0.1 M borate buffer at pH 8.8. This was incubated on ice for 45 min. The resulting TMV-alkyne was purified and concentrated via centrifuge filtration using an EMD Millipore Amicon Ultra Centrifugal Filter Unit (10,000 MW Cutoff) (4303×g) (Suppl. Information). ## Regulatory and compliance. The authors comply with the IUCN Policy Statement on Research Involving Species at Risk of Extinction and the Convention on the Trade in Endangered Species of Wild Fauna and Flora. This plant and plant use studies were approved by the University of Texas at Dallas Institutional Biosafety and Chemical Safety Committee. Seeds and plants used are not listed as endangered or threatened and were a gift from Prof. James Culver at the University of Maryland.

chemsum

{"title": "Identification and physical characterization of a spontaneous mutation of the tobacco mosaic virus in the laboratory environment", "journal": "Scientific Reports - Nature"}

the_effect_of_iron_binding_on_uranyl(<scp>v</scp>)_stability

## Abstract: Here we report the effect of UO 2 + /Fe 2+ cation-cation interactions on the redox properties of uranyl(V) complexes and on their stability with respect to proton induced disproportionation. The tripodal heptadentate Schiff base trensal 3À ligand allowed the synthesis and characterization of the uranyl(VI) complexes [UO 2 (trensal)K], 1 and [UO 2 (Htrensal)], 2 and of uranyl(V) complexes presenting UO 2, [UO 2 (trensal)Fe(py) 3 ], 6). The uranyl(V) complexes show similar stability in pyridine solution, but the presence of Fe 2+ bound to the uranyl(V) oxygen leads to increased stability with respect to proton induced disproportionation through the formation of a stable Fe 2+ -UO 2 + -U 4+ intermediate ([UO 2 (trensal)Fe(py) 3 U(trensal)]I, 7) upon addition of 2 eq. of PyHCl to 6. The addition of 2 eq. of PyHCl to 3 results in the immediate formation of U(IV) and UO 2 2+ compounds. The presence of an additional UO 2 + bound Fe 2+ in [(UO 2 (trensal)Fe(py) 3 ) 2 Fe(py) 3 ]I 2 , 8, does not lead to increased stability. Redox reactivity and cyclic voltammetry studies also show an increased range of stability of the uranyl(V) species in the presence of Fe 2+ with respect both to oxidation and reduction reactions, while the presence of a proton in complex 2 results in a smaller stability range for the uranyl(V) species. Cyclic voltammetry studies also show that the presence of a Fe 2+ cation bound through one trensal 3À arm in the trinuclear complex [{UO 2 (trensal)} 2 Fe], 5 does not lead to increased redox stability of the uranyl(V) showing the important role of UO 2 + /Fe 2+ cation-cation interactions in increasing the stability of uranyl(V). These results provide an important insight into the role that iron binding may play in stabilizing uranyl(V) compounds in the environmental mineral-mediated reduction of uranium(VI). ## Introduction Uranyl(V) 1 has been proposed as an important transient intermediate in the biological or abiotic mineral-mediated transformation of soluble uranyl(VI) compounds into the insoluble uranium(IV) dioxide (UO 2 ). These processes provide a convenient strategy to sequester uranium in the environment and, as such, are very important for ground-water remediation. In particular, stable adsorbed or incorporated uranyl(V) species have been reported to form during the U(VI) reduction by Fe(II)bearing minerals such as mica 1e or magnetite ([Fe 2+ (Fe 3+ ) 2 O 4 ]) 2,3 and the presence of iron as the second nearest neighbour has been identifed. 4 UO 2 + species have low stability in aqueous media and they quickly disproportionate to uranyl(VI) and U(IV), 5 but the incorporation into iron minerals may prevent disproportionation or further reduction of U(V) to U(IV) and thus lead to long-term immobilization of U(V). However, the role of iron binding to uranyl(V) species and their stabilization remains ambiguous in spite of its importance for the correct speciation of uranium in the environment. Dinuclear or polynuclear complexes of uranyl(V) built from the interaction of a uranyl(V) oxo group with the uranium centre from a UO 2 + moiety (UO 2 + /UO 2 + ), also known as cation-cation interaction (CCI), 6 have been proposed as intermediates in the proton promoted disproportionation of uranyl(V) to afford UO 2 2+ and U(IV) species. 1c,7 The subsequent addition of protons to these polynuclear uranyl(V) intermediates leads to complete electron transfer followed by dissociation of the resulting U(VI)/U(IV) complex. In aprotic media stable polynuclear UO 2 + /UO 2 + complexes have been isolated. 8 We showed that the addition of protons (PyHCl) to a pyridine solution of stable tetrameric UO 2 + /UO 2 + complexes leads to the immediate disproportionation of the uranyl(V) species affording uranyl(VI) and U(IV) complexes and water. 8b Disproportionation of polynuclear cation-cation complexes was also observed in the absence of protons upon addition of strong Lewis acids (Li + or U 4+ ) 8b,9 to stable uranyl(V) Schiff base complexes and was found to lead to complex mixtures of soluble mixed-valent U(IV)/U(V) uranium oxo clusters. It was also reported that the binding of strong Lewis acids or Group 1 metals to the uranyl(VI) oxo group renders more favourable the reduction of U(VI) to U(V). 10 Moreover, it has been demonstrated that the binding of strong Lewis acids such as B(C 6 F 5 ) 3 to the uranyl(V) oxo groups renders more accessible the reduction of U(V) to U(IV). 10d,11,12 A fewer studies have been directed to investigate the effect of the interaction of uranyl(V) with 3d transition metals on the stability and redox reactivity of uranyl(V) species. Moreover, in spite of the fact that several uranyl(V) complexes stable in organic solution have been isolated in recent years, 9,13 only a few examples of heteropolymetallic complexes presenting a UO 2 + / M interaction, where M is a 3d transition metal, have been prepared. 14 The few reported UO 2 + /3d complexes have shown interesting single-molecule magnetic properties. 14 The addition of FeI 2 to an unstable putative uranyl(V) dipotassium complex of a macrocyclic Schiff base Pacman ligand was reported to result in a higher stability of the uranyl(V) Pacman complexes which was corroborated by the isolation of the corresponding heterobimetallic UO 2 + /Fe 2+ CC complex. However, the effect of the interaction UO 2 + /Fe 2+ on the stability of these uranyl(V) complexes was not further investigated. 14c Here we report two new stable complexes of uranyl(V) supported by the tripodal Schiff base ligand H 3 trensal (2,2 0 ,2 00 -tris(salicylideneimino)triethylamine): the UO 2 + /K + [UO 2 (trensal)K]K, 3, and the heterobimetallic UO 2 + /Fe 2+ complex [UO 2 (trensal)Fe(Py) 3 ], 6. The reactivity of these complexes toward protons and their redox properties were compared and these studies unambiguously show the increased stability of the iron bound complexes. ## Results and discussion Uranyl(VI) and uranyl(V) complexes of trensal 3 The reaction of K 3 trensal with the nitrate salt of uranyl (VI) leads to the isolation of the uranyl(VI) complex [UO 2 (trensal)K], 1 in 59% yield. The broad 1 H NMR spectrum of 1 in pyridine suggests the presence of fluxional solution species. A higher resolution of the 1 H NMR spectrum is observed in deuterated THF and a well resolved 1 H NMR spectrum could be obtained in CD 3 OD solution (Fig. S2 †). X-ray quality crystals of 1 could not be obtained, but the addition of one equivalent of PyHCl to a pyridine solution of 1 led to the isolation of X-ray quality crystals of the neutral complex [UO 2 (Htrensal)], 2, in 60% yield. The 1 H NMR spectrum of 2 in pyridine shows the presence of 15 overlapping narrow signals in agreement with the presence of C 2 symmetric solution species (Fig. S3 †). The X-ray crystal structure of this complex is presented in Fig. 1 and shows that the uranium atom is heptacoordinated, with a slightly distorted pentagonal bipyramidal coordination geometry, by two uranyl oxygen atoms in the axial position and fve donor atoms of the trensal 3 ligand in the equatorial plane. The third protonated arm of the trensal 3 ligand is not coordinated to the uranyl cation and the phenol proton is hydrogenbonded with the Schiff base nitrogen N4. The values of the U(VI)]O bond lengths lie in the range of those typically observed for uranyl(VI) complexes (U-O3 ¼ 1.783(3) and U-O4 ¼ 1.787(3) ). 8a,d,9,15 The average U-O phenoxide (2.231 ) and the average U-N imine (2.612 bond) lengths are also in the range of those found in other reported Schiff base complexes of uranyl(VI). 8a,d,9,15 In the attempt to reduce the uranyl(VI) complex 2 we added 1 eq. of decamethyl cobaltocene (Cp* 2 Co) to pyridine solutions of 2. The 1 H NMR spectrum of the resulting reaction mixture immediately after addition shows the presence of a large number of signals in the 45 to 45 ppm range suggesting that a putative uranyl(V) intermediate complex undergoes rapid disproportionation (Fig. S4 †). This suggests that the phenol arm protonates the more basic uranyl(V) (compared to uranyl(VI)) oxo group resulting in proton induced disproportionation. In contrast, the uranyl(V) complex [UO 2 (trensal)K]K, 3, is conveniently prepared in 70% yield from the salt metathesis reaction between K 3 trensal and [(UO 2 Py 5 ) (KI 2 Py 2 )] n in pyridine (Scheme 1). The 1 H NMR spectrum of 3 in deuterated pyridine showed the presence of fluxional species with signals in the paramagnetic region (11 to 15 ppm) characteristic of U(V). Cooling down or heating up the NMR sample did not lead to a better resolution of the spectrum (Fig. S5 †). The addition of stoichiometric amounts of 2.2.2 cryptand (4,7,13,16,21,24-hexaoxa-1,10-diazabicyclo[8.8.8]hexacosane) to complex 3 resulted in a well resolved 1 H NMR spectrum (Fig. S6 †). This suggests that fluxional potassium binding to the uranyl oxygen is the cause of the broad features in the 1 H NMR spectrum of 3. The complex [UO 2 (trensal)] [K(2.2.2crypt)][K(2.2.2crypt)], 4, was obtained analytically pure as a green solid in 62% yield. The solid-state structure of 4 was determined by X-ray diffraction studies and is presented in Fig. 2. The overall quality of the crystal structure of compound 4 is rather poor (very weakly diffracting sample) but its connectivity is well determined. The coordination environment around the uranium centre is similar to that found in complex 2. In 4 the uranium atom is heptacoordinated in a pentagonal bipyramidal coordination geometry. Five donor atoms of the trensal ligand (two oxygen and three nitrogen atoms) occupy the equatorial plane of the uranium ion, while the third arm of the trensal 3 ligand does not interact with any cation and one [K(cryptand)] cation is found as an isolated ion in the unit cell of 4. The bipyramid axial positions in 4 are occupied by two oxo ligands with U-O distances (1.824(15) and 1.865( 16) ) signifcantly longer than those found in the uranyl(VI) complex 2 (1.785(3) ). These distances are in the range of those found in previously reported complexes of uranyl(V). 13a,e,13f The 1 H NMR spectrum of 4 in pyridine shows the presence of 12 narrow signals in agreement with the presence of C 2 symmetric solution species (Fig. S6 †). This indicates that the molecular anionic fragment [UO 2 (trensal)] [K(2.2.2crypt)] found in the X-ray structure of 4 dissociates in pyridine solution and the [K(2.2.2crypt)] + cation is not bound to the uranyl(V) oxo group in pyridine solution. The solid-state X-band EPR spectra measured at 298 K and 10 K revealed that the complex 3 is EPR silent. In contrast, the solid-state X-band (9.40 GHz) EPR spectrum of 4 shows an intense signal at 10 K with a ftted rhombic set of g-values (g 1 ¼ 2.44; g 2 ¼ 1.10; g 3 < 0.6), confrming the presence of uranium in the oxidation state +5 (Fig. S29 †). Notably, encapsulation of potassium enables us to obtain an EPR signal from the otherwise EPR silent complex 3. The likely presence of two potassium binding both uranyl(V) oxo groups results in a different electronic structure of 3 compared to 4 (where only one potassium cation is bound) which results in the absence of the EPR signal. 13e Complex 4 is stable up to one month in the solid state and in pyridine and THF solutions. In order to assess the stability of these uranyl(V) complexes with respect to proton induced disproportionation, we have investigated the reaction of 3 and 4 with protons. After addition of 1 eq. of PyHCl to complex 3, partial disproportionation of the uranyl(V) complex was observed by 1 H NMR spectroscopy. The addition of 2 eq. of PyHCl resulted in the complete disproportionation of the uranyl(V) to afford the uranyl(VI) complex 2 and unidentifed U(IV) products as indicated by 1 H NMR spectroscopy (Fig. S7 †) and single crystal X-ray diffraction of the isolated crystal of the complex 2. The U(IV) compounds formed in the disproportionation were identifed as the product of the hydrolysis of the [U(trensal)]Cl complex as confrmed by the 1 H NMR spectrum of a 1 : 1 : 2 mixture of [UO 2 (Htrensal)], [U(trensal)]X (X ¼ I, Cl) and H 2 O (Fig. S8 †). On the other hand, addition of PyHCl to 4 initially resulted in the formation of NMR silent species, but after 3 days the 1 H NMR spectrum shows the formation of the same disproportionation products as those found in the reaction of 3 with 2 eq. of PyHCl (Fig. S9 †). ## Iron binding to uranyl(V) complexes In view of the potential important role of iron binding in the abiotic reduction of uranyl(VI) as well as in the stabilization of uranyl(V) at iron mineral surfaces we have investigated the reactivity of complexes 1 and 3 with iron salts. The reaction of 1 with FeI 2 affords the trinuclear complex [{UO 2 (trensal)} 2 Fe], 5, in 93% yield according to Scheme 2. The solid state structure of 5 (Fig. 3) shows the presence of a neutral trinuclear complex where two [UO 2 (trensal)] moieties are held together by a Fe(II) cation bound by two trensal O, N donor atoms not involved in the coordination of the uranyl cation. Thus, the replacement of the potassium cation in 1 with a Fe(II) cation leads to formation of a trinuclear structure. In order to prepare a trinuclear uranyl(V) analogue we allowed 5 to react with Cp* 2 Co. The 1 H NMR spectrum after addition of 1 eq. of Cp* 2 Co to complex 5 revealed the formation of a complex reaction mixture. One of the products could be identifed by X-ray diffraction studies, revealing the formation of the dinuclear heterobimetallic complex [UO 2 (trensal)Fe(py) 3 ], 6. Addition of 2.5 eq. of Cp* 2 Co to a pyridine solution of 5 led to an intractable reaction mixture from which none of the components could be identifed (Fig. S11c †). Complex 6 can be conveniently prepared in 81% yield from the reaction of FeI 2 with complex 3 in pyridine in a 1 : 1 ratio (Scheme 3). The solid-state structure of 6, represented in Fig. 4, shows the presence of a neutral dinuclear complex where a [U V O 2 (trensal)] dianion binds a Fe 2+ cation through a UO 2 + /Fe 2+ CCI. cation (1.946(4) -2.132(4) ). 14c,f Similar Fe-O bond lengths ranging from 1.935(4) to 2.058(4) were reported for uranyl(VI) complexes bridged to Fe(III) via a hydroxo group. 16 The 1 H NMR spectrum of 6 in pyridine shows the presence of 12 signals over a broad range of chemical shifts (30 to + 51 ppm). The large shift of the 1 H NMR signals observed for 6 compared to complex 4 indicates that the UO 2 + /Fe 2+ CCI is present in pyridine solution (Fig. S12 †). The ESI/MS spectrum ({UO 2 (trensal)Fe(Py) + }: m/z ¼ 859.83) of 6 also indicates the presence of the heterobimetallic complex in pyridine solution (Fig. S26 †). The stability and reactivity of 6 were then investigated and compared with those found for 3 and 4 in order to elucidate the effect of the Fe 2+ ion. The addition of 1 eq. of PyHCl to a solution of 6 in pyridine results in the partial disproportionation of the uranyl(V) complex (Scheme 4) with a 2 : 1 ratio of 6 to the disproportionation product [UO 2 (trensal)Fe(py) 3 U(trensal)]Cl 7b (Fig. S20 †). The addition of 2 equivalents of PyHCl to 6 led to the complete disappearance of the signals of complex 6 in the 1 H NMR spectrum (Fig. S13 †) and to an increased intensity of the signals assigned to 7b. The presence of the uranyl(VI) complex [UO 2 (Htrensal)] as the second disproportionation product was also identifed by 1 H NMR spectroscopy. However, in both cases the disproportionation was not complete. Notably, the trinuclear cation-cation complex 7b contains unreacted uranyl(V) (Scheme 4). The formation of the [UO 2 (Htrensal)] by-product prevented the synthesis of 7b from the reaction of 6 with PyHCl (Scheme 4). However, the iodide analogue [UO 2 (trensal)Fe(py) 3 -U(trensal)]I, 7 was prepared in 80% yield from the reaction of complex 6 with 1 eq. of the [U(trensal)]I complex in pyridine (Scheme 4). This complex is a rare example of an actinidefunctionalized uranyl complex and only the third example of a uranyl(V) complex presenting a CCI between the uranyl(V) oxo group and a U(IV) cation. 9,13l The structure of complex 7 (Fig. 5) shows the presence of a cationic trinuclear complex built via CCI between the U(IV) center from the [U(trensal)] + complex and the oxo group of the uranyl(V) [UO 2 (trensal)Fe(py) 3 ] fragment. The three metal ions adopt a close to linear arrangement with a Fe-O-U angle of 170.3(3) and a U-O-U angle of 171.2(3) . The U]O bond distance for the uranyl(V) oxo group bound to the Fe 2+ remains unchanged at 1.922(6) compared to complex 6, but a signifcant lengthening of the U]O bond is observed upon binding of the U(IV) cation in 7 (1.960(6) ). The UO 2 + /U(IV) distance (2.317( 6) ) is comparable to those found in the only two other complexes reported to have a UO 2 + /U(IV) CCI (2.198(13) and 2.245(3) ). 9,13l The UO 2 + /Fe 2+ distance (2.144( 6) ) is slightly longer than in 6 but is in the range of those found in the two previously reported complexes presenting a UO 2 + /Fe 2+ interaction (1.946(4) -2.132(4) ). 14c,f These results indicate that the presence of Fe 2+ increases the stability of uranyl(V) in 6 with respect to proton induced disproportionation. Notably the addition of 2 eq. of PyHCl led to full disproportionation of the complexes 3 and 4 while it resulted only in the partial disproportionation of 6 and the formation of [UO 2 (Htrensal)] and of the Fe-U(V)-U(IV) trimer. The addition of fve equivalents of pyridinium chloride is required for the full disproportionation of complex 6 to occur affording the same uranyl(VI) and U(IV) diproportionation products as observed after addition of acid to 3. This indicates that the iron bound uranyl(V) complex 6 displays an increased stability towards the proton induced disproportionation compared to the potassium bound uranyl(V) complexes 3 and 4 (Fig. S14 †). The binding of U(IV) to the uranyl(V) oxo group was previously reported to promote partial disproportionation and formation of multimetallic U(IV)-U(V) oxo-bridged complexes. 9 The stabilizing effect of Fe 2+ compared to U 4+ can be explained in terms of the lower Lewis acidity of Fe 2+ compared to U 4+ . This results in the formation of stable UO 2 + /Fe 2+ adducts where the uranyl oxo group becomes less accessible to protonation by the Brønsted acid H + . In contrast, complex 7 is stable in pyridine solution over one month period. 1 H NMR studies show that the addition of [U(trensal)]I to complex 3 also leads to the formation of a stable unidentifed compound (Fig. S15 †). The subsequent addition of FeI 2 to this compound led to the formation of complex 7. These results suggest that stable U(IV)-U(V) complexes also form in the absence of iron bound to uranyl(V) oxo group. However, the formation of these compounds is not observed during the addition of PyHCl to 3, which undergoes complete disproportionation after the addition of 2 eq. of PyHCl. Moreover, the addition of 2 eq. of PyHCl to the U(IV)-U(V) adduct results in full disproportionation, as indicated by the 1 H NMR spectrum, suggesting that the binding of U(IV) to the uranyl(V) oxo does not lead to increased stability (Fig. S15 †). This further confrms the stabilizing role of Fe(II) binding with respect to proton induced disproportionation of uranyl(V). The 1 H NMR spectrum of 7 in pyridine shows the presence of 45 signals over a large range of chemical shifts (35 to + 53 ppm) in agreement with the presence of the trimeric complex 7 in solution. (Fig. S16 †). Additionally, the ESI/MS spectrum of 7 in pyridine solution {(UO 2 (trensal)Fe 3 U(trensal) + } m/z ¼ 1474.42) indicated that the complex 7 retains its trinuclear structure in the pyridine solution (Fig. S27 †). The addition of 1 eq. of pyridinium chloride to 7 results in partial disproportionation with a 3 : 1 ratio of complex 7 to the disproportionation products as shown by 1 H NMR spectroscopy (Fig. S17 †). The complete disproportionation of complex 7 requires the addition of 4 eq. of PyHCl. The coordination of U(IV) does not increase the stability of the uranyl(V) species in 7 with respect to 6. In view of the increased stability of 7 and 6 compared to 3 towards proton induced disproportionation, we set out to investigate how the coordination of a second Fe 2+ cation to complex 6 would affect the structure and reactivity of the U(V) centre. The 1 H NMR of the reaction mixture resulting from the addition of 0.5 equivalents of iron(II) iodide to 6 in pyridine indicated the formation of a new species (Fig. S18 †). X-ray quality crystals of [(UO 2 (trensal)Fe(py) 3 ) 2 Fe(py) 3 ]I 2 , 8, were obtained in 65% yield from this reaction (Scheme 5). The solidstate structure (Fig. 6) of 8 shows the presence of a pentametallic structure where a Fe(Py) 3 moiety bridges two iron-bound uranyl(V) [UO 2 (trensal)Fe(py) 3 ] moieties. Overall, this results in the presence of UO 2 + /Fe 2+ CCIs for both uranyl(V) oxo-groups. The central Fe 2+ cation is penta-coordinated by one oxo atom from each of the two uranyl(V) groups and three pyridine molecules. The mean Fe(2)-O(oxo) bond lengths is 1.988 . The [UO 2 (trensal)Fe(py) 3 ] moieties of the crystal structure possess the same geometry found in the mononuclear complex 6, but the additional uranyl-iron interaction results in a slight lengthening of the UO 2 + /Fe 2+ bonds compared to 6 (2.061(4) The labile binding of the central Fe(Py) 3 2+ cation in 8 does not lead to an increased stability of 8 towards proton induced disproportionation compared to 6. Notably, the 1 H NMR indicated a 2 : 1 ratio between the starting complex 8 and the disproportionation products upon addition of 1 eq. of H + per uranyl(V) which is identical to the ratio observed for the complex 6 (Fig. S20 †). ## Redox reactivity Iron binding to the uranyl(V) oxo is anticipated to have an important effect on its redox reactivity. Moreover, it has been suggested that iron binding at mica surfaces leads to the stabilization of uranyl(V) intermediates but the effect of iron binding on the redox properties of isolated uranyl(V) complexes has not been investigated. At frst, we explored the chemical oxidation of uranyl(V) by Fe 3+ . The reaction of 3 with 1 eq. FeCl 3 leads to the oxidation of the uranium center (Fig. S21 †) and to the formation of the uranyl(VI)-Fe(II) complex 2 [(UO 2 (trensal)) 2 Fe] as identifed by Xray diffraction crystallography and 1 H NMR spectroscopy. The oxidation of uranyl(V) complex to uranyl(VI) by Fe(III) is explained in terms of the respective redox potential (Fe(III)/Fe(II) ¼ 0.0 V; UO 2 2+ /UO 2 + ¼ 1.6 vs. V (Fc/Fc + )). In order to probe the possibility of obtaining a uranyl(V)-Fe(III) complex we explored the reactivity of 3 and 6 with increasingly electron-rich FeLCl n complexes (L ¼ tpa and tdmba); tpa ¼ (tris(pyridin-2-ylmethyl)amine) and H 3 tdmba ¼ (tris-(2-hydroxy-3,5-dimethylbenzyl)amine). The reaction of 6 with [Fe(tpa)Cl 3 ] led to the oxidation of uranyl(V) to uranyl(VI) with concomitant formation of [Fe(tpa)Cl 2 ] (as shown by X-ray diffraction studies and 1 H NMR spectroscopy). The reaction of 6 with the neutral Fe(III) complex [Fe(tdmba)] did not result in any change observable in the 1 H NMR spectrum of 6 (Fig. S23 †) indicating that the Fe(III) cation in [Fe(tdmba)] does not form CCIs with the uranyl(V) oxo group but does not oxidize the uranyl(V) either. In contrast, when complex 3 is reacted with [Fe(tdmba)], 1 H NMR spectroscopy indicated that a redox reaction occurs yielding uranyl(VI) and Fe(II) species (Fig. S24 †). These results are in agreement with the reported influence of chelating agents on the reoxidation by Fe(III) of biogenic products of uranyl(VI) reduction. 17 These results suggest that the presence of UO 2 + /Fe 2+ CCI stabilizes the uranyl(V) oxidation state with respect to the oxidation. In order to further probe the effect of iron binding on the redox properties of uranyl(V) species we performed comparative cyclic voltammetry studies of complexes 1, 2, 5, 6 and 8 (Fig. 6 and SCV1-SCV4 †). The voltammogram of 1 in pyridine (Fig SCV1 †) shows an irreversible redox event at 1.75 V, but when the voltammogram of 1 is measured in the presence of the cryptand a reversible redox event assigned to the U(VI)/U(V) couple is observed at E 1/2 ¼ 1.69 V vs. Fc/Fc + (Fig. 7, green curve). The voltammogram of the protonated uranyl(V) complex 2 also shows the presence of a reversible redox event at E 1/2 ¼ 1.66 V vs. Fc/Fc + assigned to the U(VI)/U(V) couple. These values compare well with the values previously measured in pyridine for other uranyl(V) complexes of tetradentate (E 1/2 ¼ 1.61 V or 1.67 V vs. Fc/Fc + ) 15a,13e and pentadentate Schiff bases (E 1/2 ¼ 1.58 V vs. Fc/Fc + ). 18 A second irreversible redox event is observed at E 1/2 ¼ 2.47 V vs. Fc/Fc + for complex 2, but not for complex 1. This event is consistent with the reduction of the metal centre (values of redox potential ranging from 2.02 to 2.88 V vs. Fc/Fc + were previously assigned to the U(V)/U(IV) couple 12a ). The possibility that this event could be related to the reduction of the Schiff base ligand is unlikely since this feature is absent from the voltammograms of the H 3 trensal, K 3 trensal ligands and of the complex 1. Moreover, the shift of the U(V)/U(IV) couple to a more positive potential in complex 2 could be explained by the presence of a proton on the complex. Similar redox events are observed in the voltammogram of complex 6 in addition to the quasi-reversible wave at E 1/2 ¼ 0.0 V vs. Fc/Fc + , assigned to the Fe(III)/Fe(II) couple. However, the U(VI)/U(V) reduction process is found at 1.03 V vs. Fc/Fc + in the voltammogram of 6 and the second reduction event occurs at E 1/2 ¼ 2.7 V vs. Fc/Fc + demonstrating that the range of stability of the uranyl(V) species is signifcantly extended compared to complex 2 as a result of Fe(II) binding. Both reduction and oxidation of the uranyl(V) cation are more difficult in the presence of Fe(II). No additional redox stabilisation was observed upon addition of two or more equivalents of Fe(II) to complex 6 as indicated by the voltammogram of complex 8 (Fig. SCV4 †). This is probably due to the labile binding of the second Fe(II) cation to the uranyl(V) oxo group in pyridine. Moreover, in the voltammogram of complex 5 (Fig. SCV2 †) the redox event assigned to the U(VI)/U(V) couple is found at E 1/2 ¼ 1.66 V vs. Fc/Fc + as in complexes 1 and 2 in spite of the presence of a Fe(II) ion bound through the Schiff base acting as a bridging ligand. These results indicate that cation-cation interaction between the uranyl(V) oxygen and the Fe 2+ is essential for the stabilization of U(V) while the presence of a Fe(II) bound through the ligand has no signifcant effect on the redox properties of uranyl(V).

chemsum

{"title": "The effect of iron binding on uranyl(<scp>v</scp>) stability", "journal": "Royal Society of Chemistry (RSC)"}

a_high_throughput_ambient_mass_spectrometric_approach_to_species_identification_and_classification_f

## Abstract: A high throughput method for species identification and classification through chemometric processing of direct analysis in real time (DART) mass spectrometry-derived fingerprint signatures has been developed. The method entails introduction of samples to the open air space between the DART ion source and the mass spectrometer inlet, with the entire observed mass spectral fingerprint subjected to unsupervised hierarchical clustering processing. A range of both polar and non-polar chemotypes are instantaneously detected. The result is identification and species level classification based on the entire DART-MS spectrum. Here, we illustrate how the method can be used to: (1) distinguish between endangered woods regulated by the Convention for the International Trade of Endangered Flora and Fauna (CITES) treaty; (2) assess the origin and by extension the properties of biodiesel feedstocks; (3) determine insect species from analysis of puparial casings; (4) distinguish between psychoactive plants products; and (5) differentiate between Eucalyptus species. An advantage of the hierarchical clustering approach to processing of the DART-MS derived fingerprint is that it shows both similarities and differences between species based on their chemotypes. Furthermore, full knowledge of the identities of the constituents contained within the small molecule profile of analyzed samples is not required.One of the manifestations of the genetic differences that distinguish one species from another is in the profile of constitutively present small molecules they contain, also known as the metabolome. Since the small-molecule profile of an organism ultimately reflects the genes that distinguish it, the information content of the metabolome might be just as well suited to genomic fingerprinting and assessment of genetic relatedness between species as the genomes themselves. There are several reasons why it would be useful to be able to accurately correlate the signature of small molecules observed within an organism to its overall systems biology. The observation of the composite of small-molecule biomarkers could provide a real-time view of gene expression activity, enable the monitoring of the status of cellular transcriptomes and proteomes, provide a means of assessing the evolutionary history of organisms, and provide an avenue for the rapid monitoring of the success of gene knockouts and knockdowns, among other uses. Although these applications can be accomplished by phylogenetic methods, the paucity of mapped and/ or annotated genes for the vast majority of fauna and flora in existence makes this approach impossible for all but a select group of mostly model systems. Convenient characterization of the defining features of an organism's real-time chemical portrait for the purpose of species classification has been hampered by several factors. These include: (1) the difficulty of acquiring a comprehensive small-molecule chemical map of an organism or its parts in real time; (2) the time-consuming nature of metabolome profiling by conventional methods; (3) the challenge of obtaining a faithful and consistent representation of defining chemical components or chemical component ratios that is divorced from biases or artifacts introduced by sample processing steps; and (4) distinguishing between chemicals that define a species and those that do not provide discriminatory information. However, the advent within the last decade of ambient ionization mass spectrometric methods that feature instantaneous real-time detection of fairly comprehensive small molecule profiles of matter in its native form, has the potential to revolutionize and simplify metabolome-and/ or chemical fingerprint-based species characterization by circumventing to a large extent the aforementioned deficiencies of conventional methods. Additionally, the utilization of the comprehensive mass spectrometry-derived fingerprint, rather than a subset of small-molecule biomarkers, provides the opportunity to subject an entire dataset to multivariate statistical analysis to aid in species classification, as well as processing of the data through hierarchical clustering in order to assess genetic relatedness and distinguish between species. Direct Analysis in Real Time (DART ® ) 1 is one of the most common of the new mass spectrometric "ambient ionization" sources 2 and was the first such source to be introduced as a commercial product. Following an early application note on the use of DART to analyze the flavor components and polyphenols in the leaves of two different basil cultivars 3 , there have been several reports on the application of DART to species biomarker identification. Fatty acid profiles measured by DART for different bacterial species have been shown to be distinct and reproducible 4 , volatiles release from Eucalypts of different species have been shown to be unique 5,6 , differentiation between red oak (Quercus rubra) and white oak (Q. alba) 7 by DART has been demonstrated, identification of printing and writing papers 8 based on chemical profile differences has been shown, and the identification of Piper betel cultivars 9 , ambiguous cubeb fruit 10 and varieties of the psychoactive plant Mitragyna speciosa ("Kratom") 11 have all been demonstrated utilizing ambient DART mass spectrometry. The U.S. Fish and Wildlife Services Forensic Laboratory has also used DART-MS to distinguish between species of Dalbergia 12 and agarwood 12 . Each of the aforementioned reports relied on visual examination of mass spectra or their corresponding heat maps for selection of m/z features that were then used in chemometric-based approaches including the unsupervised learning methods Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS), and the supervised learning methods Linear Discriminant Analysis (LDA) and Kernel Discriminant Analysis (KDA), for distinguishing between species within a single genus. Successful application of these methods requires careful albeit a priori selection of the features (mass spectral peaks) that differentiate between species. By combining mass spectrometric heat maps and chemometric protocols, we illustrate a high throughput method by which DART-MS-derived chemical signature profiles can be subjected to cluster analysis to not only distinguish between species, but provide information on genetic relations. An advantage of the hierarchical clustering approach to the processing of DART-derived fingerprint information is that it shows both similarities and differences between species based on their chemotypes as determined from DART-MS data. In contrast, previously described chemometric methods show only differences between classes, but do not indicate which classes have similar chemotypes. Furthermore, full knowledge of the identities of the constituents contained within the small molecule profile of the sample being analyzed is not required. The method is robust and rapid and the results are consistent. Here, we showcase several applications although these are by no means exhaustive and numerous other possibilities exist. In this work, we demonstrate how the method can be used to: (1) distinguish between endangered woods regulated by the Convention for the International Trade of Endangered Flora and Fauna (CITES) treaty; (2) assess the origin and by extension the properties of biodiesel feedstocks; (3) determine insect species from analysis of puparial casings; (4) identify psychoactive plant products; and (5) differentiate between Eucalyptus species. ## Results Detection of Illegally Traded Endangered Species in the Genus Dalbergia. The convention for the international trade of endangered flora and fauna (CITES) which is enforced under the Endangered Species Act (ESA), bans the trade of tree species whose harvest has been deemed unsustainable. Visual inspection of harvested plant products in which leaves, flowers and other characteristic features have been retained can enable definitive identification of endangered species. However, since timber and sawn boards generally lack diagnostic morphological features, identification of wood products has historically relied on anatomical or chemical features associated with the hardwood. This process is laborious, time-consuming and can be prone to error. The aforementioned challenge is further exacerbated not only by the plethora of colloquial terms used within the timber trade community for any one tree species, but also by the common use of a single name to refer to multiple species. For example, "Dalbergia granadillo Pittier" is a tree species of rosewood endemic to Mexico and northern Central America, whose common name is "granadillo". However, this same name has been used to describe Dalbergia retusa (Leguminosae family), and several Fabaceae family trees including Platymiscium yucatanum, Caesalpinia echinata, and C. platyloba. Furthermore, D. retusa has a large number of synonyms including Amerimnon lineatum, A. retusum, D. cuscatlantica, D. hypoleuca, D. lineata, D. pacifica, D. retusa var. hypoleuca, and D. retusa var. lineata, with many xylarium collections still using historical nomenclature. The institution of correct Dalbergia classifications and nomenclature and the identification of illegally traded endangered Dalbergia species, has been stymied by the absence of a rapid and consistent mechanism by which to distinguish between species, and routinely detect those that are banned. In this work the chemical profile derived from a DART ion source coupled to a high-resolution time-of-flight (TOF) mass spectrometer was used to rapidly and consistently identify and distinguish between Dalbergia species. We proceeded on the premise that the rare D. granadillo was a synonymous species to other less cryptic taxa. Because the known curated xylarium reference samples of D. granadillo are extremely rare (n = 11 worldwide), we decided to first compare the known xylarium-authenticated samples of D. granadillo hardwood by DART-TOF-MS to determine whether they exhibited similar chemical fingerprints. The results were then contrasted with the DART-TOF-MS spectra of five species of timber from a variety of sources that have been described by the common name of "granadillo" (i.e. D. granadillo, D. retusa, P. yucatanum, C. echinata, and C. platyloba-Supplementary Table 1). The mass spectra generated (Supplementary Figure 1) were rendered as heat maps using the Mass Mountaineer software suite. Supplementary Tables 2a-c show the corresponding measured m/z values and their abundances, and Fig. 1a illustrates the mass spectral heat maps. The results show that D. retusa and D. granadillo have similar compounds present in roughly the same relative amounts as indicated by the similar intensities of the indicated colors in the heat maps. The other three species show different and distinctive diagnostic ion patterns. The high-resolution masses of several of the molecules detected were consistent with those of compounds previously reported to be present in Dalbergia 13,14 . Figure 1b is a graphical representation of the Kernel discriminant analysis (KDA) plot generated using 104 feature masses ranging from m/z 107.037 -m/z 527.155 from a training set of 102 spectra. The plot shows that D. retusa and D. granadillo form a single cluster that cannot be differentiated using KDA. The leave-one-out cross validation (LOOCV) for the KDA classification model analysis was fairly poor at 64.29%, reflecting the fact that these Dalbergia species could not be separated. The clustering of D. retusa and D. granadillo is supportive of our hypothesis that both represent one and the same species and therefore should be described by a single name. Indeed, when D. retusa and D. granadillo were joined into a single group under one name (D. retusa), the LOOCV of the KDA model rose to 98.98%. Thus, our observations support the premise that from the chemical profile of the heartwood, D. granadillo cannot be distinguished from D. retusa and that D. granadillo and D. retusa may be synonymous. Interestingly, we found that when the heat map data were imported into a third-party hierarchical clustering program such as Cluster 3.0, the resulting dendrogram classified the various Dalbergia samples according to species and illustrated their genetic relatedness. Cluster analysis was performed using uncentered correlation of 436 variables of the spectral data, and a typical result is shown in Fig. 1c. The leaves highlighted in red in Fig. 1c are D. granadillo specimens. The dendrogram shows that the D. granadillo clusters with the D. retusa samples, supporting the hypothesis that both represent the same species. ## Inferring the Phylogeny of Biodiesel Feedstocks From Fatty Acid Methyl Ester (FAME) Profiles. Biodiesel is a renewable fuel derived from vegetable oils or animal fats by transesterification of triglycerides with an alcohol, generally methanol, in the presence of a catalyst 15 . The resulting mixture of fatty acid methyl esters (FAMEs) can be used to fuel diesel engines and is most often blended with petroleum diesel. The amount of biodiesel produced in the U.S. has increased significantly in recent years. In 2010 production was just over 300 million gallons, which increased to nearly 1 billion gallons in 2011. In 2013, production reached over 1.3 billion gallons 16 . Increased production and utilization of biodiesel has intensified interest in the properties of this fuel and how these properties impact engines and infrastructure. The feedstock from which biodiesel is derived determines many of the properties of the fuel. These properties are directly related to fatty acid makeup of different oil sources 17 . Desired properties such as cold weather operability and resistance to autoxidation are influenced by the acyl chain length and degree of unsaturation of the fatty acids in the feedstock 18,19 . If the feedstock used to manufacture a biodiesel is unknown to the user, the source may be determined from the FAME profile of the product if the unique fatty acid distribution of the source oil is known 20 . FAME profiling is commonly achieved with gas chromatography 21 . This analysis can be time consuming, particularly if a high degree of resolution is required to isolate FAMEs in more complex samples. We determined that positive-ion DART can be used to rapidly determine the FAME profile of biodiesel, allowing for quick source and properties identification. The biodiesel samples utilized in this study included the most commonly used feedstocks in the United States 22 . Arugula (Eruca sativa), Brassica (Brassica juncea), Field Pennycress (Thlaspi arvense), Cress (Lepidium sativum), Camelina (Camelina sativa), Meadowfoam (Limnanthes alba), and Cuphea (Cuphea lanceolata) seed oils were provided by the USDA National Center for Agricultural Utilization Research. The DART mass spectra (Supplementary Figure 2) obtained for hexane solutions of the aforementioned ten biodiesel feedstocks were dominated by both saturated and unsaturated FAMEs ranging in size from 11-23 carbons (Supplementary Table 3). Hexane was used to dilute the samples because they proved to be too concentrated in their native form. The most abundant species in the majority of feedstocks (i.e. Brassica, Camelina, Canola, Cuphea, Pennycress and Soy) were C 19 FAMEs (derived from C 18 fatty acids). Nevertheless, the mass spectra were consistent for samples within the same species, but very clearly different between species (Supplementary Figure 2). Fifteen feature masses were used for principal component analysis (Fig. 2b). Species level clustering was observed in the covariance PCA plot with five principal components accounting for 92.6% of the variance and the LOOCV was 98.33%. Subjection of the corresponding mass spectral heat maps (Fig. 2a) to hierarchical clustering analysis showed that each species was clearly separated from the others (Fig. 2c). All members of Order Brassicale were clustered together with the exception of Canola, which is a cultivar that has been bred to have low erucic acid content. The remaining feedstocks (including the mixed biodiesel) belonging to different orders and/or families, comprised a separate cluster. Interestingly, Meadowfoam (order Brassicale, family Limnanthaceae) was distinct from all of the other species. These observations illustrate that easily and rapidly acquired feedstock chemical profile information can be translated into dendrograms that clearly distinguish between genera to show their evolutionary relationships, and that the data can be generated in a high throughput fashion. Fly Species Identification from Insect Puparial Cases. Blowflies (Diptera: Calliphoridae) are important to forensic entomology because they are often the first colonizers of decomposing remains and can offer significant diagnostic information towards calculating an accurate minimum post mortem interval (PMI min ). Calliphoridae puparial cases are often the only persisting entomological evidence in criminal investigations involving highly decomposed remains 23 . These cases are the empty shells of the last layer of the larval stage (post feeding). Many studies have been published using larvae and pupal stages for PMI estimations , but much less research has been published on puparial cases and currently, they are rarely used in criminal investigations due to the difficulty in identifying and ageing them. However, in the past decade, some studies have suggested that invaluable information can be extracted from puparial cases and hence, new methods to identify them are being developed 23,29 . When an adult fly emerges from the case it does so from the mouth end, leaving the rest of the case intact. To correctly identify them, the same morphological features used for the pupae are examined (i.e. posterior spiracles, spines, and mouth piece if present). However, with empty cases, these morphological features have often been destroyed during emergence of the adult fly. It is well established that insect cuticular hydrocarbons have characteristic profiles for different species . The same holds true for the hydrocarbon profiles of insect puparial cases 35 . Therefore, hydrocarbon analysis is advantageous because both young and aged cases retain definitive chemical information due to the stability of the constituent hydrocarbons despite weathering effects. Positive-ion DART-MS was previously used to analyze the unsaturated cuticular hydrocarbons of awake behaving fruit flies (Drosophila melanogaster) 36 . However, this form of analysis does not give clear, unambiguous mass spectra for saturated alkanes. Recently, we reported that large polarizable alkanes, lipids and alcohols can be detected as O 2 − adducts ([M + O 2 ] − ) by aspirating sample solutions directly into the mass spectrometer atmospheric pressure orifice in the presence of the O 2 − generated by the DART ion source 37 . We applied this technique to our analyses. The mass spectra typically observed are presented in Supplementary Figure 3 and the heat map renderings of these spectra are shown in Fig. 3a. The detected C 27 -C 34 alkanes that were used as the basis for multivariate statistical analysis by supervised methods were easily observed as O 2 − adducts (see mass spectral peak assignments and molecule abundances in Supplementary Table 4). All species are distinctly separated within the PCA plot, demonstrating that their profiles have unique chemical differences (Supplementary Figure 3). However, there is less separation between L. cuprina and L. sericata. This is likely because they are from the same genus (Lucilia) and therefore their profiles share more similarities compared to the other species. It should be noted that this method does not distinguish between hydrocarbon isomers or provide any information about branching. Nevertheless, the results confirm that the hydrocarbon profiles measured by DART-MS clearly enable distinctions between species to be made. A training set comprised of blowfly puparial case hexane extract mass spectra (i.e. Chrysomya rufifacies, Lucilia sericata, L. cuprina, and Cochliomyia macellaria) as well as spectra of puparial case extracts of the common house fly (Musca domestica) was created. Feature masses from these spectra were used for KDA, with the results featured in Fig. 3b. Excellent separation between the five different insect types was observed and LOOCV gave 100% correct identification for all samples. Five sets of puparial cases labeled "A" though "E" that were provided as blind samples were then analyzed. Samples A, B, C and D were correctly identified as L. sericata, C. rufifacies, L. cuprina, and C. macellaria respectively. Sample E gave a distinctly different profile. It was later revealed that it represented puparial cases for the common housefly, Musca domestica. Although the mass spectral data for Sample B correctly clustered with that of C. rufifacies, it differed somewhat from the standard C. rufifacies samples measured one month earlier. This is illustrated in a comparison of the Sample B spectrum (Supplementary Figure 4) with that of the spectrum obtained for C. rufifacies (Supplementary Figure 3). The reason for this difference is not clear, but preliminary observations indicate that the alkane profiles for puparial cases of a given species may vary with age 35 . Although this difference is evident in the PCA plot (Fig. 3b), hierarchical clustering analysis of the mass spectral datasets that were rendered as heat maps showed the B samples and the C. rufifacies standards clustering together (Fig. 3c). The DART-MS derived small molecule fingerprint of a puparial case as a function of its age is currently being investigated by the authors, as a correlation between the two could potentially serve as a tool in post mortem investigations. ## Species Identification From Seeds of Plants Containing Belladonna Alkaloids. The genus Datura contains multiple species of ornamental flowering plants of horticultural importance. They are a well-known source of belladonna alkaloids including scopolamine and atropine, whose hallucinogenic and narcotic properties have been exploited in traditional religious rituals, herbal and mainstream medicine, and more recently in recreational drug abuse using its seeds 38 . It is often difficult to distinguish Datura species due to similarities in the appearance of both their seeds and their aerial parts, and because their morphological features can vary depending on where the plants are grown 39 . Datura plants also often bear resemblance to those in the Brugmansia genus, and this has led to the misidentification of some genus Brugmansia plants as belonging to the Datura genus and vice versa, as well as misidentification of species within the Datura genus 39 . An additional challenge is that the morphological features that allow the plants to be distinguished, most notably the flowers and fruits, take months to years to appear, making species differentiation a long-term project. Furthermore, from a chemical profiling standpoint, the belladonna alkaloids found in Datura species also appear in Brugmansia and Hyocyamus species, making identification of plant material based on the presence of belladonna alkaloid biomarkers alone indeterminate. All of the aforementioned species are members of the large group of non-model plants that are poorly annotated and whose genomes have not been mapped, making phylogenetic species identification impossible. Here, direct analysis in real time-mass spectrometry (DART-MS) and hierarchical clustering analysis tools were applied to the seeds. The approach provided a rapid high throughput and viable method to test seeds directly for identification purposes, as well as for species differentiation and classification by cluster analysis. Of the known Datura species, we used D. ferox, D. stramonium and D. inoxia, as these are commonly abused seeds. Since the belladonna alkaloids that serve as biomarkers for Datura species are also present in Brugmansia and Hyocyamus seeds, both were also analyzed to assess whether they could be distinguished as species unique from Datura. Figure 4a shows the mass spectral profiles of all five species, done in replicates of 3-5, rendered as heat maps. The corresponding raw mass spectra and the peak abundances are presented in Supplementary Figure 5 and Supplementary Tables 5a-5e respectively. The high resolution data revealed that several of the detected molecules had molecular formulas consistent with molecules that have been observed in Datura spp. such as tropine, scopoline, dihydroxytropane, hexose sugars, scopolamine, 3-tigloyloxy-6,7-dihydroxytropane, vanillin, linoleic acid and oleic acid (see Supplementary Tables 5b-5d) . By visual inspection it was apparent that the mass spectra of each species were quite unique, even for the three Datura species. A total of 31 feature masses representing diagnostic peaks were used as a training set for the Kernel principal component analysis (KPCA). The results are shown in Fig. 4b. Each of the species was well clustered and could be distinguished from the others. Nevertheless, three principal components accounted for only 40% of the variance and increasing the number of principal components to 5 accounted for only 64% of the total variance. However, the LOOCV was 96%. The power of the chemical fingerprint signatures in permitting species differentiation and classification was demonstrated when the heat map data were imported into Cluster 3.0. Processing of the data in this manner furnished a dendrogram in which each of the seeds of the same species fell within clades that were representative of species classifications based on morphological feature differences, and were clearly distinct from one another (Fig. 4c). ## Species Differentiation of Eucalypts From Mass Spectrometry-derived Tissue-dependent Chemical Fingerprints. The Eucalyptus genus covers a diverse range of flowing trees and shrubs with more than 700 species that are broadly distributed throughout the Americas, Australia, Africa, and Europe. They are commonly known as "gum trees" because of the distinct and pleasant volatile exudate that is produced in response to a tissue breach. Many species have attracted global attention as a source for fragrance oils, biofuels, a fast-growing wood source and other commercial applications 43 . DART-MS profiling of eucalypt species was previously selected as a facile method to classify temperature-dependent emissions of volatile organic compounds (VOCs) for their atmospheric contributions in relation to changing climates and global warming, and to better estimate the range of biogenic pollutants released into the atmosphere during wildfires 5,6 . In that work, VOCs from stems and leaves of several eucalypts including E. cinerea, E. citriodora, E. nicholii and E. sideroxylon were identified. A wide range of compounds from simple organics (i.e. methanol and acetone) to a series of monoterpenes (i.e. pinene, camphene, cymene, eucalyptol) common to many plant species, as well as less abundant sesquiterpenes and flavonoids, were detected. This was achieved by stepwise adjustment of the DART helium gas temperature from 50 to 100 to 200 and to 300 °C, which enabled direct evaporation of compounds up to the onset of pyrolysis of plant fibres (i.e. cellulose and lignin). The identification of compounds was facilitated by correlating the observed high resolution accurate mass data to plant library compounds, and further matching their theoretical and experimental isotopic distributions. In the current work the initial VOC temperature-dependent emission studies have been extended to chemometric-based processing of mass spectral data for species differentiation. DART-MS analyses of leaf samples at a fixed temperature of 300 °C for several eucalypts including E. bridgesiana, E. cinerea, E. globulus, E. citriodora and E. polyanthemos was conducted. The observed spectra, each of which represents the average of 5 individual spectra, are shown in Supplementary Figure 6, with the corresponding measured m/z and peak abundance values presented in Supplementary Tables 6a-6e. The heat map renderings of the spectra are shown in Fig. 5a. The results revealed the presence of a number of chemotypes common to all species including monoterpenes (m/z 137, C 10 H 17 ) and various sesquiterpenes (m/z 205, C 15 H 25 ). Several of the detected formulas are consistent with those of compounds isolated from the species (outlined in Supplementary Tables 6a-6e) . Although all the species shared most of the dominant ions, they differed primarily in the relative abundance of detected compounds. A total of 15 feature masses representing m/z values varying from 155 to 509 were used for KDA. The resulting plot is shown in Fig. 5b. Three principal components accounted for 94% of the observed variance, and the LOOCV was 83%. When the mass spectral heat maps (Fig. 5a) were processed using Cluster 3.0, the resulting dendrogram (Fig. 5c) showed excellent species level discrimination, and none of the data were misclassified, thus demonstrating the robustness of the approach of using the entire mass spectral data set in providing the information needed for species-level distinctions to be made. ## Discussion Statistical processing of the output of chemical analysis techniques for the purposes of typing and classification is not new. For example, hierarchical cluster analysis of Raman spectroscopic data has been used to classify tree pollens 47 and genus Mentha plants 48 . Multivariate statistical analysis has been applied to gas chromatographic results to classify the geographic origin of cocoa beans 49 , as well as 1 H NMR data for the analysis of wines 50 . Chemometric discrimination of coffee beans by area of origin has been demonstrated using Fourier transform infrared spectroscopy 51 . The output of various mass spectrometric methods of small molecule profiling has also been similarly analyzed with varying results. Examples include multivariate statistical analysis of data generated using: 1) Curie Point pyrolysis mass spectrometry for classification of bacteria 52 ; (2) paper spray mass spectrometry for determination of the geographic origin of coffee 53 ; (3) HPLC-tandem MS of herbal medicines to determine country of origin 54 ; (4) Ultraperformance liquid chromatography-time of flight mass spectrometry for classification of wheat lines 55 ; (5) LC-MS/MS for the assessment of the utility of using bioactive components as the basis of distinguishing between herbal medicines 56 ; (6) RPLC ESI-MS for standardization of Ginkgo biloba extracts 57 ; (7) direct injection electrospray MS for classification of coffee trees 58 ; (8) ion molecule reaction mass spectrometry for bacterial species differentiation 59 ; (9) GCand atmospheric pressure photoionization (APPI) MS for classification of natural resins 60 ; (10) GC-GC TOF/MS for characterization and authentication of edible oils 61 , and ( 11) pyrolysis GC-MS profiling of eucalypt emissions in response to climate change and wildfires 62 , among other examples. The method described here differs from those outlined in the aforementioned studies in that in general, data acquisition is simpler, a broad range of compounds spanning the dielectric constant spectrum can be detected in a single experiment, and the entire information content of the observed DART-MS-derived chemical fingerprints is subjected to unsupervised hierarchical clustering (rather than using a subset of feature masses and/or chromatographic peaks). Besides DART-MS, desorption electrospray ionization mass spectrometry (DESI-MS) is another ambient ionization mass spectrometry technique that exhibits advantages similar to those noted for DART-MS. However, relatively few studies featuring DESI-MS in metabolome profiling and/or chemical fingerprinting have appeared. Recently, Watrous et al. 63 demonstrated the use of "nanospray" DESI-MS for the in vivo metabolic profiling of bacterial colonies directly from a Petri dish. The report further illustrates the power of ambient ionization mass spectrometric methods to rapidly provide unprecedented glimpses of real-time changes in chemical fingerprint profiles in ways that are difficult and/or impossible to accomplish by more conventional methods. In this report, we show that a variety of chemotypes from a diversity of samples can be readily detected under similar conditions. The high resolution Dalbergia species results revealed the presence of several molecules with formulas consistent with those of compounds that have been identified in Dalbergia including neoflavonoid quinone derivatives such as the dalbergiones, various isoflavones, guainolide sesquiterpene lactones, auxins such as indole-3-acetic acid, pyrano-and furano-benzenes and diterpenes among many other polar and non-polar small molecules 13,14 . Both saturated and unsaturated biodiesel feedstock-derived FAMEs of from 11 to 23 carbons were easily observed in positive ion mode. In analysis of the biofuels, we observed that the biodiesel was most conveniently analyzed by first diluting it with a non-polar solvent such as hexane, in order to make it less viscous. Alkanes and alkenes from 27-34 carbons long were observed as O 2 − adducts in hexane extracts of fly puparial cases, showing distinct variations in profile and abundance as a function of species. Our approach to the analysis of the puparial cases represents the first published application of the O 2 − attachment ionization technique to address an analytical problem. This novel method enabled us to easily detect large polarizable alkanes, lipids and alcohols as [M + O 2 ] − adducts, by aspirating sample solutions directly into the mass spectrometer atmospheric pressure orifice in the presence of the O 2 − generated by the DART ion source 37 . The application of this technique necessitated the use of the solvent which, in this case, was hexane. Although these large non-volatile species could have been detected by GC-MS or field desorption, the former method is much slower than DART-MS analysis, while the latter requires introducing the sample into a vacuum on a fragile emitter. Neither approach is as convenient as the DART-MS analysis described here. In analysis of Datura, Brugmansia and Hyocyamus species plants, a range of compounds of varying polarities was observed, as illustrated in Tables 5a-e. Amines, sugars and fatty acids, among hundreds of other compound types, were all detected in seconds in positive as well as negative ion modes. In the case of the Eucalpyts, direct leaf analysis yielded spectra in which the presence of the odiferous mono-and sesquiterpenes for which this species is well known were all readily apparent. It was the consistent and reproducible comprehensiveness of the rapidly acquired small molecule fingerprint in each of the biological samples surveyed in this work that was exploited to conduct successful species classification using multivariate statistical analysis tools. Using supervised methods such as PCA, KDA and KPCA, we determined that a small number of principal components could be used to account for ~40% -90% of the observed variance, with LOOCVs from 80 -100% probability depending on the sample analyzed. Hierarchical cluster analysis was applied to the entire mass spectral data set in each case as an unbiased approach to assess the extent to which the DART-MS-derived chemical fingerprint could enable species classification. The resulting dendrograms showed that in all cases, striking species level separations were accomplished, demonstrating that genomic distinctions between even closely related species manifest themselves in small molecule profile differences. A number of previous reports have demonstrated that hierarchical clustering of the type used here can be exploited for classification purposes. However, in the majority of these cases, distinguishing biomarkers or specific spectral features, rather than the entire small molecule fingerprint, were used. A recurring observation in these studies was the appearance of misclassifications, a not too unexpected consequence of the fact that (a) the selected principal components accounted for less than 100% of the variance; and (b) the information content of the chemical data that was used as the input for multivariate statistical analysis processing was not comprehensive, in that it was acquired using extracts, or the analysis was performed by a method in which certain molecules were preferentially detected over others. For the samples used in this work, no misclassifications were observed when the entire mass spectral dataset was used. This suggests that small molecule fingerprint-based classifications that can reflect genome differences are best acquired using the full fingerprint, rather than a subset of salient features. Of note is the fact that this method does not require that the identities of the fingerprint components be known. Nevertheless, the knowledge of the molecular weights and formulas of distinguishing molecules provides important information that can be used to eventually determine compound identity. In summary, we have devised a rapid high throughput method for species identification and classification based on chemometric analysis of comprehensive DART-TOF-MS derived chemical signatures. The method entails introduction of the sample to the open air space between the DART ion source and the mass spectrometer inlet, followed by chemometric processing using the entire mass spectral dataset. A range of both polar and non-polar chemotypes are instantaneously detected, and matter in various forms (i.e. solid, liquid or gaseous) is easily analyzed with no need to change the method of sample introduction. The comprehensive small molecule signatures obtained serve as the input for unsupervised hierarchical cluster processing software, a number of open source versions of which are freely and readily available. The result of this processing tool is identification and species level classification based on the entire DART-derived chemical fingerprint. This methodology circumvents some of the pitfalls of the data selection bias that can accompany the use of supervised methods of statistical analysis on the one hand, and the deficiencies introduced by other instrument/chemical methods (such as extraction) on the other. Furthermore, it is significantly faster than conventional methods and can yield results from start to finish (including statistical analysis), in less than 3 min per sample. Given that the type of genome classification results consistently observed here are most often acquired using gene sequence information, and the time and resources required to generate it, the method outlined here provides a significant advancement in the determination of species level classifications. It supplies further evidence that inherent in the metabolome is the information content required to determine species level distinctions. In this work, we show the application of this methodology for rapid species-level identification of: endangered woods; biofuel feedstocks; insect puparial cases; plants; and tree species. These applications fall within the fields of forensic science, agronomy, agriculture, natural products chemistry, plant biochemistry and fuel chemistry among others, and it is anticipated that it could easily be used to further discoveries in a myriad of other areas. ## Methods Instrumentation. An AccuTOF (JEOL Ltd., Akishima Japan) time-of-flight mass spectrometer equipped with a Direct Analysis in Real Time (DART) ion source (Ionsense LLC, Saugus, MA) was used for all measurements. Mass spectra were stored by the JEOL Mass Center data acquisition software at a rate of 1 per second for the m/z range 60 to 1000. The mass spectrometer resolving power was 6000 (FWHM) for protonated reserpine at m/z 609.2812. The atmospheric pressure interface (API) conditions for positive-ion measurements were: orifice 1 = 20 V, ring lens = orifice 2 = 5 V. The RF ion guide voltage ("Peaks Voltage") was set to 600 V to permit analysis of ions greater than approximately m/z 60. For all analyses except those of the seeds and Eucalyptus leaves, a sample of poly(propylene glycol) with average molecular weight of 600, also referred to as "polyethylene glycol" (PEG 600), was measured in each data file as a reference standard for mass calibration. For the remaining samples, Jeffamine M600 (Huntsman, The Woodlands, TX) was used as the calibrant. Unless otherwise stated, the DART was operated with helium and a gas heater setting of 350 °C. Sample extracts were analyzed by exposing the closed end of a Corning Pyrex melting point capillary tube (Capitol Scientific, Austin TX USA) that had been dipped into the extract, to the open air space between the ion source and the mass spectrometer inlet. Mass spectral data processing. Data processing operations, including mass calibration, centroiding, spectral averaging and background subtraction were carried out with TSSPro3 software (Shrader Software Solutions). Mass Mountaineer software (RBC Software, Portsmouth, NH) was used for classification chemometrics including heat maps, principal component analysis (PCA) and linear and kernel discriminant analysis (LDA and KDA respectively). Heat maps exported from Mass Mountaineer were imported into Cluster 3.0 and Java Treeview (Stanford University) for hierarchical clustering analysis. Sample preparation and sample analysis. Dalbergia species. Because of the common practice of using a single name to refer to multiple species within the Dalbergia genus, and the fact that many of the samples we analyzed are rare and illegal to trade, we conducted our analyses on samples from xylarium collections whose species identities had been verified. We then compared these to samples from commercial sources. Wood samples of known identity were sourced from the USDA Forest Product Laboratory (FPL), the USDA Animal and Plant Health Inspection Service (APHIS), the Oregon State University Xylarium (OSU), La Xiloteca del Instituto de Biología, UNAM, Mexico City, México (XIB), Eisenbrand Inc. Exotic Hardwoods, Torrance, CA, USA (EIEH), Cook Woods, Klamath Falls, OR, USA (CW), Carlton McLendon Inc., Atlanta, GA (CMI), PFC Shanty Navarro Hurtado, the Brazilian Federal Police (SNH), and the Botany collection at the University of South Carolina (USC). Furthermore, samples from multiple countries (Mexico, Guatemala, Nicaragua, Panama, Costa Rica and Brazil) were analyzed. The number of replicates that could be analyzed depended upon and was limited by species availability. The comprehensive list appears in Supplementary Table 1. Briefly, 11 D. granadillo, 34 D. retusa, 22 P. yucatanum, 21 C. echinata and 12 C. playloba species were analyzed. For sampling, wood slivers were shaved from the heartwood of the reference specimens and placed directly in the DART helium gas stream for six seconds each. A mass calibration standard of polyethylene glycol 600 (Ultra, Kingstown RI) was run between every 5 th sample. For each species, sampling was conducted in replicates of 8-9. Biofuel feedstocks. Soy-derived, canola-derived, and mixed feedstock biodiesels were obtained from Minnesota Soybean Processors (Brewster MN, USA), Archer Daniels Midland (Decatur IL, USA), and Future Fuel (Batesville, AR, USA), respectively. Non-commercial biodiesel samples were supplied by the United States Department of Agriculture, National Center for Agricultural Utilization Research, Agricultural Research Service (Peoria IL, USA). Hexane used to dilute samples for analysis was purchased from VWR (Denver CO, USA) and used as received. Biodiesel samples were measured by dipping the closed end of a melting point capillary tube into hexane solutions of each feedstock (30 μ L of feedstock dissolved in 100 μ L of hexane), and suspending the tube between the mass spectrometer inlet and the ion source. Solutions were sampled by DART-TOF-MS as described above in replicates of 5 for each feedstock. Puparial cases. Puparial cases were provided by Dr. Jeffery Tomberlin (Texas A&M University, College Station TX USA) and Dr. Eric Benbow (Michigan State University, USA). Individual insect cases were deposited into vials containing 300 μ L of hexane (Thermo Fisher Scientific, Waltham MA USA) and allowed to stand for 5 min before DART sampling of the extract using the sealed end of a melting point capillary. The DART exit grid potential was set to + 250 V. For every species, 5 cases were sampled in replicates of 5 each. Datura, Brugmansia and Hyocyamus species differentiation. B. arborea and D. ferox seeds were purchased from Georgia Vines (Claxton GA, USA). H. niger, D. stramonium, and D. inoxia seeds were purchased from Horizon Herbs (Williams OR, USA). Individual seeds were sampled by DART-TOF-MS using a vacuum tweezer apparatus to suspend the seeds between the ion source and the mass spectrometer inlet. For analysis, seeds were cut in half and one open half of the seed was oriented so that if faced the DART ion source. For each species, mass spectra were measured in replicates of 5. Eucalypt analysis. The species of Eucalyptus analyzed were Eucalyptus polyanthemos (10 plants with 5 replicates from each plant), E. bridgesiana apple (2 plants with 25 replicates from each), E. globulus (10 plants with 5 replicates each), E. citriodora (10 plants with 5 replicates each), and E. cineraria (4 plants with 17 replicates each). All plants except E. bridesiana were purchased from Companion Plants Inc. (Athens, OH, USA). E. bridgesiana was purchased from Faddegon's Nursery (Latham, NY, USA). Plant leaves were sampled by removal of 6 mm diameter circular segments from the leaves of live soil bound plants with a paper hole punch and suspending the leaf sample in the open air space between the ion source and the mass spectrometer inlet. ## Multivariate statistical analysis. Mass-calibrated and centroided mass spectra were exported from the data processing software (TSSPro3, Shrader Software Solutions, Detroit, MI) as text files for entry into the elemental composition and classification software (Mass Mountaineer, RBC Software, Portsmouth, NH, available from mass-spec-software.com). Principal components were calculated by using the correlation matrix. Abundances used for classification were selected from each mass spectrum for the indicated number of peaks having m/z values within 0.005-015 u of the target m/z value. Heat maps were rendered as text files for import into Cluster 3.0 for single linkage hierarchical cluster analysis (Michiel de Hoon, University of Tokyo, adapted from the Cluster Program written by Michael Eisen, Stanford University, available at http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm). Dendrograms were observed using Java Treeview (written by Alok Saldanha, available at http://jtreeview.sourceforge.net/).

chemsum

{"title": "A High Throughput Ambient Mass Spectrometric Approach to Species Identification and Classification from Chemical Fingerprint Signatures", "journal": "Scientific Reports - Nature"}

structural_and_mechanistic_analysis_of_drosophila_melanogaster_agmatine_n-acetyltransferase,_an_enzy

## Abstract: Agmatine N-acetyltransferase (AgmNAT) catalyzes the formation of N-acetylagmatine from acetyl-CoA and agmatine. Herein, we provide evidence that Drosophila melanogaster AgmNAT (CG15766) catalyzes the formation of N-acetylagmatine using an ordered sequential mechanism; acetyl-CoA binds prior to agmatine to generate an AgmNAT•acetyl-CoA•agmatine ternary complex prior to catalysis. Additionally, we solved a crystal structure for the apo form of AgmNAT with an atomic resolution of 2.3 Å, which points towards specific amino acids that may function in catalysis or active site formation. Using the crystal structure, primary sequence alignment, pH-activity profiles, and site-directed mutagenesis, we evaluated a series of active site amino acids in order to assign their functional roles in AgmNAT. More specifically, pH-activity profiles identified at least one catalytically important, ionizable group with an apparent pK a of ~7.5, which corresponds to the general base in catalysis, Glu-34. Moreover, these data led to a proposed chemical mechanism, which is consistent with the structure and our biochemical analysis of AgmNAT.The discovery and characterization of enzymes involved in fatty acid amide biosynthesis has been a longstanding focus of our research 1 . One possible biosynthetic route for the fatty acid amides would be the reaction between an amine and a fatty acyl-CoA: R 1 -NH 2 + R 2 -CO-S-CoA → R 2 -CO-NH-R 1 + CoA-SH. Enzymes of the GCN5-related N-acetyltransferase family (GNAT) catalyze a similar reaction using acetyl-CoA as a substrate to generate N-acetylamides 2 . Acetyl-CoA-dependent N-acetylation by N-acetyltransferases is known for a diversity of amines 3-5 in a broad range of organisms 2,5-8 . We have long suspected that enzymes identified as N-acetyltransferases might accept longer-chain fatty acyl-CoA thioesters as substrates or that novel N-acetyltransferase-like enzymes exist that utilize fatty acyl-CoA thioesters as substrates.Drosophila melanogaster is an excellent model organism to study fatty acid amide biosynthesis. These insects are known to produce fatty acid amides 9,10 , its genome has been sequenced 11 , these organisms can be manipulated genetically 12 , and are inexpensive to maintain. In addition, two N-acetyltransferases had been identified from D. melanogaster, arylalkylamine N-acetyltransferase variant A (AANATA, also called dopamine N-acetyltransferase) 13 and arylalkylamine N-acetyltransferase-like 2 (AANATL2) 14 . Both enzymes catalyze the N-acetylation of arylalkylamines, but their respective substrate specificities, kinetic mechanisms, and chemical mechanisms were not fully defined prior to our work. A search of D. melanogaster genome using the sequences of AANATA and AANATL2 led to the identification of six other putative arylalkylamine N-acetyltransferase-like enzymes, AANATL3-8 14,15 . A complete understanding of the structural and mechanistic features of these enzymes will provide tremendous insight into rules governing acyl-chain length specificity for GNAT enzymes. An exhaustive evaluation of these enzymes for different amine or acyl-CoA substrates may yield new chemistries and define a biosynthetic route to the fatty acid amides. To define the substrate specificities of these putative N-acyltransferases, we devised a screening strategy that involved the evaluation of a collection of amines vs. a short-chain acyl-CoA or a long-chain acyl-CoA. Our screening strategy led to the discovery that D. melanogaster AANATL2 will utilize dopamine, serotonin, and long-chain acyl-CoA thioesters as substrates 16,17 . These results are likely of significance to mammals because the N-fatty acyldopamines have been identified in the brain 18 and the N-fatty acylserotonins in the gastro-intestinal tract 19 . Our application of the screening strategy to AANATL8 led to the identification of agmatine as the amine substrate with the highest (k cat /K m ) app for this enzyme. Thus, we have renamed AANATL8 as agmatine N-acetyltransferase (AgmNAT). The acetylation of agmatine points to novel agmatine related metabolites and new reactions in the degradation pathways of agmatine and arginine. A thorough study of the insect AANATs contributes to our understanding of fatty acid amide biosynthesis, enables a detailed comparison between the insect AANATs to the AANATs from other organisms, and fosters the development of insecticides targeted against insect AANATs. AANATs are suggested to be a good targets for the control of insect pests . Furthermore, AANAT inhibitors could also lead to drugs to treat circadian rhythm disorders because serotonin N-acetyltransferase catalyzes the rate-determining step in melatonin biosynthesis 24 . Agmatine, (4-aminobutyl)guanidine, was first described in 1910 25 and was later identified as the product of arginine decarboxylation 26 . Research concerning agmatine was limited until the 1990s 27 , until the discovery that agmatine is produced in the mammalian brain 28,29 . Agmatine is distributed in many tissues, including the stomach, intestine (large and small), adrenal gland, heart, aorta, spleen, lung, vas deferens, kidney, liver, skeletal muscle, and plasma . It is primarily located in cytoplasmic vesicles that are strongly associated with the mitochondria or endoplasmic reticulum 33,34 . Additionally, agmatine can translocate into the mitochondria and is likely associated with the Golgi complex, cell membrane, and nuclear membrane 39 . Agmatine is a neurotransmitter and neuromodulator in mammalian brain 27,40 , its physiological effects resulting from binding to the imidazoline (I 1 and I 2 ) 28,41,42 , α 2 -adrenergic 43 , nicotinic 44 , NMDA 45 , and serotonin receptors (5-HT2A and 5HT-3) 46 . Little is known about agmatine and its biosynthesis, degradation, elimination, and function in the fly or in insects, in general. Low levels of agmatine have been found in D. melanogaster 47 and agmatine has been reported from other insects . Likewise, little is known about arginine decarboxylase from insects 51 . Our discovery and characterization of AgmNAT may point to unappreciated role(s) for agmatine and/or N-acetylagmatine in D. melanogaster and other insects. Agmatine biosynthesis and degradation is shown in Supplementary Fig. S1. First, arginine decarboxylase (ADC) catalyzes the decarboxylation of arginine to generate agmatine , followed by agmatine degradation via two main routes: (a) hydrolysis to urea and putrescine, as catalyzed by agmatinase (AGMAT, also known as agmatine ureohydrolase) 29,52 or (b) oxidation to 4-guanidinobutanoic acid, as catalyzed by diamine oxidase (DAO) and aldehyde dehydrogenase (AlDH) . In Thermus thermophilus, polyamine aminopropyltransferase (SpeE) catalyzes the formation of agmatine N 1 -aminopropylagmatine 56,57 . In plants, agmatine coumaroyltransferase catalyzes the formation of p-coumaroylagmatine from p-coumaroyl-CoA and agmatine, p-coumaroylagmatine is thought to function in the defense system of the plant against infection 58,59 . An unexplored degradative pathway for agmatine is N-acetylation at the N1 position, catalyzed by AgmNAT to generate N-acetylagmatine -one of the subjects of this manuscript. AgmNAT is a member of the GCN5-related N-acetyltransferase family (GNAT) 2 and, in addition to the formation of N-acetylagmatine, this enzyme also catalyzes the production of N-acylpolyamines from the corresponding acyl-CoA and polyamine. We also present data showing the AgmNAT structure, substrate specificity, and kinetic and chemical mechanism for the AgmNAT-catalyzed reaction. ## Results and Discussion Crystal structure of AgmNAT. A homology model for AgmNAT was constructed using the Aedes aegypti arylalkylamine N-acetyltransferase structure 21 as a template for molecular replacement. The AgmNAT (CG15766) crystal structure was determined at 2.3, with two monomers in the asymmetric unit of the P2 1 space group (Table 1). The two monomers are nearly identical with an RMSD value of 0.262 when aligning 862 backbone atoms. Similar to the arylalkylamine N-acetyltransferase model, the new structure is primarily composed of six α-helices and seven anti-parallel α-strands (Fig. 1A). The AgmNAT structure displays a conserved GNAT fold, similar to that observed for D. melanogaster AANATA and human spermidine/spermine N 1 -acetyltransferase (SSAT) (Supplementary Fig. S2), though the sequence identity is low when compared to these N-acetyltransferase enzymes (24% with AANATA and <20% for SSAT), a known feature of GNAT enzymes 2 . Based on the functional and structural similarities between AgmNAT and other GNATs such as AANATA (PDB 3TE4) 15,60 , we predict the active site pocket to be similar, though not identical, for the binding of the acyl-CoA and amine substrates (Fig. 2). The active site is well defined in the 2Fo-Fc electron density map (Fig. 1B,C) and is located near the crystal packing interface for both monomers. Based on the structure of AANATA with acetyl-CoA bound (PDB 3TE4) 60 , the binding surface for the adenosine 3-phosphate 5-pyrophosphate moiety of CoA-SH is blocked by protein-protein interactions in the AgmNAT structure, but the rest of the active site is open. The splaying of β-strand four and five, a conserved structural feature in GNAT enzymes, is also displayed in AgmNAT, which is the binding site for the pantetheine arm of acetyl-CoA 2 . Moreover, a conserved glutamate, Glu-34, that serves as the catalytic base for other D. melanogaster N-acyltransferase enzymes, is located within an accessible pocket that can accommodate the acyl-CoA and amine substrate, similar to that observed for AANATA (Fig. 1B) 15 . Also observed in the active site pocket are the residues, Pro-35 and Ser-171 (Fig. 1B,C), which are conserved amino acids that regulate catalysis in other D. melanogaster N-acyltransferases 15,61,62 . The functional roles of Pro-35 and Ser-171 of AgmNAT are discussed in subsequent sections. ## Evaluation of acyl-CoA steady-state kinetic constants. AgmNAT showed minimal differences in the measured K m,app values for acyl-CoA substrates ranging from acetyl-CoA to decanoyl-CoA (C2-C10) (Table 2) when agmatine was used as the saturating amine substrate. However, there was an acyl chain length dependent decrease in the apparent k cat value for the acyl-CoA substrates as the chain length is increased. This apparent decrease in the turnover number of ~150-fold from acetyl-CoA to decanoyl-CoA, led to the observed acyl-chain length specific decrease in the (k cat /K m ) app value. In addition, oleoyl-CoA was not a substrate at a concentration of 500 μM. These data likely result from the acyl-chain partially (decanoyl-CoA) or fully (oleoyl-CoA) occupying the amine binding site, perturbing the productive binding of agmatine; therefore, resulting in a decrease in or complete loss of catalysis. Similar results were observed for other D. melanogaster N-acyltransferases 15,61,62 . Evaluation of amine substrate steady-state kinetic constants. We screened >50 amines as potential AgmNAT substrates using acetyl-CoA or oleoyl-CoA as the co-substrate because of our interests in fatty acid biosynthesis, structure function relationships of GNAT enzymes, and the development of novel insecticides targeted to this class of enzymes. Our amine substrate screen included the canonical amino acids (except for Cys because Cys reacts with DTNB), amino acid analogs, other biogenic amines, and different xenobiotic amines. Only six amines (Table 3) showed AgmNAT activity >3-fold higher than the level of background acetyl-CoA thioesterase activity, whereas none showed a greater rate for oleoyl-CoA. Also, we identified five polyamines as AgmNAT substrates: spermine, N 8 -acetylspermidine, putrescine, spermidine, and cadaverine (Table 3). The (k cat /K m ) app values for the polyamines were lower than that measured for agmatine, the (k cat /K m ) app,agmatine /(k cat / K m ) app,polyamine ratio ranging from 15 for spermine to 1900 for cadaverine. Structural evidence for the specificity for agmatine and different polyamines likely results from the acidic nature of the active site, similar to that observed for the human ortholog (human SSAT) (Fig. 3) 2 . A more acidic active site can accommodate an amine substrate with a basic guanidinium group better than one with a hydrophobic aromatic group, giving rise to the difference in substrate specificity when compared to an AANAT 15,60 . AgmNAT was originally named AANATL8 based on primary sequence similarity 15 ; however, the substrate specificity data reported here support a new designation: agmatine N-acetyltransferase. This is the first report of agmatine serving as the best amine substrate for an N-acyltransferase. There are only a few reports of agmatine serving as a substrate within this family of enzymes 17,62,63 and only two reports on the identification of N-acetylagmatine from a biological source 64,65 . Rats fed heavy-atom labeled agmatine yielded two major urinary products; heavy-atom labeled N-acetylagmatine and unprocessed, but labeled agmatine 64 , suggesting a similar conversion as that catalyzed by AgmNAT. Inactivation of agmatine neurotransmission by N-acetylation is an underappreciated reaction between arginine, agmatine, and human disease 27, , the search for a human ortholog of Drosophila AgmNAT could lead to a new target for drug development. Additionally, selective targeting of Drosophila AgmNAT could result in the development of novel insecticides for insect control . We found that arginine, arginine methyl ester, N-acetylputrescine, and N 1 -acetylspermidine were not AgmNAT substrates. The ~25-fold increase in k cat,app for N 8 -acetylspermidine when compared to spermidine, together with our data demonstrating that N-acetylputrescine and N 1 -acetylspermidine were not substrates all suggest that AgmNAT, most likely, catalyzes the mono-and N1-specific acetylation of these biogenic amines, similar to what is observed for the mammalian spermidine N-acetyltransferase 69,70 . The increase in the k cat,app value, together with the small ~2-fold difference in the K m,app for N 8 -acetylspermidine relative to spermidine, could result from non-productive binding of the N8-amine of spermidine in the AgmNAT active site, whereby the N1-amine is better positioned for catalysis: deprotonation and then nucleophilic attack of the -NH 2 at the carbonyl of the acetyl-CoA thioester moiety. This means both of the amine moieties can bind in the active site, but only the N1-amine is acetylated. While arginine and arginine methyl ester are not AgmNAT substrates, we further evaluated these for AgmNAT inhibition to determine if either could bind to the enzyme. Arginine methyl ester proved to weakly inhibit AgmNAT, decreasing the rate of N-acetylagmatine formation from acetyl-CoA and agmatine by ~50% at 10 mM. In contrast, we found no inhibition of N-acetylagmatine formation at both 10 mM and 25 mM arginine. These data show that a modification of the α-position of agmatine inhibits binding to AgmNAT and that the inhibition results from both electronic and steric effects. The presence of the negatively charged α-carboxylate seems to eliminate or significantly weaken AgmNAT binding, likely the result of charge-charge repulsion. Evidence for this suggestion comes from the weak inhibition by arginine methyl ester (K i,s and K i,i ≥ 1 mM, Supplementary Fig. S3), but no apparent inhibition by arginine at a concentration as high as 25 mM. ## Kinetic mechanism. A combination of initial velocity kinetic experiments, dead-end inhibition, and product inhibition studies were used to determine the AgmNAT kinetic mechanism. Our first set of experiments was to vary the initial concentrations of one substrate at different fixed concentrations of the second substrate and fit these data to rate equations for either a sequential (Equation 3) or a ping pong (Equation 4) kinetic mechanism. Equation 3, for a sequential mechanism, provided the best fit (χ 2 is 4.6 for Equation 3 while χ 2 is 5.0 for Equation 4) which resulted in intersecting double reciprocal lines for acetyl-CoA (Fig. 4A) and agmatine Blue is for positive charges and red is for negative charges. Surface electrostatic potentials reveal that the amine binding pocket for AgmNAT and SSAT are more negatively charged than the arylalkylamine binding pocket of AANATA. (Fig. 4B). These data suggest that the AgmNAT-catalyzed formation of N-acetylagmatine occurs via a sequential mechanism; catalysis takes place only after formation of the AgmNAT•acetyl-CoA•agmatine ternary complex. Next, we determined if the AgmNAT kinetic mechanism is an ordered or random sequential mechanism by using substrate analogs, oleoyl-CoA, arcaine, and arginine methyl ester, as dead-end inhibitors vs. acetyl-CoA and agmatine. The inhibitor data is summarized in Table 4 and we have included the double reciprocal plots for the inhibitors in the Supplementary Materials. Arcaine is structurally related to agmatine, with its primary amine moiety replaced with a guanidinium group. Arcaine serving as an AgmNAT inhibitor supports our conclusion that AgmNAT does not acetylate the guanidinium amine of agmatine. None of these inhibitors showed any rate of catalysis above the slow, background rate of acetyl-CoA or oleoyl-CoA hydrolysis. Oleoyl-CoA produced competitive and noncompetitive inhibition plots for acetyl-CoA and agmatine (Table 4 and Supplementary Fig. S4A,B). Arcaine produced uncompetitive and competitive inhibition plots for acetyl-CoA and agmatine (Table 4 and Supplementary Fig. S4C,D). As observed for arcaine, arginine methyl ester produced uncompetitive and competitive inhibition plots for acetyl-CoA and agmatine (Table 4 and Supplementary Fig. S3). These data demonstrate that AgmNAT catalyzes the formation of N-acetylagmatine through an ordered sequential mechanism: acetyl-CoA binding first followed by agmatine to generate the AgmNAT•acetyl-CoA•agmatine complex prior to catalysis. This is similar to the kinetic mechanism for other D. melanogaster GNAT enzymes, including AANATA, AANATL2, and AANATL7 . Support for ordered sequential mechanism for AgmNAT comes from a statistically better fit to Equation 3 (as shown in Fig. 4) and the noncompetitive inhibition of oleoyl-CoA vs. agmatine (Table 4 and Supplementary Fig. S4A,B). Additional support and further details for the kinetic mechanism are revealed by N-acetylagmatine product inhibition. N-Acetylagmatine produced uncompetitive and competitive inhibition plots for acetyl-CoA and agmatine (Table 4 and Supplementary Fig. S5). Uncompetitive inhibition by N-acetylagmatine vs. acetyl-CoA (Supplementary Fig. S5A) is inconsistent with a ping pong kinetic mechanism. In sum, the kinetic analyses are consistent with two kinetic mechanisms: (a) ordered sequential substrate binding with acetyl-CoA binding first followed by ordered sequential product release with N-acetylagmatine being released last or (b) ordered sequential substrate binding with acetyl-CoA binding first followed by ordered sequential product release with CoA-SH being being released last. Uncompetitive inhibition by N-acetylagmatine vs. acetyl-CoA would be explained by the formation of a non-productive AgmNAT•acetyl-CoA•N-acetylagmatine complex with no reversible connection between the AgmNAT•acetyl-CoA complex and the AgmNAT•CoA-SH complex. We favor the latter mechanism because we have demonstrated that CoA-SH will bind to other D. melanogaster AANATs 15,61 and many other N-acetyltransferases exhibit ordered product release with CoA-SH being released last . Proposed AgmNAT chemical mechanism. We combined the pH-dependence of the kinetic constants, primary sequence alignment to other D. melanogaster GNAT enzymes 15 , determination of three-dimensional structure, and site-directed mutagenesis of a putative catalytically important residue to provide insights into the AgmNAT chemical mechanism. First, the pH-dependence of the kinetic constants was assessed for acetyl-CoA to assign apparent pK a values to ionizable groups involved in catalysis. Both the k cat,app and (k cat /K m ) app pH-rate profiles produced a rising profile with a pK a,app of 7.7 ± 0.1 and 7.3 ± 0.2, respectively (Fig. 5). An apparent pK a of ~7.5 can be attributed to a general base in catalysis, likely either deprotonation of the primary amine of agmatine or the zwitterionic tetrahedral intermediate generated upon nucleophilic attack of agmatine at the carbonyl thioester of acetyl-CoA. A second, higher pK a,app , possibly resulting from the deprotonation of a catalytically important general acid, was not observed in our pH-activity data, a surprising result given that a pK a ~8.5-9.5 has been observed for many other N-acyltransferases 2,75,76 . Explanations for these data include: (a) AgmNAT catalysis 3) than the rate equation for a ping pong mechanism (χ 2 is 5.0 for Equation 4). does not require a general acid, (b) the general acid in catalysis is not rate-limiting under our assay conditions, or (c) the general acid in AgmNAT catalysis has an apparent pK a > 9.5. Because of the high rate of base-catalyzed acyl-CoA hydrolysis, we cannot perform experiments at pH > 9.5 to define a pK a > 9.5. Next, we combined information from primary sequence alignments, the AgmNAT structure, and site-directed mutagenesis to define potential amino acids that could function in catalysis. A conserved glutamate has been proposed as the catalytic base in two D. melanogaster arylalkylamine N-acetyltransferases (AANATs), which corresponds to Glu-34 in AgmNAT 15,16 . Additionally, the AgmNAT structure shows that Glu-34 is in the active site, a buried region with several structural waters positioned within proximity of Glu-34 (Fig. 1B), similar to D. melanogaster AANATA (PDB code: 3TE4) 15 . Ordered water molecules within the active site of other GNAT enzymes are thought to form a "proton wire" that assists the general base in catalysis 2,15,17,63, . Although only a number of water molecules (36 in total) were sufficiently ordered to be modeled in the current structure, the majority of them are in the active sites of the two monomers. The closest ordered water molecules to Glu-34 is ~ 3.7 from the Oε 1 , positioned slightly too far for a hydrogen bond; however, we anticipate that the conformational changes upon substrate binding could promote hydrogen bond interactions between ordered water molecules and the functional groups in AgmNAT and substrate. Such hydrogen bonds could facilitate proton transfer from the amine substrate to initiate catalysis. In addition, unlike Glu-33, which is exposed to the bulk solvent, Glu-34 is relatively sheltered and placed close to the hydrophobic core of the protein and next to residues such as Leu-36. This microenvironment could be responsible for a pK a shift of Glu-34, as that identified in the pH-rate profiles. Therefore, we sought to interrogate the catalytic role of Glu-34 by evaluating the kinetic constants of the E34A mutant. The E34A mutation produced a catalytically deficient enzyme, exhibiting only 0.05-0.07% of the wildtype k cat,app value indicating that Glu-34 does function in the catalytic cycle. Furthermore, Glu-34 seems to have a role in substrate binding because the K m,app values for both agmatine and acetyl-CoA for the E34A mutant differ from wildtype values, the K m,app for agmatine increases 20-fold and the K m,app for acetyl-CoA decreases 6-fold (Table 5). The data generated for the E34A mutant is consistent, but does not prove, that Glu-34 serves as the general base in AgmNAT catalysis. To further investigate the role of Glu-34 in catalysis, we generated pH-activity profiles for the E34A mutant (Fig. 6). The k cat,app profile produced a pH-dependent linear increase with slope of 0.7 and (k cat / K m ) app profile with no slope. Attempts to titrate the pH < 8.0 were unsuccessful, by which a rate of CoA-SH release was not observed above the background hydrolysis rate. The linear profile in both the k cat,app and (k cat /K m ) app pH profiles, combined with the deficiency in catalytic rate suggest that Glu-34 serves as the general base in catalysis. Our steady-state kinetic data identified an ordered sequential mechanism with acetyl-CoA binding first, followed by agmatine to generate the AgmNAT•acetyl-CoA•agmatine ternary complex prior to catalysis. After the ternary complex formation, Glu-34 functions as the general base to deprotonate the positively charged amine moiety of agmatine, most likely involving a "proton wire" of ordered water molecules, followed by nucleophilic attack of the carbonyl of the acetyl-CoA thioester to generate a zwitterionic tetrahedral intermediate. Breakdown of the intermediate ensues by the departure of coenzyme A, which is, most likely, protonated by the positively charged amine of the intermediate (Fig. 7). This mechanism is consistent with other proposed chemical mechanisms for the N-acyltransferases of D. melanogaster and other organisms 15,16,24,78 . Other amino acids in AgmNAT that function in substrate binding and modulating catalysis. In addition to Glu-34, three other amino acids were individually mutated to alanine to define their function. These residues, Pro-35, Ser-171, and His-206, are conserved between D. melanogaster GNAT enzymes 15 and are proposed to function in active site formation, substrate binding, and/or regulation of catalysis 16,17 . The P35A mutant is catalytically deficient, with a k cat,app value that is ~2% of wildtype, while exhibiting only minimal K m,app differences when compared to wildtype for both acetyl-CoA and agmatine (Table 5). Similar results were observed for the corresponding proline in other GNAT enzymes, except most exhibited a significant K m increase for the corresponding amine, suggesting a role in substrate binding. Furthermore, the structure of sheep serotonin N-acetyltransferase (PDB code: 1CJW), co-crystalized with the tryptamine-acetyl-CoA bisubstrate inhibitor, shows that the corresponding Pro-64 interacts with this inhibitor via a CH-π interaction with the negatively charged face of the aromatic tryptamine moiety 77,79 observed for other GNAT enzymes 79 . In the current AgmNAT structure, Pro-35 is stacked on top of the imidazole ring of His-206 side chain (Fig. 2). The extensive van der Waals interaction may make significant contributions to particular active site configurations. Another active site residue evaluated for its role in substrate binding and catalysis is Ser-171. The S171A mutant only retained ~9% of the wildtype k cat,app and also showed a 3-to 4-fold change in the K m,app values for the substrates (a decrease in the K m,app for acetyl-CoA and an increase in the K m,app for agmatine) (Table 5). The decrease in the k cat,app could be interpreted that Ser-171 functions as a general acid in catalysis to protonate CoA-Sas it leaves the AgmNAT active site. For Ser-171 to function as a general acid during catalysis, the pK a of the serine hydroxyl would have to decrease by ~3-5 pH units to protonate the thiolate anion of the CoA product. We did not observe an apparent pK a in the pH-rate profiles that would correspond to a general acid, arguing against Ser-171 serving in this role. Alternatively, Ser-171 could have an important role in organizing the active site architecture to accommodate both substrates to enable efficient catalysis. Ser-171 is located in the active site, where its Oγ side chain atom forms hydrogen bonds with the backbone oxygen and nitrogen atoms of Ser-168, and a water-mediated interaction with the Thr-167 backbone nitrogen atom, suggesting that the 165-169 strand region in addition to Ser-171 is important in stabilizing the active site pocket to accommodate both substrates and allow for efficient catalysis to occur (Fig. 1C). The H206A mutant resulted in a k cat,app value that is ~18-fold lower than the wild-type value, whereas the K m,app-acetyl-CoA and K m,app-agmatine increased 2.3-fold and 1.4-fold, respectively. The corresponding residue (His-220) in D. melanogaster AANATA 15 was shown to interact with Tyr-185 and Pro-48 to form part of the active site, an interaction potentially resulting from a conformational change driven by acetyl-CoA binding. We assign a similar function for His-206 in AgmNAT since its general location in the active site is similar to His-220 in D. melanogaster AANATA, and the van der Waals interaction with Pro-35, as described above, is conserved (Fig. 2B). In addition, the His-206 side chain is in van der Waals contact with Ser-168 Cα and Tyr-188 Cε2, as well as several local prolines, Pro-203 and Pro-205. This means that His-206 is contributing to the formation of the active site by interacting with multiple residues. The apo-AgmNAT structure shows Tyr-170 in a position that is not optimal for a direct interaction with His-206 (Fig. 2A), unlike that shown for the corresponding residues in the AANATA structure co-crystalized with acetyl-CoA 15,60 . Tyr-170 occupies space near the entry point for acetyl-CoA into its binding pocket; therefore, we predict that a conformational change will occur that will move Tyr-170 into position for optimal acetyl-CoA binding, possibly by interacting with His-206. The findings presented in this manuscript highlight mechanistic and structural insights for D. melanogaster AgmNAT, an enzyme that catalyzes the formation of N-acetylagmatine from acetyl-CoA and agmatine. We provide evidence for an underappreciated reaction in arginine metabolism; however, it still remains unclear if N-acetylation of agmatine by an N-acetyltransferase enzyme is biologically relevant. A combination of data provided herein and reported from other labs speaks to its relevancy, warranting further investigation into this chemical transformation as a part of arginine metabolism. Furthermore, we outline a chemical mechanism for the AgmNAT-catalyzed formation of N-acetylagmatine (and, by extension, other N-acylamides), which is consistent with the data presented herein. We also provide evidence for important active site residues involved in substrate binding and maintaining the structural integrity of the active site for efficient catalysis, though further work is necessary to provide more evidence for the dynamic nature of the AgmNAT active site. ## Methods Materials. The AgmNAT gene was codon optimized and synthesized by Genscript. Ambion RETROscript ® Kit, ProBond ™ nickel-chelating resin, and MicroPoly(A) Purist TM was purchased from Invitrogen. Oligonucleotides were purchased from Eurofins MWG Operon. PfuUltra High-Fidelity DNA polymerase was purchased from Agilent. BL21 (DE3) E.coli cells and pET-28a(+) vector were purchased from Novagen. NdeI, XhoI, Antarctic Phosphatase, and T4 DNA ligase were purchased from New England Biolabs. Kanamycin monosulfate and IPTG were purchased from Gold Biotechnology. Acyl-CoAs were purchased from Sigma-Aldrich. Cayman Chemical commercially synthesized N 1 -acetylspermidine. All other reagents were of the highest quality and purchased from either Sigma-Aldrich or Fisher Scientific. AgmNAT: sub-cloning, expression, and purification. AgmNAT was inserted into a pET-28a vector using NdeI and XhoI restriction sites, yielding the final expression vector: AgmNAT-pET-28a, that after transformation into E.coli BL21 (DE3) cells expressed a protein with an N-terminal His 6 -tag followed by a thrombin cleavage site. The E. coli BL21 (DE3) cells containing the AgmNAT-pET-28a vector was cultured using LB media supplemented with 40 μg/mL kanamycin at 37 °C. The culture was induced with 1.0 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at an OD 600 ~ 0.6, followed by an additional four hours at 37 °C. The final culture was harvested by centrifugation at 5,000 × g for 10 min at 4 °C and the pellet was collected. The pellet was resuspended in 20 mM Tris, pH 7.9, 500 mM NaCl, 5 mM imidazole, lysed by sonication, and then centrifuged at 10,000 × g for 15 min at 4 °C. The supernatant was collected and loaded onto 6 mL of ProBond ™ nickel-chelating resin, followed by two wash steps: wash one -10 column volumes of 20 mM Tris, pH 7.9, 500 mM NaCl, 5 mM imidazole followed by wash two -10 column volumes of 20 mM Tris, pH 7.9, 500 mM NaCl, 60 mM imidazole. AgmNAT was eluted in 1 mL fractions using 20 mM Tris, pH 7.9, 500 mM NaCl, 500 mM imidazole, the protein pooled, and extensively dialyzed at 4 °C against 20 mM Tris pH 7.4, 200 mM NaCl. The concentration of AgmNAT was determined using the Bradford assay indexed against BSA as a standard, and purity was assessed by a SDS-PAGE gel (proteins visualized using by Coomassie stain). Purification of recombinant AgmNAT by nickel affinity chromatography yielded pure protein (≥95%) as visualized by SDS-PAGE (Supplementary Fig. S6). AgmNAT crystallography. After nickel-affinity purification, 30 mg of AgmNAT was subjected to dialysis against 50 mM HEPES pH 8.2, 200 mM NaCl, followed by removal of the His 6 affinity-tag using 60 U of biotinylated thrombin for 18 h in a fresh batch of 50 mM HEPES pH 8.2, 200 mM NaCl leaving an unnatural Gly-Ser-His at the N-terminus. The protein mixture was again subjected to nickel-affinity chromatography to remove undigested AgmNAT. AgmNAT was eluted in the 20 mM Tris, pH 7.9, 500 mM NaCl, 60 mM imidazole fraction, whereas the His 6 -AgmNAT was retained on the column until eluted with 20 mM Tris, pH 7.9, 500 mM NaCl, 500 mM imidazole. The biotinylated thrombin was removed by using 3 mL of Pierce monomeric avidin agarose resin at 4 °C for 30 min, followed by centrifugation to recover AgmNAT, and AgmNAT concentrated to ~10 mg/mL by ultrafiltration. Further purification was performed using a HiTrap Q FF column with a linear gradient from 50 mM HEPES pH 8.2 to 50 mM HEPES pH 8.2, 0.5 M NaCl with AgmNAT eluting in fractions containing ~150 mM NaCl. A final SEC purification step was used after the ion exchange step and purified AgmNAT was concentrated to ~8 mg/ml in 50 mM HEPES pH 8.2, 100 mM NaCl for crystallization screening. The Phoenix crystallization robot and Qiagen screening kits were used to evaluate different crystallization conditions for AgmNAT. AgmNAT was crystallized using the hanging-drop vapor diffusion method in 100 mM Tris pH 8.0, 200 mM sodium acetate, 30% PEG 4000. The drop contained a 1:1 ratio of 1 μL of 8 mg/mL AgmNAT with 1 μL of well solution and incubated at 20 °C. Crystals were of elongated rod-shape. Diffraction was measured at the 22-ID-D SER-CAT beamline at the Advanced Photon Source (APS), Argonne, IL. Data were indexed, scaled, and merged with iMosflm using the CCP4 suite 80 . A homology model was constructed based on the AgmNAT sequence using the program SWISS-MODEL 81 with mosquito arylalkylamine N-acetyltransferase (PDB ID 4FD4) 21 as a template for molecular replacement. The molecular replacement program Phaser-MR was used in PHENIX. The models of refinement were first obtained using a rigid-body refinement using phenix.refine in PHENIX. PHENIX 82 and Coot 83 were used to complete the model rebuilding and refinement. For refinement, data was cut at 2.3 A due to relatively poor data quality at higher resolutions. The crystal structure has been deposited into the Protein Data Bank with accession code 5K9N. ## Construction of AgmNAT site-directed mutants. Site-directed mutants of AgmNAT were constructed by the overlap extension method. Using the primers shown in Table S1, each mutant was amplified using pfuUltra High-Fidelity DNA polymerase with the following PCR conditions: initial denaturing step of 95 °C for 2 min, then 30 cycles of 95 °C for 30 s; 60 °C annealing temperature for 30 s; 72 °C extension step for 1 min; then a final extension step of 72 °C for 10 min. Following the amplification of the AgmNAT site-directed mutant, the sub-cloning, expression, and purification procedures are the same as discussed for the wild-type enzyme. Measurement of enzyme activity. Steady-state kinetic constants for AgmNAT were determined by measuring the rate of coenzyme A release using Ellman's reagent (DTNB) at 412 nm (molar absorptivity = 13,600 M −1 cm −1 ) . The assay consisted of 300 mM Tris pH 8.5, 150 μM DTNB, and the desired concentration of acyl-CoA and amine substrates. Initial velocities were measured using a Cary 300 Bio UV-Visible spectrophotometer at 22 °C. Acyl-CoA kinetic constants were evaluated by holding the concentration of agmatine at a constant saturating concentration (5 mM). Amine kinetic constants were evaluated by holding the concentration of acetyl-CoA at a constant saturating concentration (500 μM). The apparent kinetic constants were determined by fitting the resulting data to equation 1 using SigmaPlot 12.0, where v o is the initial velocity, V max,app is the apparent maximal velocity, [S] is the substrate concentration, and K m,app is the apparent Michaelis constant. Each assay was performed in triplicate and the uncertainty for the k cat,app and (k cat /K m ) app values were calculated using equation 2, where σ is the standard error. x 2 y 2 Kinetic mechanism and inhibitor analysis. Defining the kinetic mechanism of AgmNAT was accomplished by evaluating double reciprocal plots of initial velocity data for acetyl-CoA and agmatine, followed by determining the type of inhibition for substrate analogs used as dead-end inhibitors or N-acetylagmatine for product inhibition. Initial velocities were determined by varying the concentration of one substrate, while holding the other substrate at a fixed concentration. Acetyl-CoA was evaluated at 20, 50, 100, 250 and 500 μM, whereas agmatine was evaluated at 60, 300, 750 and 1500 μM. The resulting initial velocity data was fit to equation 3 for an ordered Bi-Bi mechanism and equation 4 for a ping pong mechanism using IGOR Pro 6.34 A, where v o is the initial velocity, V max is the maximal velocity, K ia is the dissociation constant for substrate A, K b is the Michaelis constant for substrate B, K a is the Michaelis constant for substrate A, [A] is the concentration of substrate A, and [B] is the concentration of substrate B. Inhibition experiments by either substrate analogs or N-acetylagamatine were used to discriminate between an ordered, random sequential, or ping pong kinetic mechanism. Oleoyl-CoA, arcaine, and L-arginine methyl ester were used as dead-end inhibitors for AgmNAT while N-acetylagmatine was used for product inhibition. Initial velocity patterns were generated by varying the concentration of one substrate, holding the other substrate concentration at its apparent K m , and changing the concentration of inhibitor for each data set in triplicate. The resulting data was fit to equations 5-7, for competitive, noncompetitive, and uncompetitive inhibition respectively using SigmaPlot 12.0. For equations 4-6, v o is the initial velocity, V max,app is the apparent maximal velocity, K m,app is the apparent Michaelis constant, [S] is the substrate concentration, [I] is the inhibitor concentration, and K i is the inhibition constant. ## Rate versus pH. The pH-dependence on the kinetic constants for acetyl-CoA was determined using intervals of 0.5 pH units, ranging from 6.5-9.5. Buffers used to measure the pH-dependence were MES (pH 6.5 and 7.0), Tris (pH 7.0-9.0), AmeP (pH 9.0 and 9.5). The resulting data were fit to equations 8 (log (k cat /K m ) app -acetyl-CoA and equation 9 (log k cat,app -acetyl-CoA ) to determine the apparent pK a values using IGOR Pro 6.34 A, where c is the pH-independent plateau. The wild-type enzyme is reported in triplicate, whereas the E34A mutant was evaluated in duplicate. Agmatine. To a solution of putrescine (2.0 g, 22.7 mmol) in water (20 mL) was added 2-methylisouronium sulfate (2.7 g, 11 mmol). The mixture was heated to 50 °C for 6 hours, then cooled in an ice bath for 30 minutes. During this time, a white precipitate was formed, which was collected by filtration, and then washed with ice water to give agmatine (1.3 g, 44%) as a white solid that was used without further purification. 1 H NMR (500 MHz, D 2 O) δ 3.08 (t, J = 6.0 Hz, 2 H), 2.81 (t, J = 6.8 Hz, 2 H), 1.53 (br. s., 4 H) ppm. 13 N-Acetylagmatine. To a mixture of agmatine (1.0 g, 7.62 mmol) in pyridine (10 mL) was added acetyl chloride (542 μL, 7.62 mmol) dropwise. The mixture was allowed to stir at room temperature for 4 hours, then was concentrated on a rotary evaporator. The crude residue was adsorbed onto silica gel and purified by flash column chromatography (methylene chloride/methanol 19:1) to give N-acetylagmatine (400 mg, 30%) as a viscous, colorless oil. 1

chemsum

{"title": "Structural and Mechanistic Analysis of Drosophila melanogaster Agmatine N-Acetyltransferase, an Enzyme that Catalyzes the Formation of N-Acetylagmatine", "journal": "Scientific Reports - Nature"}

hollowing_out_mofs:_hierarchical_micro-_and_mesoporous_mofs_with_tailorable_porosity_via_selective_a

## Abstract: We report a new strategy for the synthesis of robust hierarchical micro-and mesoporous MOFs from water stable MOFs via a selective acid etching process. The process is controlled by the size-selective diffusion of acid molecules through the MOF windows. This method enables the fine-tuning of the porosity of hierarchical MOFs, allowing for the generation of well-defined mesopores with high mesopore volume.Because of the size-selective diffusion of acid molecules, the inherent crystallinity and external morphology of the resulting MOFs are well-maintained after acid treatment. This novel strategy may provide an alternative route towards the synthesis of diverse hierarchical MOFs. ## Introduction Metal-organic frameworks (MOFs) are crystalline porous materials constructed via the self-assembly of metal ions (or clusters) and organic ligands, 1 which have widespread applications in many felds owing to their processability, structural flexibility, and well-defned pores with large surface areas. 2 Such properties are often directly related to the size, geometry and accessibility of the pores. 3 Microporous MOFs (pore size < 2 nm) possess high surface area with structural selectivity for small molecules, but often have limited applicability because of the difficulties in mass transfer and encapsulation of functional large guest molecules. 4 To overcome these limitations, mesoporous MOFs (2-50 nm) have become a subject of great interest. 5,6 However, mesoporous MOFs often collapse upon the evacuation of guest molecules and the pore sizes of the MOFs produced by conventional solvothermal methods are usually smaller than 5 nm. 7,8 Recently, hierarchical micro-and mesoporous MOFs have emerged as a promising alternative combining the advantages of both micro-and mesoporosity. 9 For example, mesopores facilitate the mass transfer process and the concomitant micropores offer high surface area. Furthermore, hierarchical micro-and mesoporous MOFs can enhance catalytic activities, and have been utilized as hosts for large guest molecules (e.g., enzymes). 10 Therefore, several strategies have been developed to construct hierarchical porous MOFs via crystal growth as well as post-synthetic procedures, such as imperfect crystallization, gas-expanded liquid, template-assisted, modular induced, and calcination methods. 11 Although these strategies are novel and inventive, they often involve lengthy synthetic procedures and fne control of the pore sizes with high structural stability via a single step procedure remains a signifcant challenge. Recently, we reported a simple hydrolytic method to synthesize a hierarchical micro-and mesoporous MOF using a microporous MOF (POST-66). 12 However, this method is only applicable to MOFs with low water stability and the fne-tuning of pore size still needs to be addressed. Here, we report a novel strategy for the synthesis of water-stable hierarchical porous MOFs by a selective acid etching process (Scheme 1). This method not only allows fne tuning of the porosity, but also preservation of the inherent crystallinity and external morphology of the resulting MOFs due to the selectivity of the Scheme 1 Illustration of the etching process for MIL-100(Fe). Left: MIL-100(Fe) crystal with hexagonal and pentagonal windows; middle: acid diffusion into tetrahedral channels through hexagonal windows; right: resulting mesopores after etching. etching process. Although we demonstrate the principle mainly with MIL-100(Fe) 1,3,5-tricarboxylate), this strategy can be extended to the synthesis of other water-stable hierarchical porous MOFs. ## Results and discussion At the outset of this work, we thought that we could exploit the window dimensions of MIL-100(Fe) to allow selective acid etching of the MOF due to the fact that MIL-100(Fe) has large and small cages with hexagonal and pentagonal windows (d ¼ 0.89 and 0.49 nm, respectively) (Fig. 1a and S1 ‡). 13 Therefore, if we employed an appropriately sized inorganic acid as an etching agent, it may diffuse into MIL-100(Fe) through the hexagonal windows but not the pentagonal windows allowing a selective etching process, while retaining the overall crystallinity. Considering these points, we chose phosphoric acid (H 3 PO 4 , d ¼ 0.61 nm) 14 in N,N-dimethylformamide (DMF) as an etching agent, which exhibits size-selective diffusion into the 3D channels of the large cages through the hexagonal windows at room temperature. The etching process then takes place after raising the temperature (Scheme 1 and Fig. S2 ‡). With this idea in mind, we synthesized MIL-100(Fe) following the literature procedure. 13 Subsequently the dehydrated MIL-100(Fe) powder was soaked in the H 3 PO 4 solution at room temperature under sonication for enhanced diffusion, and then heated to 70 C for the etching. The concentration of the acid solution was varied (0 to 80 mM) in order to control the degree of etching. Transmission electron microscopy (TEM) images (Fig. 1b and c) showed that the acid treated sample (MIL-100(Fe)-80, treated by 80 mM H 3 PO 4 solution) has a sponge-like morphology and much higher transparency compared to pristine MIL-100(Fe) because of the enlarged pore volume. Scanning electron microscope (SEM) images (Fig. 1d-h) indicated that increasing the concentration of H 3 PO 4 (0 to 80 mM) resulted in regular enlargement of the pore sizes on the crystal surface while maintaining the morphology of MIL-100(Fe) (crystal size and shape). Also, a series of powder X-ray diffraction (PXRD) profles (Fig. S4a ‡) confrmed the maintenance of the original crystallinity and energy-dispersive X-ray spectroscopy (EDS) analyses showed that the atomic percentage of Fe 3+ ions was constant (Fig. S5 ‡). Moreover, there was no change in the oxidation state of the Fe(III) as shown by the X-ray photoelectron spectroscopy (XPS) spectra (Figure S4b ‡). The porosities of the acid treated MIL-100(Fe) series were investigated by N 2 sorption measurements. The gradual change of the adsorption-desorption isotherms from microporous type I to mesoporous type IV (Fig. 1i and S7 ‡) indicates the partial loss of micropores and the generation of mesopores (Fig. S3 ‡). 15 Additionally, the sorption isotherm has a H1 type hysteresis loop, which is characteristic of well-defned pore channels. 16 Owing to the increasing void space, the N 2 sorption capacity of MIL-100(Fe)-80 reaches up to 750 cm 3 g 1 and the mesopore volume of MIL-100(Fe)-80 is 1.15 cm 3 g 1 (Table S1 ‡). These values are comparable to that of an ultra-high mesoporous MOF. 10 The mesopore sizes, calculated from Barrett-Joyner-Halenda (BJH) analysis, gradually increased from 2.4 to 18.4 nm (Fig. 1j and S8 ‡) and had a narrow pore size distribution with high differential pore volume. Acid concentrations above 70 mM resulted in the broadening of the pore size distribution profles. Furthermore, the mesopore sizes of the acid-treated MIL-100(Fe) can also be fne-tuned by adjusting the treatment time with a fxed H 3 PO 4 concentration (Fig. S9 ‡). With the increase of acid treatment time from 2 to 12 h, the mesopore size gradually increased from 3.3 to 18.4 nm (Fig. S10 ‡). During the etching process, Fe 3+ ions and BTC ligands (constituents of MIL-100(Fe)) gradually leached from the frameworks and the leaching amount of both components was proportional to the acid concentration, as confrmed by inductively coupled plasma-atomic emission spectrometry (ICP-AES) and UV-Vis spectroscopy (Fig. S12 ‡). To investigate the etching process in more detail, the local structures of the hierarchical porous MOFs were studied by small-angle X-ray scattering (SAXS) and high resolution TEM (HR-TEM). The series of acid-treated MIL-100(Fe) samples had identical peaks in the SAXS spectra (Fig. 2a), suggesting retention of the crystallinity. The full width at half maximum (FWHM) of the diffraction peaks, (220), ( 311), ( 222) and (400), slightly increased at higher acid concentrations (Fig. S13 ‡), implying successful pore enlargement, 17 which is consistent with the power law scattering of Q 4 in the SAXS profles (typical scattering behaviour of mesoporous MOFs, Fig. S14 ‡). 18 In particular, the intensity of the (220) peak which corresponds to the lattice plane on the boundary between the large and small cages signifcantly decreased from MIL-100(Fe) to MIL-100(Fe)-80 (Fig. 2b-d). Due to the size-selective diffusion of H 3 PO 4 , the etching process probably proceeded along the large cages on the (220) plane, causing the signifcant intensity decrease of the (220) peak. Hence, the secondary building units (SBUs, Fig. 2e) of the large cages on the (220) lattice plane disassembled gradually and the large and small cages merged to form expanded pores. Additionally, the HR-TEM images of MIL-100(Fe)-40 and 80 (Fig. 3a-d) with clear lattice fringes of orientation (1-11) (d ¼ 3.68 and 3.74 nm, respectively), demonstrated that higher acid concentration resulted in a greater degree of etching. This observation indicates that the crystalline structure is maintained after the etching process and that the mesopore transformation occurred with a preferential direction along the zone axis, in accordance with the SAXS analysis. Moreover, the enlarged lattice fringe orientated along the (1-11) lattice plane (d ¼ 7.49 nm, Fig. 3d) corresponds to a doubling of the (1-11) dspacing value (blue lines, Fig. 3e). The enlargement was probably caused by the merged cages derived from etching the SBUs in between the large and small cages on the (220) plane. This clearly supports the idea that the mesopore enlargement process occurred in a structurally selective manner. Consequently, the observed etching process can be explained as follows (Scheme 1). Due to the size difference between the hexagonal and pentagonal windows, H 3 PO 4 selectively diffuses into the 3D tetrahedral channels of the large cages through hexagonal windows at room temperature (Scheme 1 and Fig. 3f). After activation (heating at 70 C), the SBUs of the large cages near the ( 220), ( 022) and ( 202) lattice planes (boundaries between the large and small cages, SAXS, Fig. 2) are gradually etched, resulting in the disassembly of the [(Fe 3 (m 3 -O))(OH)(H 2 O) 2 ] 6 and BTC linker (ICP-AES and UV-Vis spectra, Fig. S12 ‡). This disassembly leads to the merging of the small and large cages, while the majority of the walls of the small cages survive to maintain the crystallinity and the structural integrity. Finally, the local structure collapses and the mesopores are generated (TEM data, Fig. 3). To further validate the size-selective acid diffusion of the etching process, sulfuric acid (H 2 SO 4 , d ¼ 0.66 nm (ref. 19)), which has a comparable size to H 3 PO 4 , and hydrochloric acid (HCl, d ¼ 0.34 nm (ref. 20)) were employed as etching agents. In the case of H 2 SO 4 , the mesopore size of hierarchical porous MIL-100(Fe) can also be tuned in a controlled manner (Fig. S15 ‡). However, hydrochloric acid, because of its smaller molecular dimension, can easily diffuse into both the hexagonal and pentagonal windows (loss of size selectivity), resulting in the collapse of the whole framework of MIL-100(Fe) (dissolution in the acid solution). Preliminary studies have suggested that this strategy can be applied to other water-stable MOFs which have a structural selectivity for H 3 PO 4 (e.g., soc-MOF and MIL-88A). As demonstrated by the structure-property data and N 2 sorption data shown in Fig. S16, S17, S19 and S20, ‡ the porosity of soc-MOF and MIL-88A can also be controlled in a range under 20 nm by adjusting the acid concentration. Besides, the crystallinity and morphology is still retained after acid treatment as evidenced by XRD patterns and TEM images (Fig. S18, S21 and 22 ‡). ## Conclusions We have developed a general strategy to prepare hierarchical micro-and mesoporous MOFs from water-stable MOFs (MIL-100(Fe), soc-MOF, and MIL-88A). This work demonstrated that: (i) the size-selective acid diffusion strategy is a versatile method to control the etching process; and (ii) control of the acid concentration and the treatment time can produce hierarchical MOFs with the desired pore size dimensions while maintaining the original microporosity and structural stability. This simple strategy may provide an alternative route towards the synthesis of tailor-made hierarchical MOFs and hold enormous promise for facilitating the development of MOF-based materials with interesting properties. Work along this line is currently underway in our laboratory. ## General information All the reagents and solvents were commercially available and used as supplied without further purifcation. ICP-AES (IRIS Intrepid II XSP, Thermo Electron Corporation) was used for the analysis of metal ion concentrations. UV-Vis absorption spectra were collected by an Agilent Cary 5000 UV-Vis-NIR Spectrophotometer. TEM, HR-TEM and STEM images were measured using a JEOL JEM-2200FS with image Cs-corrector equipped (National Institute for Nanomaterials Technology (NINT), Korea). SEM images were collected by a JSM 7800F PRIME scanning electron microscope operating at 1 kV. Powder XRD patterns were obtained on a Rigaku Smartlab system equipped with a Cu sealed tube (wave length ¼ 1.54178 ) and a vacuumed high-temperature stage (Anton Paar TTK-450). The following conditions were used: 40 kV, 30 mA, increment ¼ 0.01 , and scan speed ¼ 0.3 s per step. NMR data were recorded on a Bruker DRX500 spectrometer. Small-angle X-ray scattering (SAXS) measurements were carried out using the 4C SAXS II beamline (BL) of the Pohang Light Source II (PLS II) with 3 GeV power and an X-ray beam wavelength of 0.734 at the Pohang University of Science and Technology (POSTECH), Korea. The magnitude of the scattering vector, q ¼ (4p/l) sin q, was 0.1 nm 1 < q < 6.50 nm 1 , where 2q is the scattering angle and l is the wavelength of the X-ray beam. All scattering measurements were carried out at 25 C. ## Gas adsorption measurements All gas sorption isotherms were measured at 77 K with BELSORP-mini volumetric adsorption equipment. Typically, a sample of as-synthesized material ($100 mg) was loaded and, prior to the measurements, residual solvents were exchanged with EtOH for 3 days, and then evacuated by heating to 200 C under a high vacuum (10 2 Pa). Mesoporous transformation procedure of MIL-100(Fe) MIL-100(Fe) was synthesized following literature procedure. 13 As-synthesized MIL-100(Fe) was frst dehydrated and then soaked in DMF (6 mL) with different amounts of phosphoric acid (89 wt%, TCI, Japan) to give solutions with varied acid concentration. All samples were sonicated for 10 minutes at room temperature and then kept at 70 C in an oven. For concentration controlled etching, the treatment time was kept constant at 5 hours. While for treatment time controlled etching, the treatment time was varied from 2 to 12 hours in 40 mM H 3 PO 4 . After the acid etching was completed, the crystalline solid materials were washed with DMF and EtOH 3 times each and dried under vacuum overnight.

chemsum

{"title": "Hollowing out MOFs: hierarchical micro- and mesoporous MOFs with tailorable porosity via selective acid etching", "journal": "Royal Society of Chemistry (RSC)"}

assessing_potential_profiles_in_water_electrolysers_to_minimise_titanium_use

## Abstract: The assumption of a highly corrosive potential at the anode bipolar plate (BPP) and porous transport layer (PTL) in a proton exchange membrane water electrolyser (PEMWE) stack often leads to selection of expensive materials such as platinum-coated titanium. Here, we develop a physicochemical model of electrochemical potential distribution in a PTL and validate it by in situ and ex situ electrochemical potential measurement. Model predictions suggest that, under typical PEMWE operating conditions, the corrosive zone associated with the anode polarisation extends only E200 mm into the PTL from the catalyst layer, obviating the need for highly corrosion-resistant materials through the bulk of the PTL and at the BPP. Guided by this observation, we present single cell PEMWE tests using anode current collectors fabricated from carbon-coated 316L stainless steel. The material is shown to be tolerant to potentials up to 1.2 V vs. RHE and when tested in situ for 30 days at 2 A cm À2 showed no evidence of degradation. These results strongly suggest that much of the titanium in PEMWEs may be substituted with cheaper, more abundant materials with no loss of electrolyser performance or lifetime, which would significantly reduce the cost of green hydrogen. The combined modelling and experimental approach developed here shows great promise for design optimisation of PEMWEs and other electrochemical energy conversion devices. Broader contextComplete decarbonisation of the heating and transportation sectors is required in the next 25 years to limit global warming to below 2 1C, and preferably to 1.5 1C. Large scale energy storage is key to this goal and, while advances in Li-ion battery technology have enabled their use in passenger vehicles and short term electricity storage, alternative energy vectors such as hydrogen are more suited to heavy duty transport and seasonal energy storage for heating. Production of 'green hydrogen' by the electrolysis of water is a promising solution but is currently hampered by the high cost of the technology. Operating costs will continue to fall with decreasing renewable electricity prices so the main challenge is to reduce the capital cost of water electrolysers. This work points the way towards cost reduction of proton exchange membrane water electrolyser stacks by establishing the underpinning theory and experimental validation required to inform new design strategies for replacement of expensive component materials with cheaper alternatives. ## Introduction Proton exchange membrane water electrolysers (PEMWEs) use renewable energy to produce green hydrogen, a key energy vector for advanced economies trying to achieve COP21 targets. The cost of green hydrogen is still high in comparison to both grey and blue hydrogen and achieving cost competitiveness in multiple sectors, expected at prices of EUSD 2 kg H 2 1 , remains a challenge. 1 Much research is being performed to decrease the cost of hydrogen produced by PEMWE, often with a focus on reducing the cost of electrocatalysts 2 or optimising cell design to increase performance. 3 A range of reports indicate that, regardless of manufacturing scale, bipolar plates (BPPs) and porous transport layers (PTLs) are among the most expensive components in a PEMWE stack. Reducing the cost of these components is therefore key to meeting Hydrogen Europe's price target of h1000 per kg d 1 of H 2 production by 2030. 7 The BPP and PTL are critical components of the stack as they perform multiple functions, including distribution of electrical current and heat, mechanical support for the stack and catalyst coated membrane (CCM), and flow distribution to ensure effective mass transport to and from the catalytic sites. At the anode side of the electrolyser, the highly oxidising potential of the catalyst layer (CL) and the presence of molecular oxygen make corrosion resistance a required property for both BPPs and PTLs if they are to operate effectively for the 4100 000 h expected of a state-of-the-art PEMWE. 8 Titanium is the conventional material of choice for the PTL and BPP since it readily forms a passivating oxide layer in contact with oxygen, which protects the bulk metal from corrosion. Unfortunately titanium is comparatively expensive to produce 6 and machine. 9 Titanium has also been identified by the European Union as a critical raw material, highlighting a potential risk to supply. 10 Furthermore, titanium oxides have an unacceptably high contact resistance and have been reported to corrode and contaminate the CCM after operation for 1000 h under constant current or current cycling operation. 11,12 Titanium is therefore typically coated with thin layers of more noble metals such as platinum, 13 gold, 14 and/or iridium, 9,15 which reduce interfacial contact resistance (ICR) and mitigate titanium corrosion. 13 Some alternative materials have been shown to degrade in the harsh environment of the anode and shorten PEMWE lifetime: for example, a carbon and 316L stainless steel-based PTL experienced significant corrosion after being operated for a relatively short period. If 316L is to be used as the PTL, an additional corrosionresistant coating is needed. These limitations on the application of cheaper materials derive only from investigations in which the PTL is substituted in its entirety. The requirement for a highly corrosion-resistant material for the BPP is based on the commonly-held assumption that a uniformly aggressive corrosive environment applies throughout the anode electrolyte. This view is often reinforced by the results of ex situ testing at elevated potential, under which conditions many materials experience significant corrosive attack. Conversely, the authors' recent report of in situ reference electrode measurements made at the PTL-current collector interface in a single cell (corresponding to the PTL-BPP interface in a stack) concluded that, due to the high resistivity of the PTL aqueous phase, the BPP remains at its open circuit potential (OCP), even when the cell voltage is 42 V. 23 This implies a 'decoupling' of the BPP from the highly anodically polarised environment of the CL and raises the possibility that the BPP, and possibly some portion of the PTL more distant from the anode CL, can be safely substituted with cheaper materials. In this work, we investigate the practicality of this suggestion by exploring the spatial profile and magnitude of the electrochemical potential (hereafter referred to simply as 'potential') through the thickness of the anode PTL, using theoretical modelling and in situ reference electrode measurements. Having predicted a much less aggressive environment at the BPP/PTL interface, we identify carbon-coated 316L stainless steel (C-316L) as a promising candidate material for replacing platinum-coated titanium (Pt-Ti), and evaluate it as the current collector material in a single cell PEMWE operated for 30 days. Due to the use of a single cell in the experimental part, the term current collector (CC), rather than BPP, will be used hereafter in this work when referring to the components behind the PTL (i.e. monopolar end plate and expanded titanium mesh) in the context of specific measurements reported. Systematic postmortem and in situ measurements reveal no degradation of the C-316L samples during testing, encouraging further long-term, full-scale testing of this low-cost material. ## Potential profile across anode components The potential of the anode CL in an operating PEMWE is necessarily in excess of 1.2 V vs. RHE in order to drive the oxygen evolution reaction. This is the reason that anode catalysts are limited to very oxidation-stable materials such as iridium oxide. However, it does not automatically follow that the potential of all components in the anode is also in excess of 1.2 V vs. RHE. While the CL, PTL and CC are all connected together with negligibly small electrical resistance, the ionic resistance between different parts of the anode may be high due to the low conductivity of the feed water. It is a combination of these resistances, with other parameters, that determines the extent to which the potential is dominated by local electrochemical reactions (such as corrosion processes) vs. those occurring in the CL (oxygen evolution). If the PTL aqueous phase were highly conductive, corrosion of the PTL material could proceed with the cathode CL (hydrogen evolution reaction) acting as counter electrode, because current density can easily be sunk from the PTL through the proton exchange membrane (PEM). Conversely, for a highly resistive PTL aqueous phase, the ionic resistance prevents the cathode CL from supporting corrosion of the PTL material -the available cathodic area is the PTL itself. The latter case is the situation of a corrosion potential, as would be encountered in an ex situ experiment where the material is unpolarised, and so the local potential would be expected to be close to its open circuit value. We describe the PTL potential in this case as being 'decoupled' from the anode CL. The water in a state-of-the art operating PEMWE is highly deionised, with a conductivity o1.0 mS cm 1 (typical specifications are ASTM Type II water or ISO 3696:1987 Grade 2), because the membrane and ionomers used as proton conductors in PEMWEs are vulnerable to foreign cation poisoning. The finite water resistivity means that the extent of local anodic polarisation of the PTL due to the applied cell voltage varies spatially, decreasing further away from the anode CL. This raises two related questions for the rational design of PTL geometry and selection of materials. First, can the potential at the CC be predicted theoretically as a function of the PTL properties? Second, for a given PTL, can the length scale be determined, theoretically and/or experimentally, over which the ## Paper Energy & Environmental Science local potential transitions from the highly oxidative conditions of the anode CL to the more benign OCP of the CC? To address the first question, a lumped theoretical physicochemical model was developed to describe the extent of decoupling between the anode CL and CC, using a ''decoupling coefficient'' suitable for situations when the PTL is made of the same material as the CC. The derivation and form of the decoupling coefficient are given in Supplementary Note 1 (ESI †). This model shows that the extent of decoupling is proportional to the electrolyte resistivity; it also depends on PTL geometry (electrochemically active surface area, porosity, tortuosity) as well as the kinetic properties of the electrochemical reaction that determines the OCP of the CC material in the electrolyte, described here simply with an exchange current density. When evaluated with suitable input parameters (Table S1, ESI †), the model prediction agrees with the previous observation of extensive decoupling under typical operating conditions. 23 To evaluate the potential profile through the PTL, a simple physicochemical model was developed and parameterised using realistic, experimentally established values. Key assumptions are described briefly in the ''Methods'' section, with full mathematical detail and the chosen parameter values described fully in Supplementary Note 2 and Tables S2 and S3 (ESI †). Predicted potential profiles are shown in Fig. 1(a) parameterised for realistic PEMWE operating conditions of 2 A cm 2 at 60 1C. Various water conductivities are compared, ranging from that of Type I deionised water up to that of pH 1 H 2 SO 4 (aq) as an unrealistic upper limit. The results illustrate how the zone of appreciable anodic polarisation within the PTL is confined more closely to the CL as electrolyte resistivity increases. Experimental validation of this trend is presented in Fig. S1 and S2 (ESI †). For the more realistic case of deionised water, the partial thickness of the PTL that is polarised to 100 mV more positive than its OCP is only E70 mm for Type I water and only E300 mm for Type II water. Finally, the model was used to predict the extent of decoupling expected in a realistic worst-case scenario of a laboratory experimental cell containing air-contaminated Type I water. Here the water equilibrates with CO 2 from the ambient air and therefore has a measured conductivity of 20 mS cm 1 at 60 1C (Fig. 1(b)), which is above the recommended acceptable water quality conductivity standard for PEMWEs of 1 mS cm 1 (ASTM Type II water). 24 The model prediction is shown in Fig. 1(b) (dashed line) and demonstrates that the potential of the CC is determined almost solely by its OCP under these operating conditions. The parameterisation used here assumes that all components are Pt-Ti, which from experiment exhibited an OCP of 1.0 V vs. RHE in oxygen-saturated water at 60 1C (Fig. 2(a)). Note that this OCP is material-dependent so the potentials shown do not constitute predictions for all scenarios. Parameterisation of this model relied on a number of parameters for which experimentally determined information was limited, so a predicted range is shown to indicate the sensitivity of the model. Previous experimental evidence has shown decoupling of the CC from the CL, but has not resolved the potential profile closer to the CL, i.e. within the PTL. The model predictions were therefore experimentally validated using an operating single cell PEMWE fitted with four in situ reference electrodes at varying depths through the PTL. The in situ reference electrode was a reversible hydrogen electrode (RHE) immersed in 0.5 M H 2 SO 4 connected to the water inside the operating cell via an insulated Nafion tube; the working principle of the in situ reference electrode has been established by prior work 23,25 and is described further in the Methods section as well as in Supplementary Note 1 and Fig. S3 (ESI †). Each Nafion tube was inserted into a hole of a defined depth in the PTL and was therefore held at a known position relative to the CL as shown in Fig. 1(c). The respective distances of the tips of the reference electrodes from the CL, as measured using XCT (Fig. 1(d)), were: 0 mm (drilled through the PTL completely), 290 mm, 600 mm, and 1090 mm. The potential difference measured between the anode electrode/CC/PTL and the reference electrode is dominated by the potential of the material in the proximity of the reference electrode tip and in this manuscript is referred to as E loc . Fig. 1(b) shows the predicted and measured E loc as a function of PTL thickness in air-contaminated Type I water; corresponding comparisons of predicted and measured E loc at different depths are shown for more conductive, less realistic solutions (0.05 mM H 2 SO 4 and 0.5 mM H 2 SO 4 ) in Fig. S1 and S2 (ESI †). Error bars are included to show systematic error on the measured values; this is estimated to be AE 100 mV, which we attribute principally to uncorrected liquid junction potentials associated with proton leakage from the acidified Nafion tip. Fig. S4 (ESI †) shows an ex situ experiment demonstrating that the magnitude of the liquid junction potential is of the order of 100 mV. Good agreement is observed between the predicted and measured values, with the results in Fig. 1(b) together with Fig. S1 and S2 (ESI †) clearly showing that the potential drops from that observed at the CL (E1.7 V vs. RHE) to E1.15 V vs. RHE within 300 mm in the case of the air-contaminated Type I water. The potential further decreases to a value typical of the OCP of Pt/O 2 (E1 V vs. RHE) at distances 4600 mm from the anode CL. The measurements in Fig. S1 and S2 (ESI †) also validate the model predictions that the extent of the potential penetration increases with increasing electrolyte ionic conductivity. It is noted that the local potential measured at the PTL/ CL interface is anomalously high, being close to the cell voltage. This is attributed to a combination of electrode shielding impacting the behaviour of the CL at the interface and pH effects on the liquid junction potential; it does not affect our confidence in the measured values within the PTL. From the physicochemical model and in situ reference electrode experiments we conclude that, for a PEMWE stack operating at typical conditions of o1.0 mS cm 1 water conductivity, 2.0 A cm 2 current density, and 60 1C, the anode BPP and a large part of the PTL will experience a potential close to their OCP in oxygen-saturated water at 60 1C. This raises the commercially significant possibility of replacing a large proportion of the coated titanium materials used in PEMWEs with cheaper alternatives. ## Ex situ evaluation of alternative BPP materials The physicochemical model introduced above predicts that, under typical operating conditions, the CC material remains largely at its OCP, as does the PTL at sufficient distances from the CL. For a stable material in oxygen-saturated solution, the OCP is determined by the metal/metal oxide redox couple and/ or the oxygen reduction reaction and is typically in the range 0.5-1.2 V vs. RHE, depending on the materials and conditions. Even these potentials are relatively oxidising, with most metals unstable with respect to oxidation. Carbon, however, is a possible candidate. While carbon is thermodynamically unstable at potentials above E0.2 V vs. RHE, it has been shown to be highly kinetically stable, with many highly graphitic carbons used in oxidising environments such as catalyst supports at the cathode of PEM fuel cells. 26 Furthermore, as carbon does not form an oxide layer it maintains a low ICR at high potentials. 27 Ready alternatives are carbon-coated stainless steels, which are ideally suited to volume manufacturing through stamping. These are cheaper than titanium and consequently are sold as BPPs in a range of fuel cell applications. 27,28 To evaluate the suitability of C-316L as a BPP material, a range of ex situ electrochemical experiments were performed. Samples were prepared using a high throughput physical vapor deposition technique that is scalable to high volume roll-to-roll production. Initially the OCP of C-316L was evaluated in oxygen-saturated dilute aqueous sulphuric acid (pH 4.5) simulating a 'worst case' electrolyser system (Fig. 2(a)). Here the OCP is E0.5 V vs. RHE, equivalent to that of uncoated 316L, suggesting that the carbon coating is sufficiently thin and porous that the OCP of the C-316L is governed largely by the underlying metal. In an operating cell, however, the CC is in electrical contact with the PTL. When this is simulated ex situ by galvanically coupling the C-316L and the Pt-Ti samples, the OCP is shifted to E0.90 V vs. RHE. An Evans diagram (Fig. S5, ESI †) shows that Pt exhibits much faster oxygen reduction kinetics compared to the oxidation kinetics of C-316L and this is responsible for pulling the mixed OCP more positive. It should be recognised that OCP can depend significantly on the mechanical and chemical exposure history of the surface as well as the material itself; here, we emphasise the observation that the carbon coating has a negligible effect, whereas galvanic coupling to a Pt-coated surface can displace the OCP to close to 1 V vs. RHE. An accelerated stress test was performed to assess the stability of the carbon coating. Coated samples were polarised to potentials between 1.0 V and 1.6 V vs. RHE with the current monitored (Fig. 2(b)), and changes in the coating thickness were evaluated by Auger electron spectroscopy (AES) depth profiling after the tests (Fig. 2(c)). These experiments were carried out under conservative conditions, with an identical coating used on 304L stainless steel (C-304L), in a more corrosive electrolyte, specifically oxygen-saturated aqueous H 2 SO 4 (pH 3) at 70 1C with 0.1 ppm HF added. HF was included to simulate plausible Nafion degradation products. 304L stainless steel is chosen for conservatism as it is recognised as a less corrosion-resistant alloy than the 316L stainless steel intended for use in the BPP; comparative dissolution properties of 304L and 316L stainless steels have been studied previously in electrolyser-relevant environments. 29 A significant corrosion current and measurable loss in thickness are only detected at potentials Z1.4 V vs. RHE, with almost complete removal of the carbon coating observed at 1.6 V vs. RHE (Fig. 2(c) and Fig. S6, ESI †). This suggests that carbon-coated stainless steels should be stable at the surface of the BPP, where E loc would be between 0.9-1.0 V vs. vs. RHE, but not at the CL where the potential is 41.5 V vs. RHE; this is compatible with the previous reported degradation of a PTL made entirely of carbon. 14 Contact resistance measurements (Table S4, ESI †) show a low ICR of 2.5 AE 0.2 mO cm 2 , meeting the 2020 DoE PEM fuel cell BPP ICR target of o10 mO cm 2 . 30 The ICR value does not change for the sample held at 1.0 V vs. RHE, but increases slightly at higher potentials. ## In situ evaluation of BPP materials Based on the promising theory and results presented above, single cell PEMWE tests (Fig. 3) were performed with a C-316L anode CC to assess the short-term stability of a real cell with this material substitution. The cell voltage and E loc of the C-316L anode CC are observed to remain decoupled throughout a 7 day test (Fig. 3(a)) with the E loc of the C-316L anode CC maintaining a steady value of E1 V vs. RHE over the entire testing period. This is consistent with the simulation predictions. Similar behaviour is observed with a bare 316L anode CC (Fig. S7, ESI †). Irrespective of the CC material (C-316L, 316L, and Pt-Ti), a similar rate of cell voltage degradation (0.44-0.66 mV h 1 ) is observed in the 7 day tests (Fig. 3(a) and Fig. S7a, ESI †), indicating a degradation mechanism arising from suboptimal CCM design rather than choice of CC material. A further 30 day single cell test was conducted using C-316L as the anode CC material, this time without an in situ reference electrode. No significant degradation is identified, as indicated by the stable cell voltage (Fig. 3(b)) and polarisation curves at the beginning and end of the 30 day test (Fig. 3(c)), which show no significant increase in cell resistance. Post-mortem analysis was performed on the C-316L material. An element-by-element comparison of the AES depth profiles of all samples (as-received and after the 30 day test) reveals no oxidation of the C-316L substrate or loss of carbon thickness, though the testing did result in minor contamination by adventitious carbon (Fig. S8b and S9, ESI †). Post-mortem analysis of the CCM by acid digestion and inductively-coupled ## Energy & Environmental Science Paper plasma -sector field mass spectrometry (ICP-SFMS) shows no build-up of typical elements present in 316L (Cr, Mn, Fe) even after 30 days of electrolyser operation (Table S5, ESI †). Furthermore, semi-quantitative ICP-MS analysis of the anode and cathode feed water before and after the 30 day test demonstrates that all elements remain at the E1 ppb level or lower (Table S5, ESI †), confirming the absence of meaningful corrosion of the 316L during the test. Interestingly, ruthenium contamination of the CC surface is observed for both C-316L (Fig. S8a, ESI †) and Pt-Ti (Fig. S8c, ESI †). This is due to the use of mixed iridiumruthenium oxides in the anode catalyst of the CCMs used in this work; ruthenium is well known to be unstable under anode CL conditions resulting in dissolution. Redeposition of the ruthenium on the surface of the CC provides further indirect evidence of the more negative potential at this location. While all of the experimental results reported here point strongly to the utility of C-316L as a stack BPP material, an experimental duration of 30 days is negligible compared to targeted PEMWE stack lifetimes of 4100 000 h. 8 The theoretical treatment has also focused only on the potential under idealised steady-state operating conditions and has not considered the influence of transient conditions or the possibility of enhanced corrosion in the presence of trace chemical species such as H 2 O 2 , radical species, or HF, which may facilitate slow but significant corrosion processes. Experimental testing of the material in a larger stack configuration for many thousands of hours is therefore required; these experiments are ongoing. ## Implications Short-term tests on C-316L as a BPP material show excellent performance and promising indications of stability, supporting the possibility of substituting the Pt-Ti in PEMWE BPPs with a much cheaper material. This view is strengthened by the predictions of the model, which reinforce the conclusion that the BPP sits at its OCP even under conservative PEMWE operating conditions. The natural question arises: is material substitution also an option for the PTL, at least in part? At the PTL/CL interface, a highly corrosion-resistant material would still be required, but it may be possible to use a cheaper material outside the corrosive region. The ex situ results indicate that C-316L is stable below E 1.2 V vs. RHE. Model predictions (Fig. S10, ESI †) for Type II water at 60 1C suggest that the local potential would only exceed this value within E200 mm of the CL. For traditional sintered PTLs with thickness in the range 1-2 mm, as in the single cell used in this work, the implication is that a cheaper material could be used for the bulk of the PTL, protected from the highly oxidising region by a thin microporous layer of corrosionresistant material (such as Pt-Ti), as shown schematically in Fig. 4. The use of a microporous layer at the PTL/CL interface has been studied by various groups with the results broadly indicating that, notwithstanding material substitution advantages as considered in this paper, such a layer improves both the mass transport and electrode kinetics. Such a scheme may therefore have benefit both through enabling the use of cheaper materials and in improving performance. We recognise that component selection in state-of-the-art PEMWE stacks may diverge in design from the sintered PTL studied here. Stack design can nonetheless clearly benefit from awareness, at the design and material selection stage, of the spatial non-uniformity of aggressivity of electrochemical conditions through the PTL thickness, to an extent that depends on material properties, especially electrolyte conductivity. Application of both our theoretical methodology and in situ reference electrode techniques to these systems may result in improved understanding and optimised stack design, with contextual consideration for cost and performance requirements. While the case for material substitution within the PTL is more nuanced than for the BPP and will depend on existing stack design, material properties and operating conditions, this work has demonstrated a powerful new approach to design optimisation and cost reduction that can be applied over a wide range of PTL materials, thicknesses and geometries. The analysis and model developed herein are also readily applicable to a wide range of other topical electrochemical energy conversion devices that have water or electrolyte circulating between an electrode and current collector. Such devices include redox flow batteries, CO 2 reduction cells and anion exchange membrane electrolysers, where, for instance, there is still debate about the optimal nature and utility of the electrolyte. ## Conclusions A new predictive modelling approach to understanding potential profiles in PEMWE anode components has been developed and validated using in situ reference electrode measurements. The model predicts that under typical operating conditions the corrosive zone associated with the elevated potential of the anode electrode extends only E200 mm into the PTL due to the low ionic conductivity of the aqueous phase. Furthermore, single cell tests with carbon-coated stainless steel anode current collectors showed no evidence of corrosion or degradation after 30 days of operation at 2 A cm 2 . The results demonstrate the potential for significant cost savings in PEMWE stack design by substitution of Pt-Ti with cheaper BPP and PTL materials. This combined modelling and experimental approach will also be applicable to design optimisation for other electrochemical energy conversion technologies, such as redox flow batteries, CO 2 reduction cells and anion exchange membrane electrolysers. ## Physicochemical model The physicochemical model predicts the spatial potential profile on the anode side of the cell in 1D (parallel to the current flow direction) subject to the following approximations: Electrolyte current in the aqueous and ionomer phases obeys Ohm's law with a constant resistivity (different values for each phase). Electrical current in the electron-conducting PTL structure obeys Ohm's law with a constant resistivity. Mass transfer effects are neglected. The PEMWE is considered to be homogeneous in the plane of the electrodes. The oxygen evolution reaction at the anode has Tafel kinetics, fit to the experimentally measured polarisation curve. 40 The PTL corrosion process proceeds with linear kinetics as a function of polarisation, with a defined exchange current density. This approximation is chosen in the absence of a more detailed characterisation for the PTL surface electrochemistry; with respect to any nonlinear kinetic model, it is conservative with respect to potential penetration into the PTL (i.e. a linear model predicts a larger polarised region). The cathode is assumed to behave ideally (zero hydrogen evolution reaction overpotential). The cell is assumed to be isothermal (uniform temperature). All porous media are treated as locally homogenised, with their morphology expressed by a local porosity and tortuosity. We choose a representative rate for the PTL passive corrosion kinetics of i 0,PTL = 2 10 4 A m 2 based on reported values of exchange current density and activation energy for the Pt/PtO surface. 41 The mathematical formulation of the current distribution model subject to the above approximations is given in Supplementary Note 2 (ESI †). All equations were solved in COMSOL Multiphysics 5.6 (COMSOL AB, Stockholm, Sweden). ## Materials The catalyst coated membrane (CCM, FuelCellsEtc, USA) comprised a Nafion 115 membrane, 3 mg cm 2 IrRuO x anode, and 3 mg cm 2 Pt black cathode. The CCM was pre-treated by submerging in 1 M H 2 SO 4 at 60 1C for 6 h, rinsing with Type I water three times, and finally immersing overnight in Type I water. The CCM was activated overnight at 1 A cm 2 before any measurement was performed. The metal sheet samples (82 mm thick) were bright-annealed 316L stainless steel and carbon-coated 316L stainless steel. The manufacturing process for the carbon-coated stainless steel has been described previously. 27,28 In brief, the oxide layer of bright-annealed 316L was removed (plasma etching) and, subsequently, the stainless-steel strip was coated with a metallic adhesion layer and carbon in a coil-to-coil, high-throughput physical vapour deposition (PVD) coating line. The production process aimed to achieve a 35 nm carbon layer and a 95 nm adhesion layer, but some variation of these values is expected over large areas. Pt-coated Ti was produced in a batch process as follows. Titanium sheets (Grade 1, 75 mm thick) were cleaned ultrasonically in 10-20 wt% NaOH + 5-10 wt% N,N-bis(carboxymethyl)alanine trisodium salt aqueous solution (Trutest AK-13 Ultra OS, A Clean Partner International AB, Sweden) at 60 1C for 10 min. Afterwards, they were rinsed in warm tap water and finally in deionised water (10 MO cm). The samples were mounted in an Ebeam vacuum chamber for batch coating. The chamber was pumped down to a pressure of 10 5 mbar and the Ti substrates were heated to 200 1C. During adhesion layer coating, the pressure was balanced to 8 10 5 mbar by Ar chamber back-fill. For the final Pt layer, Ar was turned off and the pressure stabilized at 2 10 5 mbar during coating. Nominal layer thicknesses were measured by a quartz crystal microbalance (QCM) as 70 nm for the adhesion layer and 40 nm for the Pt layer. ASTM Type I water (Elga, Purelab Ultra, UK) was used in all experiments, with the exception of the PTL potential profile experiment. Solutions of 0.5 mM and 0.05 mM H 2 SO 4 were prepared by diluting 0.5 M H 2 SO 4 (Fluka, UK) with Type I water. The pH was measured using a Semi-Micro pH electrode connected to a pH meter with a temperature probe (Jenway 3520, Cole-Parmer, UK). The pH electrode was calibrated prior to measurement using NIST traceable buffer solutions (Sigma Aldrich, UK). Conductivity was measured using a conductivity meter (Jenway 4520 Conductivity meter, Cole Palmer, UK), which was calibrated prior to measurement using a 147 mS cm 1 NIST traceable conductivity solution (Sigma Aldrich, UK). ## X-ray computed tomography (XCT) The PTL disk was mounted onto the sample holder using adhesive putty and then transferred to the rotation stage of a lab-based X-ray micro-CT system (Zeiss Xradia Versa 620 microscope, Carl Zeiss, CA, USA). 2001 X-ray projections of the PTL disk were acquired over a full rotation of 3601 at a voxel size of 16.9 mm. The exposure time of each projection was 1.5 s with a source voltage of 140 kV. The projections were then reconstructed using the built-in software Zeiss XMReconstructor, based on standard filtered back-projection algorithms. Commercial software Avizo V9.5 (Avizo, Thermo Fisher Scientific, Waltham, MA, USA) was used for the post-processing. An automatic thresholding algorithm was employed to segment the volume into pore and solid phase for porosity analysis. The tortuosity measurement was conducted in the open-source software TauFactor. 42 In situ reference electrode The in situ reference electrode consisted of a reversible hydrogen electrode (RHE, Hydroflex, Gaskatel, Germany) immersed in 0.5 M H 2 SO 4 and connected to the single cell with an acidfilled Nafion tubing. Further details have been reported in our previous work. 23,25 In this work, holes were drilled in the anode piston to accommodate either 1 single reference electrode or 4 reference electrodes in a cross configuration. As the cell area is small, inhomogeneity due to lateral flow distribution is assumed to be negligible. For the 7 day in situ reference electrode measurement, the Nafion tube end tip was positioned at the interface of the metal sheet and mesh. Metal sheet samples were sandwiched between the piston and the gridded mesh. To measure the potential profile within the PTL, holes in the PTL were drilled to different depths and the Nafion tube end tips were positioned at the bottom of the respective holes. Measurements at each point within the PTL were conducted three times in each electrolyte (0.5 mM H 2 SO 4 , 0.05 mM H 2 SO 4 and Type II water) to ensure repeatability. Measurement of E loc with the reference electrode was made in a two-electrode arrangement by connecting the working electrode cable to the anode piston and the reference electrode cable to the RHE. Polarisation curves were recorded by firstly ramping the current density at defined values with a 5 min dwell time up to 2 A cm 2 and then lowering it at defined current density steps with 1 min dwell time and 30 s recording time. ## Electrolyser testing The electrolyser single cell consisted of an acrylic circular cellhousing with an active area of 8 cm 2 . The titanium cell components comprised a piston, mesh (1 mm thick), and sintered porous transport layer (PTL, 2.1 mm thick) coated ## Paper Energy & Environmental Science with 100 nm of Pt (Teer Coatings, UK). All components were ultrasonicated sequentially in acetone (VWR Chemicals, UK), isopropanol (VWR Chemicals, UK), and Type I water for 15 min each before assembly. An extra step of ultrasonication in hexane (VWR Chemicals, UK) for 15 mins prior to the other organic solvents was added for the C-316L and 316L metal sheets to remove impurities arising from sample manufacture. In all tests, the metal sheet was sandwiched between the piston and the mesh. The hardware was tightened using 4 M6 bolts, nuts, and spring washers at a torque of 1 N m and then pneumatically compressed using 20 bar nitrogen from the cathode side. 80 ml min 1 of feed water was circulated in a closed loop to both the anode and cathode using a peristaltic pump (Watson Marlow, UK), with individually separated tanks for each electrode. The water was pre-heated to 60 1C using a water bath (Grant Optima TC120, Grant Instruments, UK). ## Scanning electron microscopy (SEM) The as-received and tested materials were characterised using high resolution scanning electron microscopy (FEG-SEM, Zeiss Sigma VP or Zeiss Ultra 55) in combination with energy dispersive X-ray spectroscopy (EDS, Oxford Aztec software and X-max 50 mm 2 SDD detector or Oxford Inca energy 450 system). The SEM and EDS analyses were performed using a primary electron energy of 5 keV. ## Auger electron spectroscopy (AES) AES analyses were performed with an ULVAC-PHI 700 Xi scanning Auger nanoprobe with an acceleration voltage of 10 kV and a primary beam current of 10 nA. Auger depth profiles were performed with Ar + ion sputtering at an accelerating voltage of 2 kV, giving a sputter rate of 27 nm min 1 at the sample position (as measured on a 100 nm thick Ta 2 O 5 reference material at the same position) using 1 mm 1 mm rastering of the ion beam. On uncoated samples a 2 mm 2 mm raster was used, giving a sputter rate of 8.3 nm min 1 . PHI MultiPak (Physical Electronics, Chanhassen, MN, USA) was used to evaluate the Auger depth profile data. Linear least squares fitting was used to extract different chemical states of the elements found. ## Trace metal analysis Analysis of trace metals in the CCM was conducted by an accredited lab (ALS, Luleå, Sweden); the details have been previously described by Novalin et al. 28 In addition, samples of both anode and cathode feed water before and after the 30 day test with a C-316L anode CC were analysed for trace metals using an inductively coupled plasma mass spectrometer (ICP-MS, Agilent 8900). The analysis was semi-quantitative, giving results as indicative mass fractions for a range of elements. The samples were matrix-matched in 1% HNO 3 to a semiquantitative standard, which was PA tuning solution (Agilent). ## Ex situ OCP measurement Prior to ex situ measurement, the reverse of the metal sheet samples was ground with sandpaper and connected to an ethylene tetrafluoroethylene (ETFE)-insulated copper wire (RS Components, UK) using a silver epoxy resin (RS Components, UK). The silver resin connection was dried at room temperature overnight followed by final curing at 65 1C for 2 h. The connection and the sample reverse face were then covered with an epoxy (Araldite Rapid Epoxy Adhesive, RS Components, UK) and left to dry overnight followed by curing at 60 1C for 2 h. The OCP of the sample was measured against a RHE immersed in 0.5 M H 2 SO 4 with Nafion tubing as the bridge to the electrolyte in a water-heated electrochemical cell. The electrolyte was dilute H 2 SO 4 (pH 4.5) to mimic the electrolyser water condition. 23 The electrolyte was oxygen-sparged and heated to 60 1C. The galvanically coupled potential of a Pt-Ti PTL contacting the C-316L metal sheet was measured by immersing both samples and combining the electrical connection as the working electrode. In order to obtain the Evans diagram, the potential of a Pt working electrode (+ = 2 mm, CH Instruments, USA) was scanned from OCP to 0.5 V vs. OCP at 10 mV s 1 and that of a C-316L working electrode (encapsulated with Araldite leaving an active area + = 13 mm) was scanned from OCP to +0.5 V vs. OCP at 10 mV s 1 . The counter electrode was a Pt wire (20 cm, + = 1 mm, Goodfellow, UK) and the reference electrode was the RHE/Nafion tubing configuration described above. ## Ex situ corrosion testing Ex situ corrosion testing of 316L, C-316L and C-304L coupons was conducted using an electrochemical setup accommodating eight working electrodes with a shared electrolyte and a Pt-Ti mesh counter electrode (Fuel Cell Store, Texas, USA), referred to as ''MultiCell''. The working electrodes were coupons made from a (coated) metal strip and were sealed against the cell body (polytetrafluoroethylene, PTFE) with a Viton O-ring exposing a 3 3 cm 2 geometric area to the electrolyte to avoid edge effects (uncoated portions due to cutting) for coated samples. Areas on the coupons not exposed to the electrolyte served as reference (not exposed to electrochemical conditions). A RHE (MiniHydrogen, Gaskatel, Germany) was placed in the middle of the exposed area of each working electrode at a distance of o1 mm. All eight RHEs were combined into one reference electrode signal at the potentiostat (ECi-211, Nordic Electrochemistry, Denmark), which helps to compensate for bubble formation and unstable reference potential. The cell was heated via a jacketed glass mantle and a thermostat (CC-205B High Precision Thermoregulation, Huber, Offenburg, Germany). The cell was initially heated to 70 1C over the course of 2.5 h and only when the correct temperature had been reached was the electrolyte added. Here, 217 ml of H 2 SO 4 and 1.7 ml of HF (40% EMSURE s ISO for analysis, Merck Chemicals and Life Sciences, Solna, Sweden) were added to achieve pH 3 H 2 SO 4 (49 ppm) and 0.1 ppm HF. This electrolyte is recommended by the US Department of Energy (DoE) to simulate the local BPP environment in PEM fuel cells. 26 In another experiment to mimic the single cell PEMWE water conditions, 6.7 ml of H 2 SO 4 (498%, EMSURE s for analysis, Merck Chemicals and Life Sciences, Solna, Sweden) were added to achieve pH 4. All coupons investigated in the ''MultiCell'' were cleaned by subsequent sonication in toluene (Z98%, Technical grade, VWR Chemicals, Sweden), acetone (Z99.8%, AnalaR NORMA-PUR s ACS, VWR Chemicals, Sweden), ethanol (absolute, Z99.5%, TechniSolv s , VWR Chemicals, Sweden), and ultrapure water (18 MO cm, Millipore Elix s Essential 15 UV and Milli-Q s Integral 3 system, Merck Chemicals and Life Science AB, Solna, Sweden) for 10 min each and dried in an argon stream (Laboratory Argon 5.7, 99.9997%, Linde Gas AB, Solna, Sweden) immediately before the experiment. After the coupons had been sealed against the cell, 8 L of ultrapure water was added and oxygen sparging (Z99.5%, Linde Gas AB, Solna, Sweden, 1000 ml min 1 , three inlets) commenced.

chemsum

{"title": "Assessing potential profiles in water electrolysers to minimise titanium use", "journal": "Royal Society of Chemistry (RSC)"}

molecular_engineering_of_β-substituted_oxoporphyrinogens_for_hydrogen-bond_donor_catalysis

## Abstract: A new class of bifunctional hydrogen-bond donor organocatalyst using oxoporphyrinogens with increased intramolecular hydrogen-bond donor distances is reported. Oxoporphyrinogens are highly non-planar rigid macrocycles containing a multiple hydrogen bond forming binding site. In this work we report the first example of non-planar OxPs as hydrogen-bond donor catalysts. The introduction of β-substituents is key to the catalytic activity and the catalysts are able to catalyze 1,4-conjugate additions and sulfa-Michael additions, as well as, Henry and aza-Henry reactions at low catalyst loadings (≤ 1 mol%) under mild conditions. Preliminary mechanistic studies have been carried out and a possible reaction mechanism has been proposed based on the bi-functional activation of both substrates through hydrogen-bonding interactions. ## Introduction Organocatalysis 1 has been widely studied over the past two decades, demonstrating synthetic utility for a wide range of transformations involving various methods for substrate activation. Hydrogenbonding promoted catalysis 2 has become commonplace as a mode of activation in organocatalytic processes with different structural motifs reported as being useful including ureas, thioureas, and squaramides. 3 The incorporation of additional basic moieties adjacent to the H-bond donor site gives a new bifunctional catalyst capable of the simultaneous activation of multiple substrates. 4 However, organocatalysts can suffer from poor efficiencies leading to high catalyst loadings, and their activities can be limited to particular reactions leading to reaction-specific catalysts. The overall catalyst structural motif, including the inter-hydrogen-bond donor distance and requirement for additional proximal functionality in these systems, suggested to us that non-planar porphyrinoids (Figure 1) might be of interest for catalytic applications. The comparison of literature reported intramolecular NH-NH hydrogen-bond donor distances between thioureas, squaramides and non-planar porphyrinoids, such as oxoporphyrinogens (OxPs), suggested that an increase in the distance between the donors may be achieved. 3 This is interesting towards applications as a synthetic mimic for the oxyanion hole, which are active site in enzymes capable of stabilizing negatively charged oxygens through H-bonding interactions. 5 Many studies have been carried out in synthetic mimics for this purpose include several common H-bond catalyst motifs including thioureas. 6 Studies have shown that the hydrogen bond donor intramolecular orientations and distances have significant effects on catalytic activity, suggesting new motifs with tunable properties and increased H-bond donor distance may be of interest. Metallated porphyrins 8 and porphyrinoids 9 are highly effective catalysts, including naturally occurring metalloporphyrins, such as hemes. 10 However, the use of porphyrinoids as organocatalysts has been largely neglected because the H-bonding sites are inaccessible within the plane of the porphyrin macrocycle 11 requiring interruption of macrocycle planarity to render them organocatalytically active, which is typically achieved by β-substitution, 12 protonation 13 or N-alkylation. 14 Senge and co-workers have demonstrated that some non-planar porphyrins catalyse a sulfa-Michael reaction. 15 Calix pyrroles (non-conjugated porphyrinoids) have also been used as organocatalysts despite their conformational flexibility, which leads to less well-defined binding sites. 16 These limited examples demonstrate that porphyrins and porphyrinoids are a promising class of compounds for organocatalysis. The required characteristics of non-planarity, H-bonding site availability and synthetic flexibility are also fulfilled by OxPs derived from the oxidation of meso-tetrakis(3,5-di-t-butyl-4hydroxyphenyl)porphyrin. 17 OxPs are essentially calix pyrrole macrocycles with rigidifying conjugated substituents at their meso positions with the additional attractive feature that they can be selectively N-alkylated, 18 for instance at N21 and N23, leading to formation of a highly non-planar rigid macrocycle containing a multiple hydrogen bond forming reaction site. Non-planar OxPs may be βsubstituent free or contain different multiplicities of β-substituents and are not charged. 19 These features have allowed their study for different applications including enantiomeric excess estimation, 20 water sensing 21 and alcohol differentiation. 22 In this case, their propensity to bind analytes, including anions 23 or solvents, 24 through oxygen atoms of oxo or carbonyl groups, through hydrogen bonding interactions along with their high degree of non-planarity and ease of synthetic modification suggested that they might be suitable materials for organocatalytic transformations. Herein, we describe the development of a new class of bifunctional H-bond organocatalysts based on β-substituted OxPs and demonstrate their utility in several transformations. The OxPs prepared are active at low catalysts loadings under mild conditions, can be used in several different reactions, and are easily recovered following reaction completion. ## Results & Discussion Catalyst design, synthesis and structure. Non-planar OxP 1a 17 and di-alkylated 2a, which renders only one face of the macrocycle accessible to interactions, have been reported. 25 However, contrary to previously reported non-planar porphyrinoids, 15,16 compounds 1a and 2a were not catalytically active for the addition of 2,4pentanedione to β-nitrostyrene (Table 1) despite the intramolecular H-bond donor distance for dialkylated OxP, which is typically 3.17 , 18,23,25,26 being similar to squaramide catalysts (Figure 1). 3d In order to induce catalytic activity further functionality analogous to that observed in squaramides was required, 4 such as basic functionality in proximity to the proposed catalytically active site of 2a. In order to achieve this β-substitution on the same face as the H-bond donors is required, giving the general structure shown in Figure 1b. The β-substitution of OxP was achieved by treatment of the Ni(II) or Cu(II) complexes of mesotetrakis(3,5-di-t-butyl-4-hydroxyphenyl)porphyrin 27 with Vilsmeier reagent 28 with an appropriate work-up leading to Vilsmeier adduct hydrolysis and acid induced demetallation, followed by treatment of the crude material with hydroxide to induce oxidation under aerobic conditions to give OxP-β-CHO (1b). Selective N-alkylation of 1b, here with 4-bromobenzyl bromide to aid solubility, can be accomplished 29 with pyrrolic NH and the formyl on the same face of the molecule (2b). Regiochemistry was confirmed by 2D NMR spectroscopy (see Figures S1,S2) of 2b and X-ray crystallography of its derivative 2i (Figure 3) with further confirmation of substituent introduction found in the X-ray crystal structure of a nickel complex of a precursor (Figure 4). Tetra-alkylated 3b was synthesized as a negative control to confirm catalytic activity of the pyrrole units. Reductive aminations with a structurally diverse range of amines were carried out with 1*b, 2b, and 3b to synthesize a library of catalysts containing amino functionality at the β-position of OxP proximal to the H-bonding site (Figure 2) allowing for preliminary structure-activity relationships to be probed. An interesting property of these compounds is that due to the removal of symmetry from the porphyrin macrocycle caused by the β-substitution, compound 2b and those derived from it are chiral. The macrocyclic conformation has been demonstrated to interconvert between its enantiomeric forms 20c but the presence of the 4-bromobenzyl alkylating groups hinders the macrocycle inversion through the plane of the molecule and therefore lock the macrocycle conformation, causing the formation of a stable, but racemic, enantiomer pair. However, the purpose of this body of work is to determine how to introduce catalytic activity into the macrocycle core, as 1a, 2a and 3a are not organocatalytically active, and therefore the chirality has not been investigated in this study. Catalysis results and substrate scope. Catalysts were screened for activity in the 1,4-conjugate addition of 2,4-pentanedione (5a) to βnitrostyrene (4a) as this has been used as a benchmark reaction for a number of hydrogen-bond donor catalysts. No conversion was observed for compounds lacking β-substituents (1a, 2a, 3a, Table 1, Entries 3-5) while the introduction of a dimethylamino group (2c, Table 1, Entry 6) lead to 10% conversion to the expected product 6aa. The use of a secondary amine pentylamino group (2d, Table 1, Entry 7-8) in place of the tertiary amine group resulted in a dramatic increase in reactivity with complete conversion observed to 6aa at 1 mol% catalyst loading. Catalyst loading was subsequently decrease to 0.5 mol% in order to more accurately compare the activity with other derivatives synthesized with a decrease in loading of 2d to 0.5 mol% gave a reduced in conversion to 60%. The tetra-N-alkylated derivative 3d and the acetylated form 2e exhibited no catalytic activities (Table 1, Entries 9-10), indicating that both pyrrolic and β-substituent amine functionality is required. Exchange of the pentylamino of 2d for 3-picolyl (2f) and 2-picolyl (2g), both lead to a decrease in catalytic activity to less than 10% conversion (Table 1, Entries 11-12). However, increasing the linker length between the amino group and the pyridyl substituent by a single carbon (2h, Table 1, entry 13) restored activity to a level above those previously obtained for 2d. The reasons for this increased activity are unclear although X-ray crystallographic structures of the precursor shown in Figure 4 and derivative 2i (Figure 3) indicate that the pyridyl in 2h may be positioned remote from the catalytic site, further suggesting that in 2f and 2g the pyridyl groups interfere with the catalysis reaction. To confirm catalytic activity, the OxP compounds lacking pyrrole N-alkyl groups (1h), tetra-N-alkylated (3h) and β-position aminoacylated (2i) were studied with all exhibiting lower or no catalytic activity relative to 2h (Table 1, Entries 14-16). Also, pyridine (Table 1, Entry 2) did not catalyze the reaction and the phenyl analogue 2j gave a similar conversion (Table 1, Entry 17) indicating that pyridyl groups introduced to the βsubstituent do not participate in the catalytic process and are furthermore not responsible for product formation. After having successfully used derivatives of 2 to catalyze the Michael addition of 2,4-pentanedione (5a) to β-nitrostyrene (4a) an optimization of reaction conditions was performed (see Table S1). At 0.5 mol% loading of 2h, solvents containing carbonyl-type oxygen atoms capable of competitively hydrogen-bonding to the active site of oxoporphyrinogens (e.g. acetone, ethyl acetate) hinder the reaction and significantly lower conversions were obtained. Surprisingly, solvents with alcohol moieties, which typically also bind to OxPs, gave large improvements in conversion with ethanol giving quantitative conversion to 6aa at 0.5 mol% catalyst loading whilst the background rate was shown to remain negligible. It is important to note the significance of a solvent such as ethanol successfully catalyzing these reactions, firstly for its status as a green solvent compared to more typical solvents employed for these reactions, such as DCM and toluene. 30 Secondly, protic solvents are capable of competitively forming H-bond and therefore can interrupt interactions required for the successful outcome of H-bond catalysts, interestingly this does not appear to effect the activity of catalysts such as 2h. The substrate scope of the reaction with optimized conditions was examined based on variation of aryl-substituted nitrostyrenes using 5a as the nucleophile for the Michael addition (Scheme 1). The variation of substrate aryl substituent was tolerated with greater than 90% yields maintained regardless of the electron-donating (6ba, 6ca, 6ia) or electron-withdrawing (6da, 6fa, 6ga) properties and substituent position (6ja, 6ka, 6la, 6ma). Compound 6na was obtained in a lower isolated yield despite appearing to have occurred quantitatively according to TLC analysis of the reaction mixture due to losses incurred during chromatographic purification. Substrate 4e is poorly soluble in the reaction solvent but increasing the catalyst loading to 2 mol% led to an improved yield of 6ea to 97%. Nitroolefin 4h gave an acceptable yield of 72% for the formation of 6ha. Interestingly, the products (6) are insoluble in the reaction media and precipitate from the reaction mixture facilitating their isolation by filtration albeit in slightly lower yield than obtained using purification by column chromatography. The catalysts can also be recovered by using this method because of their intense color and high polarity relative to the product. Scheme 2. Nucleophile substrate scope in the 1,4-conjugate addition of 5a-j to 4a catalysed by 2h. Reaction conditions: 5 (1.36 mmol, 2 equiv.) was added to 4a (102 mg, 0.68 mmol, 1 equiv.) and 2h (5.46 mg, 3.42 µmol, 0.5 mol%) in EtOH (2 mL) and stirred under N2 for 16 hours. Isolated yields after purification by column chromatography are reported. Variation of the nucleophile used led to greater variation in the yields of the Michael addition products (Scheme 2). The substitution of the terminal methyl groups of 2,4-pentanedione with simple phenyl groups (6ab, 6ac) resulted in only a small reduction in reactivity while the substitution of one ketone of the 1,3-dione with an ester (6ad, 6ae, 6af) gave decreases in yield correspondent with the bulk of the ester. The steric and electronic properties of the remaining ketone could be modified (6ag, 6ah) without considerable loss in product yield in both examples. The reaction conditions also allowed for the use of nitriles, with both malononitrile (6ai) and cyanoacetate (6aj) adducts successfully synthesized in high yields. Low diastereoselectivity (dr) was observed for the unsymmetrical substrates studied here with the majority isolated as essentially 1:1 mixtures of diastereoisomers with the highest dr observed for 6ab as 1:1.8. Variation of nucleophile 4 had a greater effect on the yields of the Michael addition products (Scheme 2), which is consistent with the mechanism proposed in Figure 3 as it is the substrate being activated by the catalyst. ## Mechanistic Studies. A plausible mechanism is shown in Figure 5 based on that reported by Soós and co-workers, 31c 1 H NMR binding and kinetic studies were carried out in order to support the proposed mechanism. However, it is important to note that this mechanism has been proposed by comparison to mechanistic and computational studies for a number of literature reported catalysts. 31 5 (centre). Binding constants were determined using the two exchangeable pyrrolic NH and the benzylic type protons at the β-position (shown in Figure 5), these were used due to their large shift in the NMR and there being no interfering protons throughout the course of the titration. To ensure that competing analytes were not present all NMR solvents had been neutralized prior to use and the measurement temperature was kept constant to avoid shifts in exchangeable protons. The highest binding constant of 0.43 M -1 was observed with 5a, whereas both 4a and 6aa showed lower binding constants of 0.26 M -1 and 0.025 M -1 , respectively. It should be noted here that the saturation point of 6aa was reached at 212 equivalents, which could lead to binding constant inaccuracy. Based on the magnitude of changes in the 1 H NMR shifts, and by comparison with reported mechanisms, 31 dicarbonyl 5a interacts with all three NH functionalities (2h-5a) through two H-bonds to the pyrrolic NH and a third possible interaction due to deprotonation of substrate 5a by the amino functionality in the β-substituent of 2h. Indeed, this deprotonation step appears to be the key role of the βsubstituent as non-β-substituted OxPs have previously been shown to form H-bond interactions but these compounds, such as 2a, are not catalytically active in the above studied reactions, demonstrating the importance of the β-substituent for catalytic activity. Nitrostyrene 4a interacts with a single pyrrolic NH and the β-substituent which suggests a complex such as 2h-5a-4a. The preorganisation of the substrates, as well as the substrate activation due to H-bond interaction, may facilitate and promote bond formation to give 2h-6aa. No variation in the benzylic protons chemical shifts was observed in titrations of product 6aa with 2h, indicating that following product formation H-bond interaction with the β-substituent is substantially attenuated to give 2h-6aa'. Weak binding of 6aa may allow catalysis to proceed since it is released (or displaced by fresh substrate through competitive binding) regenerating 2h, which can re-enter the catalytic cycle. Competitive binding studies were also performed by monitoring the effect of the addition of 5a to a preformed complex of 2h-4a and vice versa (see Figures S6,S7). Addition of 100 equivalents of 4a to the catalyst causes a shift in the NMR signals of <0.1 ppm. However upon addition of 100 equivalents of 5a to the 2h-4a complex a large shift of 0.6ppm for a pyrrolic NH, amongst other variations, is observed. In the opposite case, the formation of the 2h-5a complex causes a large shift in the 1 H NMR signals which is followed by only minor changes upon addition of nitrostyrene 4a. This result supports the hypothesis that the catalyst forms a stronger interaction with the nucleophile (5a) than electrophile (4a) and demonstrates that 5a exhibits a strong binding interaction even in the presence of 4a supporting the proposed mechanism in this report, more specifically the initial formation of 2h-5a. The mechanism proposed by Takemoto and co-workers 4 was believed to activate the electrophile through H-bonding interactions whereas later studies, including mechanistic studies by Soós and coworkers 31c amongst others, 31 demonstrated nucleophile activation, which is more in-line with results obtained with 2h. In order to further support the proposed mechanism, kinetic parameters were determined by carrying out the reaction in CDCl3 at 24 ± 2 °C utilizing the pseudo-first-order kinetics method (Figure 6). 31a Through separate 1 H NMR studies with variations in the reagent equivalents and catalyst loading the reaction was shown to be pseudo first order in 4a, 5a and 2h, meaning that the rate equation for this reaction is: Rate = k[2h] [4a][5a]. As only reaction events that occur at or prior to the rate-determining step (RDS) are included in the rate equation, and can be studied in these kinetics experiments, the formation of 2h-5a adduct and related catalyst protonation cannot be the RDS due to the presence of 4a in the rate equation. Due to the first order kinetics observed for 4a, the RDS is most likely the tertiary complex formation (2h-5a-4a) or subsequent C-C bond formation. The weak binding of 6aa with the catalyst, determined by 1 H NMR titrations, would suggest that the release of the product from the catalyst is a facile process, excluding this as the RDS. Unfortunately, mechanistic studies could not be performed in methanol or ethanol due to deuteriumhydrogen exchange rendering the protons of interest unobservable in the 1 H NMR spectra and due to the poor solubility of the catalyst in these solvents causing signal broadening. However, upon repetition of the control reactions with 3h and 2i in ethanol, no conversion to 6aa was observed suggesting that 2h continues to act by the same mechanism despite the use of a polar protic solvent capable of competing with H-bonding interactions. ## Reaction Scope. Scheme 3. Screening of reaction scope for catalyst 2h. Conversion was determined by 1 H NMR spectroscopy with an internal standard. Preliminary investigations into the synthetic utility of catalyst 2h for a range of other transformations (Scheme 3) typically catalyzed by H-bond donor catalysts were performed. 15 It was found that 2h is catalytically active in Michael addition reactions with other substrates, such as the addition of malononitrile (5i) to chalcone (7). Additionally, 2h is highly effective for the activation of sulfa-Michael reactions leading to products 11 and 13 in quantitative conversions. The Henry and aza-Henry reactions were also successfully catalyzed by 2h in moderate conversions furnishing products 15 and 17. Further studies were carried out for the sulfa-Michael reaction (Table 2), in order to gain a direct comparison with previously reported porphyrinoid catalysts for this transformation. Firstly, when compared to the pioneering work in the field reported by Senge and co-workers, where 80% conversion was obtained for the formation of 11, 15 catalyst 2h gave quantitative conversion to 11 under the same conditions (Table 2, Entry 1). Secondly, it was found that for non-β-substituted (2a, 3a) and tertiary amine (2c) catalysts analogous results were obtained as to those for the addition of acetylacetone to nitrostyrenes outline above, further demonstrating the importance of the βsubstituent. Finally, with the modified conditions used earlier in this study, catalyst loading could be reduced by 1/3 that used in the previously reported porphyrinoid-catalyzed formation of 11, 15 whilst maintaining quantitative conversion (Table 2, Entry 7) simply by utilizing ethanol as the reaction solvent. ## Conclusion In conclusion, a novel class of OxPs has been synthesized and their application as bifunctional H-bond donor catalysts has been investigated. The catalysts possess multi-reaction activity with wide substrate scope and operate at low catalysts loadings (≤ 1 mol%). We have demonstrated that βsubstitution is essential to establish catalytic activity, and preliminary mechanistic studies indicate that they interact with substrates in a similar fashion to already reported catalysts. The catalysts exhibit activity in the Michael addition reactions and preliminary results reveal the catalyst's excellent activity for sulfa-Michael additions compared to literature reported porphyrinoids, as well as moderate activity for Henry and aza-Henry reactions. We believe that these OxP systems offer an extremely adaptable scaffold for the development of H-bond catalysts due to a concave 3-dimensional structure at their binding sites, which can be modified to affect and optimize any catalytic processes occurring there. These OxP systems possess excellent potential for the design of supramolecular catalysts beyond what is currently possible. We are currently continuing to investigate their applications as organocatalysts.

chemsum

{"title": "Molecular Engineering of \u03b2-Substituted Oxoporphyrinogens for Hydrogen-Bond Donor Catalysis", "journal": "ChemRxiv"}

lytic_polysaccharide_monooxygenase_(lpmo)_mediated_production_of_ultra-fine_cellulose_nanofibres_fro

## Abstract: The production of cellulose nanofibres (CNFs) typically requires harsh chemistry and strong mechanical fibrillation, both of which have negative environmental impacts. A possible solution is offered by lytic polysaccharide monooxygenases (LPMOs), oxidative enzymes that boost cellulose fibrillation. Although the role of LPMOs in oxidative modification of cellulosic substrates is rather well established, their use in the production of cellulose nanomaterials is not fully explored, and the effect of the carbohydratebinding module (CBM) on nanofibrillation has not yet been reported. Herein, we studied the activity of two LPMOs, one of which was appended to a CBM, on delignified softwood fibres for green and energyefficient production of CNFs. The CNFs were used to prepare cellulose nanopapers, and the structure and properties of both nanofibres and nanopapers were determined. Both enzymes were able to facilitate nanocellulose fibrillation and increase the colloidal stability of the produced CNFs. However, the CBMlacking LPMO was more efficient in introducing carboxyl groups (0.53 mmol g −1 ) on the cellulose fibre surfaces and releasing CNFs with a thinner width (4.3 ± 1.5 nm) from delignified spruce fibres than the modular LPMO (carboxylate content of 0.38 mmol g −1 and nanofibre width of 6.7 ± 2.5 nm) through LPMO-pretreatment followed by mild homogenisation. The prepared nanopapers showed improved mechanical properties (tensile strength of 262 MPa and modulus of 16.2 GPa) compared to those obtained by conventional CNF preparation methods, demonstrating the potential of LPMOs as green alternatives for cellulose nanomaterial preparation. † Electronic supplementary information (ESI) available: Sugar analysis, mechanical properties data, gene sequences, restriction pattern analysis, colony PCR, SDS-PAGE and western-blot. See ## Introduction Extensive use of plastics in consumer products has caused wide-ranging environmental issues due to the long-term persistence of plastics in the global ecosystem, ingestion of microplastics by organisms, and their accumulation along food chains. 1,2 Therefore, there is a pressing need to develop biodegradable materials with good properties from naturally occurring polymers, such as cellulose. In recent years, cellulose has gained popularity in sustainable materials research in the form of cellulose nanofibres (CNFs), which are three orders of magnitude smaller than intact fibre cells. Cellulose nanomaterial-enabled products are lighter, stronger, biologically compatible, and environmentally benign, which makes CNFs ideal replacements for the conventional synthetic polymer materials. 3,4 To produce CNFs, cellulose fibres from natural sources need to be disintegrated into nanoparticles. 5 Strong intra-fibre hydrogen bonding and a tendency to aggregate in water make cellulose fibres resistant to mechanical disintegration, and consequently strong disintegration methods that require high energy or chemical modifications of cellulose fibres are needed. Various chemical pre-treatment approaches, such as 2,2,6,6-tertamethylpiperidine-1-oxyl (TEMPO)-mediated oxidation 6,8 or quaternisation via nucleophilic addition, 9 are able to facilitate the fibrillation of cellulose with a kitchen blender and produce CNFs with the corresponding negative or positive surface charges and also with a uniform width of 3-4 nm, which is similar to the width of cellulose microfibrils in plants. However, there are still major environmental issues related to some chemical modifications in CNF production, particularly when periodate oxidation, TEMPO-mediated oxidation, or sulfuric acid hydrolysis is employed. The dicarboxylic acid hydrolysis method has been shown to be sustainable in cellulose nanomaterial production owing to the acid recovery. 11 Enzymatic pre-treatment using glycoside hydrolases is more environmentally friendly and requires reduced energy consumption for fibrillation compared to the untreated sample. 12,13 However, multiple passes (at least 30 passes) through the microfluidizer are still required in order to prepare CNFs with a thinner width of 4-12 nm. 13 Thus, using carbohydrate-active enzymes to assist the production of CNFs with tailored surface properties and an ultrafine width and significantly reduce the energy consumption for mechanical homogenisation still remains a challenge. In nature, recalcitrant wood cell wall components, including cellulose, are effectively decomposed under mild, aqueous reaction conditions mainly by fungi. 14 Fungi can accomplish this by utilising an array of efficient carbohydrate-active enzymes, including lytic polysaccharide monooxygenases, LPMOs. Most fungal LPMOs are classified in the auxiliary activity family AA9 in the carbohydrate-active enzyme database (http://www.cazy.org). These recently discovered enzymes are capable of oxidative modification of recalcitrant polysaccharides, such as cellulose. 15,16 On cellulose, LPMOs perform a highly specific oxidation reaction at either the C1 or C4 position, the reaction products being aldonic acids or gem-diols, respectively, of which the former contains an ionisable carboxyl group. 17 When acting on low-molecular weight cellulosic substrates, C1-specific LPMOs release soluble cello-oligomers containing carboxyl groups at their reducing ends. 18 On fibres, the oxidative chain cleavage does not yield soluble products, but instead carboxyl groups are formed on the fibres, mainly at the crystalline areas. 19,20 The C1-localised carboxyl groups have been shown to facilitate cellulose fibrillation owing to the increased colloidal stability of the produced CNFs, in a much similar manner to that of C6-carboxyl groups derived from TEMPO-oxidation. The unique aspect of LPMO-oxidation is the chain breaks introduced by the enzyme, which might also enhance cellulose fibrillation by decreasing the crystallinity. A separate carbohydrate-binding module (CBM) possessed by some LPMOs is thought to enhance the LPMO activity by bringing the enzyme's active site into contact with the crystalline regions of cellulose, 25 but little is known about the effect of the CBM on the subsequent mechanical fibrillation process, as it may also restrict the movement of the LPMO. 26 In the present work, we developed an environmentally friendly and energy-efficient method to produce CNFs and cellulose nanopapers with robust mechanical properties. We used two C1-active LPMOs from the ascomycetous fungus Neurospora crassa (NcLPMO9E and NcLPMO9F) to boost the fibrillation of delignified wood fibres. NcLPMO9E contains a family 1 CBM and NcLPMO9F is non-modular. The delignified wood fibres, i.e. holocellulose, were prepared from spruce wood by using peracetic acid (PAA). Treatment with PAA selectively removes lignin while preserving the native wood fibre structure without the defects typically present in commercial wood pulp fibres, and the process is environmentally sustainable. 27 The LPMOs were removed after the enzymatic reaction, and the fibres were subjected to mild and short mechanical homogenisation. The structure and morphology of the resulting nanofibres after the reaction with a CBM-containing LPMO and a non-modular LPMO were characterised and compared. The LPMO-oxidised nanofibres were further used to prepare cellulose nanopapers by vacuum filtration and their mechanical properties were studied in uniaxial tensile tests and compared to those of the unmodified holocellulose fibres. ## Preparation of holocellulose The holocellulose fibres were prepared according to a previously reported method. 27 Briefly, softwood spruce sticks were first soaked in water under vacuum, and subsequently treated with 4% (w/w) peracetic acid (PAA) for 45 min at 85 °C. The PAA solution was prepared by diluting 38-40% (w/w) PAA solution (Sigma-Aldrich), followed by adjusting the pH to 4.6 using NaOH before the reaction. The weight ratio of PAA and softwood was 35 : 100. Four rounds of PAA treatment were performed until the wood sticks were disintegrated into individual fibres. Most of the lignin was removed after this treatment as 1.5% (w/w) residual lignin was present in the final samples compared to the initial content of 21% (w/w). No mechanical stirring was applied during the treatment to avoid fibre damage. To increase the enzyme accessibility of holocellulose fibres, the hemicelluloses were partially removed with pressurised hot water extraction (autoclave) followed by extensive washing in hot water before LPMO treatment (Table S1, ESI †). ## LPMO selection and sequence analysis We selected LPMOs that have been previously shown to possess C1-oxidation ability, i.e. NcLPMO9E and NcLPMO9F from the ascomycetous fungus N. crassa. 28,29 The amino acid sequences of the LPMOs were obtained from UniProt ( protein-IDs: Q1K4Q1; Q7RWN7). Native signal peptides were predicted with SignalIP software 4.1 (http://www.cbs.dtu.dk/services/ SignalP/). The theoretical molecular masses of the mature recombinant LPMOs were calculated from the amino acid sequences without the signal peptides with ExPASy ProtParam (http://web.expasy.org/prot-param/). The calculated theoretical molecular masses of NcLPMO9E and NcLPMO9F are 32 kDa and 24 kDa, respectively. ## Gene synthesis and cloning Genes encoding the LPMOs were commercially synthesised and codon optimised by GenScript, USA. Codon-optimised gene sequences for NcLPMO9E and NcLPMO9F are shown in Fig. S1 and S2, ESI, † respectively. The synthetic genes encoded native secretion signals at the N-terminus and (His) 6 tags at the C-terminus of the corresponding proteins. The genes were inserted into the EcoRI-NotI position of the P. pastoris expression vector pPICZB, in which the c-myc epitope and polyhistidine tag were omitted. The constructs were transformed into E. coli TOP10 cells (Thermo Fisher Scientific) by electroporation, and the transformants were cultivated in low salt Luria-Bertani (LB) broth containing 25 µg mL −1 zeocin (Invitrogen). The plasmid DNA was extracted using a GeneJet Plasmid Miniprep kit (Thermo Scientific). The constructs were confirmed by restriction pattern analysis (EcoRI and NotI -Thermo Scientific) (Fig. S3, ESI †), and the positive clones were sequenced by Eurofins Genomics, Germany. Constructs with verified sequences were linearised using PmeI and approximately 10 µg of DNA was used to transform P. pastoris X33 cells by electroporation. The clones were selected on YPDS agar containing 100 µg mL −1 zeocin. Integration of the heterologous genes into P. pastoris genomic DNA was confirmed by colony PCR utilising AOX1 primers and a standard protocol (Fig. S4, ESI †). ## Production and purification of the LPMOs The LPMOs were produced by cultivating the P. pastoris transformants in 250 mL flasks containing 30 mL of the growth medium. The cultivation conditions were 30 °C and 230 rpm. After overnight cultivation in BMGY, the cells were suspended into BMMY for a final cell density (OD600) of ∼1.0. BMGY and BMMY contained 10 g L −1 yeast extract, 20 g L −1 peptone, 1.34% YNB, 4 × 10 −5 % biotin and 100 mM potassium phosphate buffer at pH 6.0. BMGY also contained 1% glycerol. The cultures were supplemented daily with methanol and CuSO 4 at final concentrations of 2% (v/v) and 0.1 mM, respectively. The recombinant protein production was followed daily with SDS-PAGE and western-blot analyses as shown in Fig. S5 and S6, ESI. † The His-tagged LPMOs were purified by immobilised metal-ion affinity chromatography (IMAC) using an Äkta purifier FPLC system (GE Healthcare). The pH of the culture supernatant was adjusted to 7.4 before loading on a 5 mL IMAC FF HiTrap column (Bio-Rad), which was previously charged with a metal (Cu 2+ ) and equilibrated in 50 mM NaCl -5 mM imidazole -20 mM sodium phosphate, pH 7.4 (for NcLPMO9E) or 50 mM NaCl -20 mM sodium phosphate, pH 7.4 (for NcLPMO9F). Gradient elution was performed using 500 mM imidazole and 500 mM NaCl. The pooled proteins were gel filtered on a Biogel P-6 (Bio-Rad) column to 20 mM sodium phosphate, pH 7.4. Finally, the proteins were concentrated in 5 kDa cut-off Vivaspin® 20 centrifugal concentrators (GE Healthcare). Protein concentrations were determined using a Pierce™ BCA protein assay kit (Thermo Scientific). The molecular masses of LPMOs were estimated on SDS-PAGE. The samples were diluted in sample buffer containing 200 mM β-mercaptoethanol, 0.05% bromophenol blue, 5% SDS, and 50% glycerol in 225 mM Tris-HCl pH 6.8, and heated at 95 °C for 5 min before loading on Mini-protean® TGX™ 4-20% gels (Bio-Rad). PageRuler Prestained protein ladder (Thermo Scientific) was used as a molecular mass marker, and the gels were stained using a PageBlue™ protein staining solution (Thermo Scientific). SDS-PAGE separated proteins were transferred using a Trans-Blot SD Semi-Dry Transfer Cell (Bio-Rad) to 0.45 µm nitrocellulose membranes (Bio-Rad). Mouse anti-His anti-bodies were used as primary antibodies and anti-mouse IgG antibodies conjugated to HRP were used as secondary antibodies (Invitrogen). The membranes were washed in TBS buffer that contained 2.42 g L −1 Tris-base, 8 g L −1 NaCl and 0.1% (v/v) Tween®20, pH 7.4. Visualisation was performed using Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare) according to the manufacturer's instructions. The membranes were photographed using a Fujifilm LAS1000 camera running Image Reader LAS-1000 software version 2.6. ## LPMO activity assay The peroxidase activity of the purified LPMOs was determined spectrophotometrically (Cary 50 UV-VIS) against the soluble synthetic substrate 2,6-dimethoxyphenol (Sigma Aldrich). Formation of the reaction product (coerulignone) was monitored at a wavelength of 469 nm (ε 469 nm 53 200 cm −1 M −1 ). The assay contained appropriate amounts of the enzymes, 1 mM 2,6-DMP, 100 µM H 2 O 2 and 50 mM sodium phosphate buffer pH 6.5 in a final volume of 1 mL, as described by Breslmayr et al. 30 The reaction temperature was 25 °C. Three technical replicates were used for all assays. To determine their stability, the LPMOs were incubated for 0, 6, 24, 48, 72, and 120 h at 25 °C, after which their residual activity was assayed at 25 °C as described above. ## LPMO oxidation of holocellulose and fibrillation Holocellulose fibres (300 mg) were gently dispersed in 50 mM sodium phosphate buffer, pH 6.5 and soaked overnight. For the LPMO reactions, the fibres were diluted to a final concentration of 0.4% (w/v). The LPMO reactions contained 122 mg g −1 NcLPMO9E or 88 mg g −1 NcLPMO9F at an approximate molar ratio of 4 µmol g −1 of enzyme to cellulose, and 10 µM L-ascorbic acid as the electron donor for the enzymes. Reactions without LPMOs were conducted as controls. The reactions were performed at 25 ± 2 °C under magnetic stirring for 2 days. The reactions were stopped by separating the fibres from the suspension. To remove the enzymes, the fibres were dispersed in 10 µM NaOH pH 9.0 and incubated overnight. NaOH was removed by extensive washing with ultrapure water in a vacuum filtration unit equipped with a Durapore® 0.65 µM DVPP membrane (Merck). The LPMO-treated and the control holocellulose fibres were suspended in 500 mL distilled water and homogenised for 2 min at 25 000 rpm with an IKA T 25 digital ULTRA-TURRAX©, corresponding to 0.02 kW h. ## Preparation of nanopapers 100 mg of the homogenised fibres was gently suspended in ultrapure water at a final concentration of 0.05% (w/w). Water was removed from the suspensions with a vacuum-filtration setup equipped with a filter funnel (Φ = 70 mm) and filter membrane (0.22 µm, DVPP). The wet films obtained from filtration were carefully peeled off and placed between two stainless steel meshes. The films were dried under pressure at 50 °C for 24 h. ## Characterisation The chemical composition of the LPMO-treated and unmodified holocellulose fibres was determined by carbohydrate analysis. 100 mg of the fibres were hydrolysed with sulphuric acid, and the monosaccharides were quantified using a Dionex ICS-3000 chromatography system (Thermo Fisher Scientific, USA). The amount of glucose in glucomannan was estimated based on the glucose and mannose contents using a 1 : 3 ratio of glucose : mannose and subsequently subtracted out when calculating the cellulose content. Soluble sugars released from the holocellulose fibres by NcLPMO9E and NcLPMO9F treatment were determined by the phenol-sulphuric acid method. Samples of 100 µL were taken from the supernatants of the LPMO reactions and mixed with 100 µl of 5% phenol and 500 µl of 98% sulphuric acid. The absorbance was measured after 10 min at 490 nm, and the soluble sugars were quantified as glucose equivalent units against a standard curve of glucose. 31 The extent of the fibrillation was examined under an inverted light microscope (Nikon Eclipse Ti-S, USA) running imaging software Nis-Elements version 4.60. The solid content of the homogenised holocellulose fibres was adjusted to 0.1% (w/w) before observation. Transmission electron microscopy (TEM) observations were conducted using a Hitachi Model HT7700 transmission electron microscope operated in high-contrast mode at 100 kV. The specimens were prepared by depositing a drop of the dilute water suspension of cellulose nanofibres on a freshly glow-discharged, carboncoated grid and stained with 1% uranyl acetate as a negative stain. Atomic force microscopy (AFM) analysis was performed on a MultiMode 8 Atomic Force Microscope (Digital Instruments, Inc., USA) in the ScanAsyst® mode. The X-ray diffraction (XRD) patterns were recorded using a Philips X'Pert Pro diffractometer (model PW 3040/60) in the reflection mode (5-35°2θ angular range, steps of 0.05°). CuKα radiation (λ = 1.5418 ) was generated at 45 kV and 40 mA and monochromatised using a 20 μm Ni-filter. Diffractograms were recorded from rotating specimens using a position sensitive detector. The crystallinity index (CI) of cellulose was calculated from the ratio between the intensity of the 200 peak (I 200 ) and the intensity of the minimum (I AM ) between the 200 and 110 peaks. 32 The zeta (ζ) potentials of the holocellulose fibres were measured using a Zetasizer nano ZS instrument (Malvern, Worcestershire, UK) at 25 °C following the Smoluchowski method. For the measurements, the fibres were suspended in distilled water at a final concentration of 0.05% (w/v) and disintegrated by sonication (amplitude 30%, duration 30 s, pulse 50%). The amounts of the carboxyl groups introduced by the LPMOs were determined by conductometric titration. 100 g suspensions containing 50 mg of the holocellulose fibres were adjusted to a pH value of 2.5 with 0.1 M HCL. The suspensions were titrated with 0.01 M NaOH and the conductivity was monitored with a conductometric station (SevenCompact, Mettler-Toledo). The titration curve showed the typical presence of strong and weak acid groups. The amount of strong acid corresponded to the added HCl, and that of the weak acid corresponded to the carboxyl content. Fourier transform infrared spectroscopy (FTIR) was performed by using a PerkinElmer Spectrum 2000 instrument equipped with an MKII Golden Gate Single Reflection ATR system from Specac Ltd, UK. The ATR crystal was an MKII heated diamond 45°ATR top plate. The measurement was performed in a spectral range of 600-4000 cm −1 with a resolution of 4 cm −1 . The nanopaper surfaces and their cross-sections after freeze-fracture were analysed by field-emission scanning electron microscopy (FE-SEM) (Hitachi S-4800). The samples were attached with carbon tape onto metal stubs and then coated with platinum/palladium using a sputter coater (Cressington 208HR). The mechanical properties of the nanopapers were tested under uniaxial tension with a Universal testing machine (Instron 5944, USA). The prepared nanopapers were cut into strips of 3 mm in width and conditioned at a relative humidity of 50% for 2 d. At least ten specimens were tested from each sample and the results are reported as an average of at least five specimens. The test was performed by using a strain rate of 2 mm min −1 and a gauge length of 20 mm. The modulus was determined from the slope of the initial low strain region of the stress-strain curve. Toughness, defined as work to fracture, was calculated as the area under the stress-strain curve. ## Activity of NcLPMO9E and NcLPMO9F The apparent molecular weights of the recombinant enzymes, 43 kDa and 25 kDa for NcLPMO9E and NcLPMO9F (Fig. 1), respectively, corresponded well to those reported in a previous study where the enzymes were produced from P. pastoris using a similar protocol, 33 but without the C-terminal purification tag (His) 6 utilised in this study. In order to confirm the activity of the LPMOs, we utilised a spectrophotometric assay that detects coloured oxidation products of the synthetic substrate 2,6-dimethoxyphenol (2,6-DMP) from an LPMO-catalysed peroxidase reaction. 30 The peroxidase reaction is dependent on the presence of catalytic copper on the enzyme's active sites, and as it can only be demonstrated by LPMOs that are correctly processed and folded by the host, it could be used to confirm the activity of the recombinant LPMOs. Indeed, both of the N. crassa LPMOs were active on 2,6-DMP at pH 6.5 and 25 °C, and these conditions were also selected for later experiments using a more complex, natural cellulosic substrate. Under these conditions, the activity towards the synthetic substrate 2,6-DMP was observed to be higher for the NcLPMO9E than the NcLPMO9F, the highest observed specific activities being 105 nkat mmol −1 and 25 nkat mmol −1 , respectively (Fig. 2). The observed differences in the activities between the two enzymes corresponded to previous results utilising synthetic substrates. 33 While the peroxidase activity on the soluble synthetic substrate cannot be used to predict the activity on the native insoluble cellulosic substrate, 28,29,34 the assay is a convenient method to characterise the stability of the enzymes under certain reaction conditions to estimate the appropriate reaction time. The results showed that the modular NcLPMO9E demonstrated good stability at 25 °C, retaining 40% and 20% of its original activity after 48 h and 120 h, respectively. The non-modular NcLPMO9F was less stable under the same conditions, as the activity was observed to be lost after 48 h of incubation. ## LPMO-assisted holocellulose fibrillation Spruce holocellulose was incubated with NcLPMO9E and NcLPMO9F for 48 h before brief mechanical homogenisation. The enhanced fibrillation resulting from the LPMO treatment was clearly visible in the fibre suspensions after the mechanical treatment (Fig. 3a-c). This was further examined by using inverted light microscopy (Fig. 3d-f ) and TEM (Fig. 3g and h). Microscopy analysis revealed extensive fibrillation down to the nanoscale resulting from the LPMO treatments, while the control non-enzymatically treated fibres remained relatively intact and could not be analysed by TEM. The nanofibres obtained after the LPMO treatments were well individualised with some aggregations due to drying during sample preparation for TEM observation. Of the two studied enzymes, the CBM-lacking NcLPMO9F yielded smaller nanofibres with an average width of approximately 5 nm (Fig. 3h), while the nanofibres resulting from the CBM-containing NcLPMO9E-treatment had an average width of around 9 nm (Fig. 3g). This was also confirmed by AFM analysis as shown in Fig. S7, ESI. † The NcLPMO9F oxidised nanofibres had a width ranging from 2 to 8 nm and an average value of 4.3 ± 1.5 nm, while NcLPMO9E oxidised nanofibres had a width from 3 to 14 nm and an average of 6.7 ± 2.5 nm, respectively. The width measured by TEM was slightly higher due to the negative staining. The length of both nanofibres was longer than 1 µm and the fibre ends were not apparent in the TEM images due to aggregation. The widths of these nanofibres are significantly smaller compared to what has been previously observed for kraft pulp treated with other LPMOs. 20,22 The widths are also smaller than that of the nanofibres prepared by using a pre-treatment with glycoside hydrolases followed by much stronger mechanical disintegration using a microfluidizer (10-40 nm). 12,35 These results suggest that the N. crassa LPMOs made the holocellulose more susceptible to mild mechanical homogenisation (25 000 rpm, 2 min) and facilitated the production of CNFs. The yield of the nanofibres was assessed by mass recovery compared to the control sample. The yields of nanofibres from the non-modular NcLPMO9F and CBM-containing NcLPMO9E treated holocellulose were 81.8 ± 5.0% and 65.4 ± 0.5%, respectively. The analysis of soluble sugar in the supernatant confirmed that the lower nanofibre yield obtained with the CBM-containing NcLPMO9E correlated with a higher amount of released glucose equivalents (25.2 ± 1.2 mg g −1 of cellulose), compared to the non-modular NcLPMO9F, which released 10.7 ± 0.9 mg g −1 glucose equivalents. The different mass recoveries and varying amounts of soluble sugars most likely resulted from different modes of action of the modular and non-modular enzymes. A similar effect of the CBMmodule on the sugar release has been recently reported by Chalak et al. 36 To assess the predominant mechanism by which the N. crassa LPMOs were able to enhance the mechanical fibrillation of holocellulose, total sugar analysis, X-ray diffraction (XRD) and ζ-potential measurements were carried out to determine the sugar compositions, crystallinities and surface charges of the LPMO-oxidised fibres, respectively. Before the LPMO reaction, the holocellulose was autoclaved and the hemicelluloses (59 wt% of xylan and 22 wt% of glucomannan) were partially removed (Table S1, ESI †). The purpose was to expose the cellulose surface and avoid hemicellulose precipitation on cellulose after the delignification step using PAA, and increase the LPMO accessibility. The hemicellulose content did not change significantly after the LPMO treatment (Table S1, ESI †), which is different from the TEMPO-mediated oxidation method where the hemicelluloses are partially removed after the oxidation step. 37 The glucose content of the LPMO-treated holocellulose decreased significantly due to the difficulties in hydrolysis of aglycones after the formation of aldonic acids at the reducing ends. The results from XRD revealed that both LPMOs induced a moderate reduction by 2% in the cellulose crystallinity compared to the control sample (Fig. 4), indicating that the LPMOs mainly targeted the crystalline regions of cellulose. 38 However, no significant difference in crystallinity was found between the holocellulose nanofibres oxidised by the CBM-containing NcLPMO9E and those oxidised by the non-modular NcLPMO9F. The surface charges obtained from the ζ-potential measurements showed a 1.5-fold increase in both enzymetreated holocellulose fibres compared to the non-enzymatically treated fibres (Fig. 5). The obtained ζ-potential values (−37.6 ± 3.1 mV for NcLPMO9E and −37.7 ± 2.4 mV for NcLPMO9F) suggested stable colloidal suspensions, which were speculated to result from the presence of deprotonated C1-carboxyl groups on the fibre surfaces. 19 The ζ-potential of the control holocellulose fibres (−25.4 ± 1.5 mV) corresponded well to previous results obtained with the same type of fibres. 27 Conductometric titration on the fibres also revealed that the LPMOs increased the carboxyl content by 27% and 88% compared to the control. The obtained carboxyl contents were 0.38 and 0.53 mmol g −1 for NcLPMO9E and NcLPMO9F, respectively (Fig. 5). The carboxyl content was observed to be 38% higher when the fibres were treated with the CBM-lacking NcLPMO9F, compared to the modular LPMO. The difference in carboxyl content could result from more restricted movement of the CBM-containing NcLPMO9E on the cellulose surface. However, as inferred from the ζ-potential, the particle charge density appeared to be similar for the nanofibres produced by both enzymes. This is probably due to the fact that the width of the nanofibres produced by NcLPMO9E was almost twofold compared to that of the nanofibres produced by NcLPMO9F, as measured by TEM (Fig. 3). These results indicate that the fibrillation was facilitated by better dispersion of the fibres resulting from the repulsion between the LPMO-generated C1 carboxyl groups on fibre surfaces, rather than the disruption in the ordered crystalline areas of the fibres by the LPMOs. The carboxyl content obtained by using the CBM-lacking NcLPMO9F was 0.53 mmol g −1 , similar to that of CNFs generated by partial carboxymethylation of fibres followed by fibrillation, which has a charge of 515 μequiv. g −1 . 7 This value is five times higher than the previously reported value when LPMO was used together with endoglucanase and xylanase on hardwood kraft pulp. 22 Interestingly, TEMPO-mediated oxidation of spruce was also found to be more efficient with higher carboxylate contents and higher stability of the nanofibre dispersion than the oxidation of eucalyptus. 37 The presence of the LPMO-induced C1-carboxyl groups in the nanofibres was also confirmed by FTIR. The FTIR spectra showed distinct changes in the region associated with carboxyl groups derived from the LPMO treatment (Fig. 6). The observed peaks at 1733 cm −1 were attributed to the CvO stretching frequency of carbonyl groups from the partially acetylated galactoglucomannans in the holocellulose. A new peak at approximately 1595 cm −1 was observed for the LPMOoxidised holocellulose, which corresponds to the asymmetric stretching vibrations of the deprotonated carboxyl groups from the aldonic acids at the reducing ends. The new peak was more distinct for the NcLPMO9F-oxidised nanofibres, which correlates with the conductometric titration results. ## Structure and mechanical properties of the nanopapers The nanopapers were prepared from water suspensions of the nanofibres by vacuum filtration akin to the papermaking procedure. As shown in Fig. 7c and e, a fibrous nanofibril network structure with a random-in-plane orientation was present on the surfaces of the nanopaper prepared from the LPMO-oxidised nanofibres, while large cellulose fibril aggregates were observed in the control nanopaper from holocellulose without the LPMO treatment (Fig. 7a). A uniform layered structure was apparent in the cross-sections of the nanopapers prepared from the LPMO-oxidised nanofibres, resulting from nanofibre deposition during vacuum filtration (Fig. 7d and f ). The crosssection of the control nanopaper also showed a layered structure, but the layers were heterogeneous with large aggregates and voids (Fig. 7b). This is due to the fact that the holocellu- lose was only partially fibrillated by mild homogenisation without LPMO treatment. Fig. 8 shows typical stress-strain curves under uniaxial tension for the holocellulose nanopapers, and the physical and mechanical properties of the nanopapers are summarised in Table S2, ESI. † The mechanical performance of the nanopapers was substantially improved when the nanofibres were prepared using the LPMOs, particularly with NcLPMO9F which did not contain a CBM. The tensile strengths were 257.0 ± 6.2 MPa and 262.2 ± 10.1 MPa for NcLPMO9E and NcLPMO9F, which are 73% and 76% higher than that of the control (148.7 ± 4.2 MPa), respectively. The strain-to-failure values of the nanopapers prepared from the LPMO-oxidised CNFs were also higher (4.2 ± 0.7% and 3.7 ± 0.4%) compared to the control paper, which failed at 2.6 ± 0.3%. The improved mechanical properties derived from the LPMO oxidation were attributed to a better fibrillation and production of nanofibres with a width below 10 nm, which led to enhanced entanglement of the nanofibre network and allowed inter-fibril slippage during uniaxial tension before fibril breakage. This resulted in higher strain-to-failure and higher tensile strength after strain hardening, and thus higher toughness, i.e. work of fracture, for the nanopapers. This is consistent with the scaling law of the mechanical properties of cellulose nanopapers: the smaller, the stronger and the tougher, as reported by Zhu et al. 39 The tensile strength (262 MPa) and modulus (16.3 GPa) of the nanopaper obtained with NcLPMO9F-oxidised CNFs were also higher than those of the MFC films prepared by enzymatic pre-treatment using glycoside hydrolases, which had a tensile strength of 214 MPa and a modulus of 13.2 GPa, 35 owing to a much thinner width of LPMO-oxidised CNFs compared to the MFC nanofibres prepared using cellulases. ## Impact of the carbohydrate-binding module The CBM-lacking NcLPMO9F efficiently introduced carboxyl groups on the cellulose fibre surfaces, resulting in better fibrillation and production of nanofibres with a thinner width compared to the CBM-containing NcLPMO9E. This also resulted in better mechanical properties of the prepared nanopapers. While the surface charge densities (ζ-potentials) of the prepared nanofibres were similar after oxidation with both C1active LPMOs from N. crassa, the carboxyl contents were significantly different, as the CBM-lacking NcLPMO9F was able to generate 38% more carboxyl croups (0.53 mmol g −1 ) than the CBM-containing NcLPMO9E (0.38 mmol g −1 ). This difference in activity is most likely due to enhanced movement of the CBM-lacking enzyme on the cellulose surfaces, leading to a less localised oxidation pattern during the reaction on cellulose. Indeed, it has been previously shown that a CBM might increase the residence time of LPMOs on the substrate, confining the oxidation to a more specified area, 26 indicating that an LPMO without an appended CBM might be free to perform oxidation in a more disperse manner over the entire cellulose surface. It was also found that the nanofibre yield was higher when the fibres were treated with the non-modular NcLPMO9F, while the CBM-containing NcLPMO9E was able to release more soluble sugars from the holocellulose substrate. This is most likely due to the localisation of the CBM-containing enzyme and subsequently higher amount of proximate cleavages of the cellulose chains enabling the release of shorter products compared to the freely moving enzyme. Similar ζ-potentials and different carboxylate contents after the LPMO oxidation might be due to the fact that the ζ-potential reflects the surface charge density rather than the total charges, as a higher total charge in nanofibres is often related to a thinner width and a larger surface area. Such a result was also observed in TEMPO oxidation of cellulose from different origins, where the selective oxidation of C6 primary hydroxyls yielded nanofibres with similar ζ-potentials regardless of the total carboxylate contents. 40 ## Conclusions In summary, we have successfully produced CNFs from delignified softwood fibres by a completely green and energyefficient method utilising oxidative enzymes, C1-active LPMOs from N. crassa. Lignin was first removed from the spruce fibres with PAA, followed by enzymatic oxidation of cellulose with the LPMOs. After these pre-treatments, CNFs with a thin width (<10 nm) could be prepared by a rather mild mechanical homogenisation. Thinner nanofibres (width of 4.3 ± 1.5 nm) with a higher carboxylate content (0.53 mmol g −1 ) were obtained from spruce holocellulose by using a CBM-lacking LPMO, NcLPMO9F, compared to the modular LPMO, NcLPMO9E, owing to the free movement of the CBM-lacking enzyme and more homogeneous oxidation over the fibre surface. The nonmodular NcLPMO9F also enabled the production of higher yields of nanofibres, 82% from the original fibre mass. These results suggest that LPMOs without CBMs might be better suited for cellulose nanofibrillation purposes than CBM-containing LPMOs. The LPMO-oxidised nanofibres were further used to prepare cellulose nanopaper films. The films showed a high tensile strength of 262 MPa and Young's modulus of 16.2 GPa under uniaxial tension, higher than those for nanofibres from enzymatic treatment with cellulases and TEMPOmediated oxidation. These results demonstrate the unique utility of LPMOs in cellulose nanomaterial preparation, reducing the use of toxic chemicals and high energy for nanofibrillation. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Lytic polysaccharide monooxygenase (LPMO) mediated production of ultra-fine cellulose nanofibres from delignified softwood fibres", "journal": "Royal Society of Chemistry (RSC)"}

molecular_weight_distribution_of_kerogen_with_maldi-tof-ms

## Abstract: Kerogen is an amorphous organic matter (AOM) in fine grain sediments, which produces petroleum and other byproducts when subjected to adequate pressure and temperature (deep burial conditions). Chemical characteristics of kerogen by considering its biogenic origin, depositional environment, and thermal maturity has been studied extensively with different analytical methods, though its molecular structure is still not fully known. In this study, conventional geochemical methods were used to screen bulk rock aliquots from the Bakken Shale with varying thermal maturities. Organic matter was isolated from the mineral matrix and then a mass spectrometry method was utilized to quantify molecular weight distribution (MWD) of four different kerogens at various thermal maturity levels (immature to late mature). Furthermore, to complement mass spectrometry, Fourier transform infrared (FTIR) spectroscopy was employed as a qualitative chemical and structural investigation technique. The MWD of four samples was obtained by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, and the results are correlated with the absorption indices (CH 3 /CH 2 ratio and aromaticity) calculated from the FTIR attenuated total reflectance (ATR) method. The results showed when the degree of maturity increases, the aliphatic length shortens, and the branching develops, as well as the aromatic structure becomes more abundant. Moreover, based on the MWD results, higher maturity kerogen samples would consist of larger size molecular structures, which are recognized as more developed aromatic, and aliphatic branching stretches. The combination of infrared spectroscopy (AFT-FTIR) and mass spectrometry (MALDI-TOF) provided MWD variations in kerogen samples as a function of maturity based on varying absorption indices and revealed the rate of change in molecular mass populations as a function of thermal maturity. ## Introduction In the past few years, there has been growing attention to organic-rich shale reservoirs due to their importance in hydrocarbon production and CO 2 sequestration . The high abundance of total organic carbon (TOC) in shale layers, which are deposited in these rocks hundreds million years ago, would lead to the generation of oil and gas through thermal maturation. This process takes place due to the exposure of rocks containing organics to adequate temperature and pressure as the depth of burial increases. During this geologic process, the organic matter known as kerogen, which has a complex and amorphous molecular structure, breaks down and goes through a significant structural and compositional transformation . Thermal maturity of organic matter is an important phenomenon that still requires further studies at the molecular scale from various perspectives to help us better explain this process which enables us to better explain the phenomenon that'd generate hydrocarbons from the organic matter. In this regard, there has been a limited number of studies to investigate molecular structure of organic matter (kerogen) at different stages of thermal maturity using various spectroscopy methods such as: 13-C nuclear magnetic resonance ( 13 C-NMR), X-ray photoelectron spectroscopy (XPS), and X-ray absorption near-edge structure (S-XANES) , Raman , and infrared (IR) spectroscopy. Even though these studies have clarified the structural evolution of kerogen to some extent, this task still demands further analysis since kerogen has different biogenic origins (e.g., lacustrine and marine algae and plankton or terrestrial higher-order plants). One of the methods in better characterizing the molecular structure of organic matter during thermal maturation is measuring its molecular weight (MW) and relating that to the existing chemical compounds or functional groups at every stage of maturity. Mass spectrometry (MS) technique has been widely utilized in biology and chemistry for decades to identify the molecular weight of various materials such as proteins and synthetic polymers, which is now extended to petroleum and fossil fuels arena. This being said, in the first attempt to characterize kerogen with MALDI by Li et al., , a relationship between the minimum laser power sufficient for activating the MALDI-ionization process and the thermal maturity level (reported in the age of samples) of kerogens, was established. This study was later continued by Herod et al. , where several isolated kerogens from various thermal maturities and mixed depositional environments (marine, lacustrine and terrestrial) were analyzed by MALDI-TOF-MS (Matrix-assisted laser desorption ionization-time of flight mass spectrometry). Authors found that spectra of isolated kerogens from 'younger' deposits (less mature or less buried) showed larger molecular mass distributions compared to the spectra obtained from extracts of older sediments. Furthermore, kerogen molecular weight estimation measured by MALDI-TOF was in agreement with the values obtained by conventional techniques, such as SEC-chromatograms . The issue with these two studies that employed MALDI spectroscopy is that neither of them studied a single origin kerogen specimen, and there was a mixture of samples without a systematic maturity trend among them. The importance would be to reveal the rate of molecular weight alteration at specific thermal maturity stage and the rate of change in the population of molecular fragments, which is missing. Recently, molecular weight of the organic matter that was artificially matured in a semiopen pyrolysis system to various levels of thermal maturities from Green River Shale Formation was quantified using various mass spectrometry methods including laser desorption laser ionization (L 2 MS), surface assisted laser desorption ionization (SALDI) , and matrix-assisted laser desorption ionization (MALDI) . Results showed a systematic change in molecular weights and correlation between elemental analysis and mass spectroscopy results. It should be mentioned that although chemical information that is provided by MALDI-TOF is considered accurate for both identification and in quantification of chemical substances in a material, it would be difficult to determine the existing functional groups (e.g. carboxyl/amide/carbonyl/methylene/methine functional groups), and non-degradable fractions in the studied compounds. This adds a drawback to MALDI-TOF-MS, which makes it inevitable to be coupled with other analytical techniques such as IR, Raman or nuclear magnetic resonance (NMR) spectroscopy, as was done on these limited studies of kerogen, to provide us more detailed information about the components and materials that are being analyzed . In this study, four isolated (extracted) kerogens from the mineral matrix at four different stages of natural thermal maturity (immature to late mature) from the Bakken Shale were examined by MALDI-TOF-MS combined with FTIR spectroscopy for molecular weight quantification along with chemical and structural analysis. This study enabled us to identify structural characteristics of organic matter as it undergoes thermal advance in nature to relate these variations to both qualitative information from chemical structures and quantitative molecular weight distributions (MWD). Additionally, the outcome helps us to delineate the rate of change in molecular weight intervals populations as thermal maturity progresses. ## Samples Kerogen samples (type II representing a marine environment) were collected from the Bakken Formation which is one of the largest unconventional shale oil plays in North America and is currently being studied for potential CO 2 -EOR and sequestration . Samples were retrieved from four different wells that are drilled at different locations in the Williston Basin in ND, where the Bakken has different thermal maturities. Aliquots of the bulk samples were examined with programmed pyrolysis (Rock Eval 6) for determining the thermal maturity level of the samples. To be more accurate, samples were also analyzed for the solid bitumen reflectance in the absence/scarcity of the vitrinite maceral, and the measured values then converted to the equivalent vitrinite reflectance using the appropriate conversion equations. Results are summarized in Table 1, where it shows sample A is immature and sample D is late mature. Selected kerogen samples were isolated using HCl and HF, and liquid state hydrocarbon was extracted before MALDI-TOF-MS experiment. Further details about these samples and conventional geochemical analysis, organic matter extraction and the Bakken Formation can be found in Khatibi et al. , and Abarghani et al. . ## ATR-FTIR Infrared spectra of kerogen samples were recorded in adsorption between 450 and 4000 cm -1 wavelengths using a Thermo Fisher Scientific, Nicolet iS50 FTIR Spectrometer. Four kerogen samples were pulverized using a ball mill before the characterizations. Fourier transform infrared (FTIR) spectroscopy using attenuated total reflectance (ATR) was utilized to analyze the isolated kerogens. Unlike transmittance FTIR, ATR does not require that kerogen samples to be mixed with potassium bromide and form into pellets under high pressure, which reduces the time needed to prepare the samples. Kerogen samples were placed in contact with an internal reflection material, and IR spectra were obtained based on the excitation of the molecular vibrations of chemical bonds by the absorption of light. The stretching absorption of a vibrating chemical bond is observed at higher frequencies (wavenumbers) than the corresponding bending or bond deformation vibrations . Although the bending vibrations can support to obtain more details from the chemical structure, we'd face limitations in the analysis of integrated bands in bending vibration areas because kerogen is a complex macromolecule. In this study, it was decided to analyze the C-H set of stretch vibrations that are observed in kerogen samples, which is also done in previous studies . It should be noted that recorded bands of the absorption were identified through comparison with the published spectra shown in table 2. Table 2 Saturated aliphatic group and aromatic ring group frequencies . -H). These are the most common characteristics of an organic compound containing the aliphatic fragments. The intensities observed are corresponding to the asymmetric and symmetric stretching and bending of the C-H bonds of the central carbon atom. The C-H stretching vibrations for the saturated aliphatic occur between 3000 and 2800 cm -1 , and the aromatic ring stretch vibrations occur between 3130 and 3070 cm -1 . IR structural evaluations, which has been established in earlier studies , were calculated from acquired spectra. The intensities of bands assigned by the functional groups were deconvoluted through a curve fitting procedure . The CH 3 /CH 2 ratio (I (2970-2950) /I (2935-2915) ) indicates the average chain length of aliphatic and the degree of chain branching. Moreover, aromaticity index (I (3130-3070) /(I (2970-2950) +I (2935-2915) ) represents the degree of aromatic structures to aliphatic chain structures . These ratios were considered to better analyze MWD that is acquired by MALDI-TOF at different stages of maturity and were related to alterations in the overall molecular weight of the specimens. ## MALDI-TOF-MS The AB SCIEX TOF/TOF 5800 mass spectrometry system was used for the identification and relative quantitation of kerogen molecular weight. A dilution series of kerogen powder were prepared for two different test setups, in the absence and the presence of the matrix. In the second set up where the matrix is presented, dilutions were mixed with α-Cyano-4hydroxycinnamic acid (α-CHCA) in 1:1 ratio. The matrix, α-CHCA, is commonly used in conjunction with organic molecules, particularly for relatively higher-weight ones . Each sample (1 μL) was spotted onto a standard stainless steel plate and allowed to air-dry. Spectra were acquired with MS reflector mode for 1,000 shots, and the analysis was conducted over a range of 60 to 5000 Da by use of an adjusted accelerating voltage. It should be noted that MALDI-TOF-MS analysis has been utilized to characterize kerogen in a very limited scope. However, previous studies conducted on heavy fossil fuels and aromatic compounds have demonstrated the applicability of MALDI-TOF to characterize similar macromolecules . These studies suggested that the matrix would not always be necessary to obtain MALDI mass spectra, because lower molecular weight aromatics, can play the role of the matrix for the ablation of the higher molecular weight compounds. Even though suggestions have been made regarding the matrix, its role in the analysis of kerogen has not yet fully understood. Therefore, in our study, the MALDI-TOF spectrum of kerogen samples at various maturations is obtained in the absence and presence of the matrix, and the results are compared and reorganized to delineate the MWDs more accurately and help us investigate the necessity of having a matrix. In order to confirm accurately the compound measurements validating the technique, we performed control procedures for MALDI-TOF . MS technique is possible to detect varying components (fragments) of a complex structure. Hence, we made a 1:1 and 1:1:1 mixture of model compounds including kerogens spanning the mass range occurred for kerogen samples. Then, we checked that MALDI can detect all of these model compounds with about the same cross-section, and that the results give a spectrum with peaks regarding references and kerogens. Additionally, we confirmed that the measured molecular weight (WD) is not adjusted changed by the laser pulse energy. Furthermore, to find optimum laser power for kerogen samples, the multiple tests with model compound were performed to obtain clear and exact signals. Hence, we set of laser intensity at 5000 (instrument-specific units) for the absence of matrix with ??? Hz pulse rate and at 4200 for the presence of matrix with ??? Hz pulse rate. ## Results and Discussions Since kerogen consists of amorphous and complex chemical compounds, spectroscopy analysis would be inadequate to evaluate its chemical structure quantitatively. As a result, in this study, FTIR and MALDI-TOF-MS are integrated, to assist the chemical characterization of this geomacromolecule. We have interpreted spectra obtained from several extracted kerogens at different stages of thermal advance to reveal the relationship between various functional groups and molecular weight distributions and chemical alterations in a type II kerogen. ## Structural Characterization by FTIR Considering the FTIR intensities, a given absorption band that is assigned to a specific functional group has a direct relationship with the amount of that functional group overall existence in the structure of the macromolecule . We noted that the methyl group (CH 3 ) is considered as the terminal for the aliphatic chain, where a methylene group (CH 2 ) is linked to a neighboring group which can also be attached to another methylene group. This framework presents a carbon to carbon bond, forming the aliphatic chain. Because each section of this group has its corresponding C-H stretching and bending vibrations, identification of this group and skeletal frequency would help to estimate a carbon-based organic compound such as the kerogen. Especially, a strong methylene band (2935-2915 cm -1 ) and a weak methyl band (2970-2950 cm -1 ) indicates a longer-chain aliphatic structure. In contrast, the strong methyl band and a comparatively weaker methylene band indicate shorter-chain branching. Figure 1 represents the intensities measured for frequency absorptions assigned to methyl, methylene, and aromatic on a relative basis in the spectrum of each kerogen samples. The reason for the overlapped original spectra is due to the same amount of energy required for various vibrations . To analyze each functional group at desired frequencies, the FTIR spectrum area from 2700-3200 cm -1 was fitted (Figure 2). In order to do so, intensities and frequencies of the bands in the desired regions were estimated by curve fitting where the peak separation and quantitative calculation were performed using the Fourier self-deconvolution method. In other words, when kerogen thermal maturation is advanced, relatively, the aliphatic chain length will become shorter, and branching is developed. As a result, the aromaticity index (Figure 3-Right), representing the abundance of aromatic structures in the kerogen molecule is also increased as thermal maturity is increased. Furthermore, at a higher degree of thermal maturation where the abundance of aromatic rings and the reduction of methyl/methylene is observed, it is expected this phenomenon to result in a significant change in the molecular weight of the remaining substance. The observed trends in Figure 3 are consistent with the results from previous findings, which proved that the aliphatic chain lengths have shortened and increased aromaticity during the maturation . It should be stated again during the maturation process, as a result of organic matter exposure to temperature, hydrocarbons are generated and expelled from the organic matter and the remaining molecule is undergone major molecular alteration as being investigated here. ## Molecular Characterization by MALDI-TOF MS The ionization process in MALDI-TOF proceeds through the capture of a proton, which forms a charged adduct with the molecular species of the sample. As the number of charged adducts reflects the signal intensity of the molecular weight, quantitative MWD analysis of any chemical compound is conducted. Four kerogen samples of this study were analyzed by MALDI-TOF to attribute maturation effects on the MWD of organic matter. Furthermore, systematic naturally matured kerogens with single biogenic origin have rarely been researched with MALDI-TOF, which limits the guidelines for an optimal and effective matrix selection based on the literature. Hence, in our study, kerogen samples were examined in both the presence and in the absence of the matrix (α-CHCA) for comparison and more accurate results. Figure 4 exhibits the mass spectra of kerogen samples obtained both in the presence and in the absence of the matrix. In all of these spectra for the presence of the matrix, unidentified highintensity signals were observed in the relatively low-mass region between 200 and 400 m/z. Also, we could not find a regular ionization and pattern overall. This confirms that kerogen does not have a specific chemical structure and contains various functional groups which are differently ionized, which hinders the signal interpretation from analyzing molecules in kerogen. Also, overall spectra representing the samples in Figure 4 does not show a particular relationship with signal intensities unlike polymers and proteins. However, we found that signal intensity changes with respect to the degree of thermal maturation. As a result of this alteration in the intensity, it is conclusive that the maturation process would change the molecular structure of kerogen in terms of the molecular mass of the organic matter. This infers that the chemical structure of kerogen has been evolved while bond-breaking and forming have occurred . Thus, we can at least conclude that MALDI result confirms the chemical structure changes by maturation. Sample A, the immature kerogen, exhibits signals between 250 to 350 m/z with the highest intensity regardless of the presence or absence of the matrix. In addition, when the maturation is progressed, strong signals become evident in the presence of the matrix, near 170 m/z and 370 m/z. On the other hand, in the absence of the matrix, the highest intensity signals are recorded in the similar mass region in all maturation stages, while molecular alterations have occurred in the higher mass region of the spectra. When the degree of maturation is higher, it was seen that signals become more transparent in the relatively heavier mass regions. This observation was not quite similar to the changes of signals in the presence of the matrix. In the latest experimental conditions, the spectra do not display any clear signals over 800 m/z. Therefore, it is probably understood not only that α-CHCA is not the best matrix to investigate complex mixtures of organic matter (kerogen) for the higher mass regions, but also that an excess of matrix inhibits to precisely conduct the ionization process. Based on this, in our study, the mass spectra without any matrix was utilized for the MWD analysis of the kerogen samples to avoid erroneous results. 4). Furthermore, this particular partitioning (distribution) of MWs is decided through trying different molecular weight intervals and plotting MWDs of these four different samples when the most distinct and meaningful trends are delineated. We also note that the quantitative value in Figure 5 disregards total MW and solely pertains to the comparative change of MWD regarding the maturity since the area ratio was adopted. It was found that, for relatively immature kerogen (A and B), around 60% of the signals are under 1 kDa range, whereas the majority of recorded signals for relatively higher maturity kerogen samples (C and D) are found between 1 to 2 kDa range. Immature sample (A) has the highest signal in the range under 500 Da; this indicates that relatively lighter molecular structures exist in kerogen compound prior to maturation. Meanwhile, the MWD for samples A and B continuously decreases in higher mass ranges. Remarkably, as the degree of the maturation increases, the higher signal distribution in the heavier range is observed in Figure 5. It is verified that the MWD variation is certainly a function of the degree of thermal maturation, reflecting the existence of separate molecules detected through the MALDI-TOF-MS process. Although in all kerogen samples, the distributions are almost similar in the ranges over 5 kDa, generally higher maturity kerogens should contain larger size molecular compounds and fragments than lower maturity ones. Additionally, the rate of change in MW is also investigated for each maturity stage over the entire range from 0 to 5 kDa. The smaller graph in Figure 5 is obtained through taking the derivatives of integral of MALDI-TOF spectra to represent the rate of change in the quantity of that specific MW interval for each sample. The most important aspect of this graph is that at a specific MW (around 2.5 kDa) all of four samples start to show relatively similar increasing trends in the rate of MW quantities, and then this becomes constant regardless of their maturity at higher MW region (> 4 kDa). This means there isn't any change in the molecule's population at higher weights regardless of the maturity. In sample A, the most immature kerogen, relatively large negative rate of change in the lower MW region is observed and the degree of change presents a decreasing trend as the molecular weight increases. The rate of change in sample B rapidly decreases up to around 1.5 kDa MW, then alleviates in higher MW ranges showing similar positive rate over 2 kDa similar to sample A. Unlike immature kerogens, two higher maturity specimens (C and D) have comparable MW rate of changes over the entire MW range. Positive change is observed near lower MW region (< 2kDa), which means the molecules in 1 to 2 kDa are becoming more abundant than other MW regions, then the rate of popularity decreases and becomes unchanged like other samples. The difference in MWD and rate of change in MWs between the immature (A and B) and late mature (C and D) kerogens could refer to the boundary from pre-oil to oil generation window and to the phenomenon where hydrocarbons are generated and expelled from the OM through a nonlinear trend . This major change in molecular structure corresponds to the onset of oil-window, when a major alteration is expected to take place (molecular structure and molecular weight) around 2.5 kDa where all MW rate of change graphs in the smaller image in Figure 5 coincide and start to perform with a similar increasing trend. These two graphs combined illustrate, although we have a relatively linear increase in thermal maturity index (in Table 1) from samples, the variations are nonlinear which is at the onset of oil generation. The MWD results can be correlated to the structural indices from the FTIR as well. As we discussed in this study, the carbon chain (aliphatic) length has shortened and branching has developed as the degree of maturation is increased, as well as the aromatic structure becomes more abundant. From Sample A to D, the abundancy in lower MW intervals decreases which corresponds to the increase in CH 3 /CH 2 ratio. On the contrary, the higher quantity of heavier MW intervals is correlated to the aromaticity index where it increases from Sample A to D. These two combined, as a function of increasing the maturity, kerogen should contain a greater number of heavier molecules that are more developed aromatic and aliphatic branching structures. We state that the trend could not directly convey the change of total weight of kerogen. In other words, the structural changes that kerogen undergoes due to the abundance of the aromatic and shorten aliphatic chain length as its maturity progresses is well correlated with the heavier molecular weight of the product. ## Conclusion This article presented a characterization of kerogen at 4 different maturity levels using infrared spectroscopy and mass spectrometry. Spectrums obtained by FTIR and MALDI-TOF from organic matters (kerogens) isolated from the bulk rock aliquots collected from the Bakken Shale were analyzed. The results were useful to better understand hydrocarbon generation and maturation process through investigation of a chemical structure and molecular weight variations during the expulsion of hydrocarbons as maturity progresses. Based on the results, the following conclusions can be made:  MALDI-TOF spectra in the presence of the matrix (α-CHCA) show irregular ionization and pattern between 200 and 400 m/z for all maturity stages. It follows that kerogen does not have a specific chemical structure and hardly can be analyzed regarding its composition and molecular weights for specific fragments in the presence of the matrix. Since kerogen is a complex macromolecule, the matrix (α-CHCA) maybe inappropriate to be used during the ionization/mass spectroscopy experiments and existing lower MW aromatic fragments can take the role of the matrix.  It was found while the increase in the maturation takes place, the results of MALDI-TOF in the absence of the matrix demonstrates that kerogen consists of heavier molecules. The result without the matrix shows obvious signals in the heavier mass regions (over 1,000 m/z). Based on this, MWD was generated from MALDI-TOF spectra in the absence of matrix.  The MWD variation provided different molecular weight ranges at different thermal maturity stages in the absence of the matrix. Relatively immature kerogens exhibited around 60% of the signals under 1 kDa range, with the strongest signals in the range under 500 Da, whereas the majority of signals of mature kerogens were found between 1 to 2 kDa range.  In terms of major changes (molecular structure and molecular weight) through maturation, the boundary from pre-oil to oil generation window can be distinguished based on the rate of change of integrated MALDI-TOF signals with respect to molecular weight. After a specific MW (over 2.5 kDa), all kerogens would show a similar increasing trend in the rate of change of the population of MW intervals.  The shorter aliphatic chain length and abundant aromatic structure are expected and delineated based on FTIR structural indices (CH 3 /CH 2 and Aromaticity) when the degree of maturity increases which was correlated to the MWDs. The structural kerogen alteration in the abundance of the aromatic and shorten aliphatic chain length as its maturity progress tends to make the weight of molecules/fragments in kerogen heavier.

chemsum

{"title": "Molecular Weight Distribution of Kerogen with MALDI-TOF-MS", "journal": "ChemRxiv"}

anticancer_metallohelices:_nanomolar_potency_and_high_selectivity

## Abstract: A range of new helicate-like architectures have been prepared via highly diastereoselective self-assembly using readily accessible starting materials. Six pairs of enantiomers [Fe 2 L 3 ]Cl 4 $nH 2 O (L ¼ various bidentate ditopic ligands NN-NN) show very good water solubility and stability. Their activity against a range of cancer cell lines in vitro is structure-dependent and gives IC 50 values as low as 40 nM. In an isogenic pair of HCT116 colorectal cancer cells, preferential activity was observed against cell lines that lack functional p53. Selectivity is also excellent, and against healthy human retinal pigment epithelial (ARPE19) and lung fibroblast (WI38) cells IC 50 values are nearly three orders of magnitude higher. Cisplatin is unselective in the same tests. The compounds also appear to have low general toxicity in a number of models: there is little if any antimicrobial activity against methicillin-resistant Staphylococcus aureus and Escherichia coli; Acanthamoeba polyphaga is unaffected at 25 mg mL À1 (12.5 mM); Manduca sexta larvae showed clear evidence of systemic distribution of the drug, and rather than any observation of adverse effects they exhibited a significant mean weight gain vs. controls.Investigation of the mode of action revealed no significant interaction of the molecules with DNA, and stimulation of substantial cell death by apoptosis. ## Introduction The main purpose of current anticancer therapies is to eradicate tumour cells without damaging overall patient health. However, side effects limit the dosage of chemotherapeutic drugs which may be safely applied, and as a result, cancer cells often remain. This leads to poor outcomes in the clinic and the evolution of drug-resistant tumours. 1 Hence, while the potency of a drug is a very important consideration, drug selectivity towards cancer cells is key to ensuring both safety and effectiveness. 2 While we might hope that more effective cancer chemotherapies would come from drugs designed to address specifc biomolecular targets, 2 this is far from uniformly the case. 3 Such drugs may be too targeted since tumours can circumvent the blockade of a specifc pathway by switching to anotherso-called tumour plasticity. 4 Compounds with polypharmacology (action against multiple targets) are thus currently of considerable interest to the pharmaceutical industry. This coincides with the resurgence of phenotypic drug discovery, where the targets of a drug are established after the observation of the useful biological effect. This strategy has led to a disproportionately high number of frst-in-class drugs with novel mechanisms of action (1999-2008) 8,9 The accompanying challenge for synthetic chemistry is to discover, perhaps without reference to some specifc biomolecular target, new classes of drug candidates which are both potent and selective. Lehn recognized the potential of helicates in medicinal chemistry, 10 and this was borne out in early studies, particularly in the area of cancer. We have argued, 16 however, that in order for helicates to be capable of translation to the clinic a number of criteria need to be addressed: optical purity and stability, solubility and chemical stability in water, availability on a practical scale, and synthetic diversity. Our recent work has attempted to address these matters 17 using a new strategy whereby the absolute confgurations of individual metal centres are controlled 18 and linked together to form the prototype helicate-like architectures of Fig. 1. Of these flexicates, 19 [Fe 2 L 1 3 ] 4+ contains a diamine linker while [Fe 2 L 2a 3 ] 4+ is based on a dialdehyde. 19 Promising results were reported in a number of disease areas, 16, including good activity against a range of cancer cell lines. 20 Here we report the discovery of a new series of highly potent (40 nM) anticancer compounds of the dialdehyde class related to [Fe 2 L 2a 3 ] 4+ that preferentially kill cancer cells that lack functional p53, are nearly three orders of magnitude less toxic to healthy human cell lines tested and have low toxicity to microbes, amoeba and caterpillar larvae. While DNA does not appear to be the target, the compounds are triggering signifcant apoptotic cell death as part of their mode of action. ## Synthesis of ligands and Zn II systems The dialdehyde units of Fig. 2(a) and (b) include various linker rigidities and orientations designed to probe structural viability and biological activity. They were synthesized via simple etherifcations of 5-hydoxypicolinaldehyde. 22 Treatment with Zn(ClO 4 ) 2 $6H 2 O and (R)-1-phenylethan-1amine, in appropriate proportions, led to the rapid selfassembly of the bimetallic flexicates in acetonitrile solution at ambient temperature. For the majority of these new Zn II complexes NMR spectra indicated that within the limits of the experiment single diastereomers were formed (vide infra). The sole exception was the 1,3-phenylene system L Zn -[Zn 2 L 2e 3 ][ClO 4 ] 4 $4H 2 O, which gave more complex 1 H NMR spectra [Fig. 2(c)]. At 253 K the phenethylamine methyl group doublet region 1.4-1.7 ppm contains one more intense doublet and two broader signals in the ratio ca. 10 : 1 : 1. The proportion of the minor species increases with temperature and the resonances sharpen somewhat, such that by 313 K two of the smaller doublets corresponding to the minor species are relatively sharp and resolved while a third overlaps with the main resonance. By 353 K the smaller peaks had again broadened considerably and the ratio of the two sets of resonances was ca. 10 : 9. The imine region (8.5-7.6 ppm) behaved in a corresponding manner (253 K, three peaks in ratio 10 : 1 : 1 : 1; 353 K, ratio 10 : 3 : 3 : 3). These observations are consistent with the presence of two speciesone of high-symmetry and one lowin thermodynamic equilibrium (ratio ca. 1 : 0.3 at low temperature, increasing to almost 1 : 1 at high temperature) but with the involvement of other related conformers particularly at higher temperatures. The processes leading to the observed NMR behaviour may correspond to exchange between these conformers, or indeed between isostructural low symmetry species. While the spectra are not sufficiently well resolved to determine kinetic parameters, we sought to investigate this molecular system by computational means. ## Computational studies Following extensive searching, six conformers of L Zn -[Zn 2 L 2e 3 ] 4+ were located and minimised [Fig. 3]. These fell into two classes: those where the three m-xylenyl groups were oriented away from the central cavity i.e. exo, and those where one such group was oriented endo. No conformers were observed in which two or three m-xylenyl groups were oriented into the cavitythis caused too much torsional and steric strain. Structure endo1 was found to be the lowest in energy, the next lowest being endo2 (ca. +5 kcal) which differs only in the fold of one of the linkers. For these structures the Zn-Zn distances are ca. 11.7 and 11.8 respectively. The structure exo1 (+7 kcal) has a large central cavity but a similar Zn-Zn distance (11.8 ). The structure exo2 (+8 kcal) has a considerably shorter Zn-Zn distance at ca. 9.5 with accompanying concertinaed fold. Furthermore, higher energy conformers exo3 and exo4 differed principally in how the m-xylenyl groups folded towards each metal centre. Both were found to have a short Zn-Zn distance of 9.4 and 9.5 respectively. While prediction of an equilibrium population from the above calculations is complicated by statistical and entropic contributions from the total number of possible structures and the differences in structural flexibility, the detection of two distinct structural classes is clearly consistent with observations in solution. We propose that the species detected by NMR displaying high symmetry (D 3 ) comprises exo conformations while the asymmetric (C 1 ) species is endo. Examination of the structures indicates that the barrier to conversion within the exo or endo manifolds would be low since it would involve relatively simple concertina-type processes, but conversion between exo and endo conformations requires the rotation of the m-xylenyl linker through a strained, high energy transition state. ## Synthesis of water soluble compounds Pairs of water-soluble Fe II flexicate enantiomers [Fe 2 L n 3 ]Cl 4 (n ¼ 2b-2e) were synthesised in high yield by heating the appropriate dialdehyde linker with either (R)-or (S)-1-phenylethan-1-amine and FeCl 2 in methanol. 1 H-NMR spectra were similar though slightly broader than the analogous Zn II perchlorate complexes and are consistent, along with 13 3 ]Cl 4 displayed poor solubility in water and methanol and could not be fully characterised. ## Stability in aqueous media Absorbance spectra indicated that little decomposition of the flexicates occurred in water at pH 7 over weeks, but half-lives for decomposition could readily be recorded in hydrochloric acid (0.2 M) via the 540 nm MLCT absorbance band of the complex. Even under such conditions, frst order kinetic plots gave t 1/2 values in the region 10-20 h. This very favourable aqueous stability of flexicates probably arises from the presence of extensive (hydrophobic) p-stacking. 23 ## Biological activity & selectivity Cytotoxicity. The activities of the new compounds and cisplatin were investigated in human tumour cell lines: (a) MDA-MB-468 (human epithelial breast adenocarcinoma); 24 (b) HCT116 p53 +/+ and (c) HCT116 p53 / . 25 The HCT116 p53 +/+ and HCT116 p53 / cancer cells are human colorectal cancer cell lines that are genetically identical (isogenic) except for the presence or absence of functional p53. 25 These were chosen to enable screening of the effects of p53 status as the loss of p53 function is common genetic event in patient tumours and is strongly associated with increased resistance to many conventional chemotherapeutic agents. 25,26 In the cisplatin-sensitive (2.5 AE 0.5 mM) MDA-MB-468 cells, the new flexicates showed a range of activities [Fig. 5 Toxicity in healthy human cells. The most active flexicates in HCT116 p53 / cancer cells (2a and 2c), along with cisplatin, were investigated in human non-cancer retinal pigment epithelial cells (ARPE19) 27 and normal lung fbroblasts (WI38) [Fig. 5(d)-(e)]. These are healthy human cells with a stable diploid karyotype which senesce after multiple passaging as is characteristic of non-cancer cells. 27 In Fig. 5(f) we depict an in vitro selectivity index (SI) which compares the activity of these compounds in ARPE19 and HCT116 p53 / cells. While for cisplatin SI was found to be signifcantly less than 1, meaning that it is actually more toxic to these healthy cells than it is to the cancer cells, the flexicates tested gave SI substantially higher, and for D Fe -[Fe 2 L 2c 3 ]Cl 4 SI ¼ 836 AE 280. This excellent selectivity prompted us to investigate the toxicity of the compounds against a number of organisms. Toxicity to microbes. The compounds were screened against cultures of the gram-positive bacterium methicillin-resistant Staphylococcus aureus, USA300 JE2 (ref. 28 and 29) (MRSA) and the gram-negative Escherichia coli, TOP10 (E. coli). 30 Kanamycin 30 was used as a positive control. The new flexicates had very modest antimicrobial activity (Table 1) or did not signifcantly inhibit microbial growth at concentrations well over 3 orders of magnitude higher than the IC 50 values observed in cancer cells. Toxicity in amoebae and M. sexta larvae. We further tested the potential toxicity of these compounds using a single cell protist organism, the well-established amoeba model 3 ]Cl 4 showed comparable weight gain to controls suggesting no oral toxicity and no adverse effect on feeding behaviour (Fig. 6). Interestingly the larvae exhibited an increased mean weight gain of approximately 30% (P < 1). Also we noted that larvae that ingested the flexicate solutions turned a bright purple colour over the course of the assay, suggesting that these compounds were persisting in the insect and not being rapidly metabolized or excreted. Systemic toxicity was further tested by injection of 50 mg (0.25 mM) of the compounds directly into the hemocoel of cohorts of 5 th instar larvae (n ¼ 3). The cohorts were then allowed to continue feeding. Despite becoming purple, all larvae proceeded to develop into the pupal diapause stage as per the buffer control injections. ## Mode of action The mode or modes of action of such a new and different system will require intensive investigation and is likely to involve multiple targets and pathways. Here, we describe two preliminary studies towards this end. Denaturation of ct-DNA. We have previously concluded that the induction of DNA damage is not involved in the mode of action of earlier flexicates, despite particular examples binding in a cell free environment. 20,32,33 We investigated the effect that the new flexicates had on the denaturation temperature (T m ) of ct-DNA to screen for any indications of DNA binding. Isolated ct-DNA (0.5 mg mL 1 ) was mixed with each flexicate (7.5 mM) in buffered conditions (10 mM tris, 1 mM EDTA at pH 7.0), to give 10 bases: 1 flexicate complex, and the absorbance at 260 nm between 25 C and 90 C was recorded (0.4 C min 1 ). T m for each experiment was calculated from the frst derivative of a Boltzmann sigmoidal ft of the plot of absorbance versus temperature. T m of untreated ct-DNA (0.25 mg mL 1 in 10 mM tris, 1 mM EDTA at pH 7.0) was measured to be 68.3 AE 0.5 C. Most of the new flexicates had no signifcant effect on the denaturation of ct-DNA (Fig. 7); the small (DT ca. 1 C) reduction for L 2e enantiomers can be ascribed to an electrostatic effect. 34 We are therefore satisfed that DNA is unlikely to be the target of this panel of compounds. Induction of cell death by apoptosis. The chemosensitivity observed could be due to cytostatic or cytotoxic effects, and cell death can occur by several different mechanisms. These include programmed cell death by apoptosis, inflammatory necrosis, autophagy or 'self-eating', necroptosis and pyroptosis. 35 One of the hallmarks of cancers is the evasion of apoptosis, thus enabling the long-term survival and proliferation of cancer cells. 36 We thus investigated whether the most active flexicates are stimulating apoptotic death in cancer cells as part their mode of action. HCT116 p53 +/+ cancer cells (24 h post-seeding) were incubated in fresh media containing flexicate or no flexicate (control) and were then analysed after 48 h for levels of apoptosis and necrosis. As cells start to undergo apoptosis, one of the frst cellular changes is the externalisation of the membrane protein phosphatidylserine (PS). This can be detected and quantifed by fluorescently labelled annexin V 37,38 which can selectively bind externally exposed PS but is membraneimpermeable. This enables cells in the early stages of apoptosis to be distinguished from necrotic cells and cells in the late stages of apoptosis both of which have lost membrane integrity and will therefore also stain with the membrane-impermeable DNA stain propidium idodide. ## View Article Online Flexicates L Fe -[Fe 2 L 2a 3 ]Cl 4 and D Fe -[Fe 2 L 2a 3 ]Cl 4 were tested and both induced signifcant levels of apoptosis that were $2.6 fold (L Fe -[Fe 2 L 2a 3 ]Cl 4 ) and $4.4 fold (D Fe -[Fe 2 L 2a 3 ]Cl 4 ) above background control levels in the HCT116 cancer cells at 48h (Fig. 8). A signifcant proportion of late apoptotic/necrotic cells were also detectable by 48 h, with levels $2.3-2.5 fold above background control levels (Fig. 8). These preliminary investigations indicate induction of apoptosis by these new flexicates as part of their mode of action. ## Experimental Synthesis (E)-5,5 0 -(But-2-ene-1,4-diylbis(oxy))dipicolinaldehyde (0.13 g, 0.44 mmol) and (R)-1-phenylethan-1-amine (0.11 g, 0.88 mmol) were dissolved in acetonitrile (10 mL) with Zn II perchlorate hexahydrate (0.11 g, 0.29 mmol) and the solution was stirred at ambient temperature for 20 h. Ethyl acetate was added dropwise to cause precipitation of a white crystalline solid, L Zn -[Zn 2 L 2c 3 ][ClO 4 ] 4 $10H 2 O. Yield 0.214 g, 57%. 1 H NMR (400 MHz, 298 K, CD 3 CN) d H 8.06 (6H, s, CHN), 7.49 (6H, dd, 3 J HH ¼ 8.5 Hz, 4 J HH ¼ 3.5 Hz), 7.36 (6H, d, 3 J HH ¼ 8.5 Hz), 7.14 (6H, d, 3 J HH ¼ 3.5 Hz), 7.09 (6H, t, 3 J HH ¼ 8.0 Hz), 6.95 (12H, t, 3 J HH ¼ 7.5 Hz), 6.64 (12H, d, 3 J HH ¼ 7.0 Hz, Ar), 6.12 (6H, m), 5.38 (6H, q, 3 J HH ¼ 6.5 Hz, CH), 4.64 (12H, s, CH 2 ), 1.61 (18H, d, 3 J HH ¼ 6.5 Hz). 13 (E)-5,5 0 -(But-2-ene-1,4-diylbis(oxy))dipicolinaldehyde (0.1 g, 0.32 mmol) and (R)-1-phenylethan-1-amine (0.08 g, 0.63 mmol) were dissolved in methanol with Fe II chloride (0.03 g, 0.21 mmol). The solution was stirred at reflux (75 C) for 48 h and all volatiles were removed under reduced pressure to yield a dark purple solid, L Fe -[Fe 2 L 2c 3 ]Cl 4 $9H 2 O. Yield 0.388 g, 97%. 1 H NMR (400 MHz, 298 K, MeOD) d H 8.80 (6H, s, CHN), 7.47 (6H, s), 7.13 (6H, t, 3 J HH ¼ 7.0 Hz), 7.04 (12H, t, 3 J HH ¼ 7.0 Hz), 6.64 (12H, d, 3 J HH ¼ 7.0 Hz), 6.44 (6H, s), 6.05 (6H, s, Ar), 5.26 (6H, q, 3 J HH ¼ 6.0 Hz, CH), 4.65 (12H, s, CH 2 ), 4.59 (6H, br s, CH), 1.99 (18H, d, 3 J HH ¼ 6.0 Hz, CH 3 ). 13 ## Molecular modelling Models of a number of possible conformers of L Zn -[Zn 2 L 2e 3 ] [ClO 4 ] 4 were constructed and optimised. Starting points for geometry optimisations were taken from crystallographic data. Monometallic structures were frst optimised using the B3LYP-D3(BJ) 40 functional and the 6-31g* basis set, with convergence criteria of 0.0001 a.u. as implemented in the Firefly quantum chemistry package, 41 which is partially based on the GAME-SS(US) source code. 42 Bimetallic systems were optimised using ligand feld molecular mechanics (LFMM) 43 as implemented in the DommiMOE program, 44 before being annealed at 500 K for 1 ns prior to re-optimisation. Single point energy calculations of all structures were performed using the B3LYP-D3(BJ) 40 functional and the deff2-TZVP basis set with energy convergence criteria of 0.0001 a.u. as implemented in the Firefly quantum chemistry package. 41 The calculations were conducted by employing the RIJCOSX approximation with SCF convergence criteria set to 'tight', both of which are defned internally as part of the ORCA DFT quantum chemistry package. 45 Where relevant, acetonitrile solvate correction was performed using the conductor-like screening model (COSMO) 46 as implemented in ORCA. 45 Biological activity MIC values were established using a macro broth dilution method in cation-adjusted Müller-Hinton (MH) broth. 96-well plates (200 mL of 128 mg mL 1 , 64 mM) complex in MH broth, diluted 2 n mg mL 1 , inoculated with each bacterial strain (10 3 cfu mL 1 ) were sealed and growth was monitored over 20 h at 37 C with an iEMS 96-well plate reader (see ESI †). IC 50 values were determined by incubating cells in 96-well plates (2.0 10 3 cells per well) for 24 h at 37 C, 5% CO 2 prior to drug exposure. Compounds were added (100 mM to 5 nM in cell medium) for a further 96 h. 3-(4,5-Dimethylthiazol-1-yl)-2,5diphenyl tetrazolium bromide solution (0.5 mg mL 1 , 20 mL per well) was added for a fnal 4 h. Upon completion all solutions were aspirated, dimethyl sulfoxide (150 ml) was added and absorbance (540 nm) was recorded with a Thermo Scientifc Multiskan EX microplate photometer. Oral toxicity was established by feeding cohorts of Manduca sexta 31 one-day-old neonate larvae with each flexicate (25 mg mL 1 in artifcial wheat germ diet) for 7 d at 28 C and weighing to assess growth rate. Systemic toxicity assays 47 were conducted by injecting an ethanol (70% v/v) swabbed region of frst day ffth instar M. sexta larvae with each flexicate (0.5 mg mL 1 [0.25 mM] in PBS), before allowing them to continue feeding for 7 d at 28 C, using physical stimulus to assess their status. ## Mode of action Denaturation of ct-DNA was measured by mixing ct-DNA (0.5 mg mL 1 , 7.5 10 5 per base, as determined by absorbance at 200 nm) with each complex (7.5 mM) in buffered conditions (10 mM tris, 1 mM EDTA at pH 7.0) to give 10 base: 1 complex. The absorbance at 260 nm as a function of temperature (every 1 C, 25-90 C) was measured in a 1 cm masked quartz cuvette at a rate of 0.4 C min 1 and run in triplicate. T m was calculated from the frst derivative of a Boltzmann sigmoidal ft of the plot of absorbance at 260 nm against temperature for each complex. Induction of apoptosis was determined by incubating HCT116 p53 +/+ cells (5 10 5 cells/flask, 10 mL complete RPMI-1640 medium) for 24 h at 37 C in 5% CO 2 , before treating with each flexicate (20 mM in fresh complete media for 48 h) or fresh media containing no drug (control). The supernatant containing any non-adhered, floating cells was then collected and pooled with cells harvested by trypsinisation. This pooled single cell suspension was washed twice with PBS and incubated with propidium iodide and Annexin-V-FLUOS (Roche) to stain apoptotic cells in accordance with the manufacturer's instructions. The proportion of early stage apoptotic cells and late stage apoptotic/necrotic cells were then quantifed by flow cytometry as previously described. 37,38 ## Conclusions Our approach to metallohelix assembly has allowed us to generate a panel of biologically-compatible enantiomers incorporating various bridging groups. This was possible because in this so-called flexicate platform the stereochemistry of the metal complex units is predetermined very efficiently and largely independently of the bridges, and by a mechanism that also provides water-compatibility. 23,48 In contrast, in a conventional "helication" approach the bridging units are structure-determining, so a mechanism of stereoselection would need to be designed for each example. A further advantage of the flexicate platform is beginning to emerge in that we may be able to develop asymmetric molecules from symmetric ligands via the kinds of conformational abnormalities caused by bridges that partially oppose the predetermined stereochemistry e.g. L 2e . We have already shown that asymmetric (as opposed to merely chiral) optically pure assemblies are available using directional ligands. 49 Further, this modular self-assembling system will allow us to probe the effects of peripheral functionality and lipophilicity. The activity of these new assemblies against cancer cells is strongly dependent on structure, with a range of potencies from 30 mM to as low as 40 nM. The most active compound D Fe -[Fe 2 L 2c 3 ]Cl 4 shows a selectivity index (versus healthy cell lines) approaching 10 3 , demonstrating superiority over the clinically used anticancer drug cisplatin in vitro (SI < 1). This selectivity is substantiated in tests with various models; bacteria and amoeba exposed to high concentrations were essentially unaffected, and in Manduca sexta larvae, where the systemic stability of the drug is evidenced, there is arguably a pro-biotic effect i.e. the insects appear to thrive. In respect of mechanism or mode of action, the lack of binding to DNA indicates that this is unlikely to be the general target in this panel. In fact only one early flexicate 19 ([Fe 2 L 1 3 ] 4+ , Fig. 1) in our growing library shows signifcant interactions with nucleic acids, and while there are fascinating selectivities with various motifs 19,32,33 there is no DNA damage akin to that induced by e.g. platinum drugs and alkylators. 34,49 Instead, relevant examples of protein interaction and enzyme inhibition have been characterised. 21,32 To achieve drug safety and cancer selectivity, mechanistic classes which do not involve induction of DNA damage are attractive, and this may well be the source of the excellent selectivities we describe in this manuscript. Mode of action studies indicate that these compounds can induce substantial cell death by apoptosis independent of any DNA damage. Extensive studies are now required to understand how this complex process, normally subverted in cancers, is induced by these compounds. The above observations of remarkable selectivity alongside very high potency and large enantiomeric differences are however all consistent with a subtle mechanism involving the targeting of oncogenic drivers.

chemsum

{"title": "Anticancer metallohelices: nanomolar potency and high selectivity", "journal": "Royal Society of Chemistry (RSC)"}

directed_evolution_of_flavin-dependent_halogenases_for_atroposelective_halogenation_of_3-aryl-4(3h)-

## Abstract: In this study, we engineer a variant of the flavin-dependent halogenase RebH that catalyzes siteand atroposelective halogenation of 3-aryl-4(3H)-quinazolinones via kinetic or dynamic kinetic resolution. The required directed evolution uses a combination of random and site-saturation mutagenesis, substrate walking using two probe substrates, and a two-tiered screening approach involving analysis of variant conversion and then enantioselectivity of improved variants. The resulting variant, 3-T, provides >99:1 e.r. for the (M)-atropisomer of the major brominated product, 25-fold improved conversion, and 91-fold improved site-selectivity relative to the parent enzyme on the probe substrate used in the final rounds of evolution. This high activity and selectivity translates well to several additional substrates with varied steric and electronic properties. Computational modeling and docking simulations are used to rationalize the effects of key mutations on substrate scope and site-and atroposelectivity. Given the range of substrates that have been used for atroposelective synthesis via electrophilic halogenation, these results suggest that FDHs could find many additional applications for atroposelective catalysis. More broadly, this study highlights how RebH can be engineered to accept structurally diverse substrates that enable its use for enantioselective catalysis. ## Introduction Site-selective catalysis allows for precise manipulation of molecular structure by controlling which of two or more identical functional groups (e.g. ketones, C-H bonds, etc.) in nonequivalent positions undergoes a reaction of interest. 1 In addition to introducing new functionality, this capability can provide access to otherwise unfavorable scaffold conformations or lock a scaffold in a particular conformation. 2 In the latter case, site-and stereoselective modifications can also control the absolute stereochemistry of the scaffold. For example, a single atropisomer can be generated upon installation of a functional group that hinders the rotation of two fragments joined by a sigma bond. 3 When the barrier to rotation is high enough, conformationally-locked atropisomers maintain unique three-dimensional structures that can interact with other molecules in a stereoselective manner. This phenomenon is found in natural products 4 and is increasingly exploited in the pharmaceutical industry 5 since a single atropisomer of a compound often exhibits improved properties relative to mixtures of atropisomers or to derivatives containing fragments that can rotate freely about an axis of chirality . In response to the growing appreciation of atropisomerism and its effects on molecular function, chemists have developed a variety of methods to access single atropisomers of diverse compounds, particularly those containing an axially chiral biaryl motif. 3 Among these methods, selective functionalization of biaryl compounds has been used to resolve atropisomers via kinetic resolution, or, if the starting atropisomers can readily interconvert, via dynamic kinetic resolution. In several notable examples of the latter approach, 9 the Miller group has established that suitably designed peptides catalyze enantioselective synthesis of brominated atropisomers of benzamides, 10 3-aryl-4(3H)-quinazolines, 11 and other biaryl compounds 12,13 . These reactions are believed to proceed via dynamic kinetic resolutions in which a peptide selectively binds one atropisomer of a substrate and facilitates monobromination of the substrate by an electrophilic brominating reagent to set the axis of chirality (Figure 1A). 9 In all cases reported to date, however, subsequent electrophilic bromination also occurs, so additional synthetic steps are required to remove or functionalize these groups if they are not desired in the final product. 11,12,14 Site-selective catalysts for atroposelective electrophilic bromination could therefore improve the efficiency of these processes by installing only the bromine substituent required to set the stereochemical axis and enable subsequent transformations. Our group and others have established that flavin-dependent halogenases (FDHs) can be engineered to brominate a diverse range of substrates with high site selectivity. We have also developed FDH variants that catalyze enantioselective chlorination of methylenedianilines 21 and halolactonization of olefins 22 . FDHs are believed to achieve this selectively by binding substrates such that only a single site is presented to HOX (X = Br, Cl) that is selectively bound and activated by a lysine residue within their active site. Given this capability, we envisioned that a suitably engineered FDH could enable site-and atroposelective halogenation of biologically relevant compounds. Biocatalytic methods for atroposelective synthesis are dominated by hydrolysis or oxidation/reduction of functional groups on atropisomeric compounds 26 . Atroposelective lipasecatalyzed macrolactonization 27 and P450-catalyzed oxidative biaryl coupling have been developed, 28,29 but modest yields and/or selectivities have typically been observed for non-native substrates in the latter case to date 30 . Suitably engineered FDHs could therefore both improve the efficiency of electrophilic halogenation for atroposelective synthesis and expand the range of enzymes that catalyze atroposelective transformations via substrate elaboration. Herein, we establish that previously engineered FDHs catalyze atroposelective halogenation of 3-(3'aminophenyl)-4(3H)-quinazolinones and that directed evolution can be used to improve the site selectivity and activity for these reactions to enable efficient access to mono-brominated atropisomers of these compounds (Figure 1B). We also demonstrate that these products can be transformed via cross-coupling involving the installed bromide or heterocycle formation involving the aniline amine substituent with no erosion of enantiomeric purity. These results further establish the utility of FDHs for enantioselective catalysis and chemoenzymatic synthesis. ## Results Inspired by Miller's studies on atroposelective tribromination of 3-(3-hydroxyphenyl)-4(3H)quinazolinones (Figure 1A) 11 and building on our previous study of enantioselective chlorination of methylenedianilines, 21 we evaluated the brominase activity of 46 wild type FDHs and 95 engineered variants on 3-(3-amino-5-methylphenyl)-2-methyl-4(3H)-quinazolinone, 1 (see supporting data_file). The assay yields of 1a and its isomer 1b were evaluated for these enzymes in cell lysate using UPLC. The highest yielding variants were then purified, and their atroposelectivity was evaluated by chiral SFC (see supporting data_file). This analysis revealed that RebH variant 6-TL F465P, originally reported in a study aimed at engineering RebH variants with altered site selectivity on tryptamine, 31 provided roughly ~5% assay yield of both 1a and 1b (Figure 2). More importantly, 1a was formed with >99% enantioselectivity (Figure S3). Given this finding, we sought to improve the activity of 6-TL F465P using directed evolution. A library of 800 variants of 6-TL F465P containing an average of 1-2 base pair mutations/gene was generated using error prone PCR. The two-tiered screening protocol outlined above was used to evaluate the yield and atroposelectivity of 1a provided by each variant. In general, all variants with improved activity and site selectivity in the lysate screen maintained near-perfect atroposelectivity. One variant from this screen, 1-L (6-TL F465P F402L), provided 20-fold higher selectivity for 1a over 1b and 6-fold better conversion (Figure 2). On the other hand, preliminary analysis of 1-L activity on other 3-aryl-4(3H)-quinazolines revealed that substitution on the quinazoline benzene ring lowered both yield and atroposelectivity substantially (Figure S6), indicating that screening on substrate 1 alone was insufficient to identify enzymes with desirable substrate scope. To identify variants with improved activity and substrate scope, a second library of 1000 1-L variants, again generated by error-prone PCR, was evaluated on methyl-substituted substrate 2. Analysis of this library revealed three variants, 2R (1-L W455R), 2S (1-L N467S), and 2G (1-L S469G), that improved conversion of substrate 2 by 1.3-1.7-fold (Figure 2). Combining these mutations to generate double mutants 2-RS, 2-RG, and 2-SG and triple mutant 2-RSG led to further improvements in activity, as 2-RSG, for example, provided 2.2-fold improved conversion for 2a over 1-L. Given the beneficial effects of mutations at residues 455, 467, and 469, individual site saturation libraries at these positions were constructed from 2-RSG using degenerate NNK codons. Two variants from these libraries, 3R (2-RG S467R) and 3T (2-RG S467T), exhibited 1.8-fold improved activity over 2-RSG for conversion of substrate 2 and 1.4-fold improved activity for substrate 1, although 3-R had slightly lower site selectivity for 1a/1b compared to 3-T. In total, 3-T is 14-and 25-fold improved for the conversion of substrates 1 and 2 and gives 35-and 91-fold improvement to site selectivity over 6-TL F465P (Figure 2). The substrate scope of 3-T toward a variety of substituted 3-(3-aminophenyl)-2-alkylquinazolin-4(3H)-ones was next evaluated in preparative-scale halogenation reactions. Isolated yields of the desired products ranged from 17-92%, and high enantioselectivity (≥95:5 e.r.) was observed for all reactions. Single crystal X-ray diffraction was used to establish that the (M) atropisomer of 1a was formed using 2-RSG from the 3-T lineage (Figure S27), and that selectivity was presumed for the remaining products. Given that a methyl substituent at the 6-position of the quinazolinone core of 2a led to lower conversion of this substrate relative to 1a using FDH variants early in the quinazolinone lineage, the activity of 3-T on substrates with differential methyl substitution on both the quinazolinone core and the 3-aminophenyl group was examined. Methyl substitution of the 6-and 7-positions of the quinazolinone was well-tolerated, and only a slight loss in site selectivity was observed. Substitution of the 5-position led to decreased yield, but high site-and enantioselectivity was obtained with slightly increased catalyst loading. Relatively low (5:1) site selectivity was obtained with a methyl group at the 8-position, though enantioselectivity remained high. Methyl substitution of the 4-position of the 3-aminophenyl group in substrate 7 was not tolerated, and analogous substitution of the 6-position of this group leads to stable atropisomers in substrate 6, providing the opportunity to establish whether 3-T could be used to resolve these atropisomers via classical kinetic resolution. Even though the 6-methyl group occupies the position typically halogenated by 3-T, 6a was formed in 15% yield with an enantiomeric ratio of 95:5, constituting an E value of 22. Additional steric bulk of the quinazolinone 2-position, which is critical for restricting rotation of the biaryl axis, was tolerated as evidenced by highly selective halogenation of both ethyl-and benzyl-substituted compounds 8 and 9. The corresponding 2-isopropyl substituted compound 10, however, was not halogenated by any enzyme in the lineage. Varying electronic properties of substituents at the quinazolinone 6-position showed that while electron-donating through modestly electron-withdrawing substituents were well-tolerated (i.e. OCH3, H, and Cl), a strongly electronwithdrawing CF3 group led to low conversions and significantly reduced site selectivity. Variation of substituents on the 3-aminophenyl group revealed a similar trend, as a 5-methyl substituted compound 1 led to the highest conversion, followed by unsubstituted compound 14 and 5-chloro substituted 15. In general, we observed substantial improvements in yield from 1-L to 3-T for the halogenation of all substrates examined excepting 3-amino-4-methylphenyl substituted substrate 7. In addition to the improved yields, 3-T either improved or maintained site-and enantioselectivity when compared to 1-L for each product. While site selectivity for most reactions exceeded 10:1, regioisomeric ratios for substrates 5 and 13 were relatively low (5:1 and 2:1, respectively). Traditional resolution substrate 6 and 2-benzyl substituted 9 were also halogenated with reduced reduced site selectivity (3:1 and 9:2, respectively), perhaps relating to the more complex structures of these substrates. Figure 3. Substrate scope of 3-T. Relevant compound numbering is provided in the inset. Product yields are isolated % yields, enantiomeric ratio (e.r.) was determined by chiral SFC analysis, and regioisomeric ratio (r.r.) was obtained from the ratio of integrals for products 1-15a:b obtained from UPLC analysis (r.r. >25:1 denotes second isomer not detected; N.D. indicates value was not determined). a 1 mol% 3-T b 2.5 mol% 3-T c Starting material exists as stable atropisomers. d 500 µM substrate, 20 equiv. NaBr, 5 mol% 3-T e See product characterization in the SI for details. As previously noted, a key benefit of site-and atroposelective bromination is simplified functionalization of the brominated aromatic ring 9 . Both the added bromo group and the amino group of the 3-aminophenyl fragment represent potential points of diversification for the products generated in this study. Despite the fact that 1a is a relatively hindered aryl bromide, this substrate underwent Suzuki-Miyaura cross-coupling with p-methoxyphenylboronic acid under mild conditions 32 to give tri-aryl product 1c in 75% yield without a detectable change in enantiomeric ## Discussion In this study, we establish that RebH variants can be engineered to catalyze site-and atroposelective halogenation of 3-(3-aminophenyl)-4(3H)-quinazolinones. The required directed evolution proceeded via a two-tiered screening approach involving initial analysis of variant conversion using UPLC followed by analysis of the enantioselectivity of improved variants using chiral SFC. Because our parent enzyme, RebH variant 6TL F465P, was identified from a collection of enzymes that already contains many active site mutations, we turned to random mutagenesis to identify sites required to improve activity on biaryl substrates. Saturation mutagenesis at the sites of beneficial mutations led to variant 3-T. Variant 3-T has 14 mutations relative to RebH (protein sequences are available in the SI). Four of these mutations, F402L, Y455R, N467T (obtained via N467S then S467T), and S469G were introduced in the current study; Y455R and N467T are in the active site, while F402L and S469G are in the second sphere. We previously established that N467T led to improved activity on large aromatic substrates. 15 While we have observed beneficial effects of mutations at Y455, 21,31 the specific mutation identified in the current study is unique, and we had not characterized variants with mutations at the two non-active site residues. Docking simulations were used to help rationalize how these mutations might lead to improved activity and selectivity on 3-(3-aminophenyl)-4(3H)-quinazolinones. Models of 6TL F465P, 1-L, and 3-T were generated using Alphafold, 33 and substrate 1 was docked in the active sites of each model using AutoDock Vina 34 . In general, FDHs bind their substrates such that the halogenation site projects toward an active site lysine residue that is believed to activate HOX for electrophilic attack via H-bonding. 35 For example, in the crystal structure of the RebH, a tryptophan 7halogenase, C7 of tryptophan is 3.9 from K79 (Figure 5A). The RebH active site is relatively small, but 6TL F465P has several active site mutations that increase its volume. F465P in particular is predicted to open the left side of the active site to facilitate binding of 1 (Figure 5B) such that the quinazolinone core is sandwiched between H109 and L111 just as other aromatic substrates bind to RebH variants containing F111 (Figure 5A). Consistent with the low activity of this enzyme, no poses were oriented such that the 2-position of the 3-aminophenyl fragment projects toward K79. On the other hand, F402L is predicted to push W466 toward bound 1, rotating 1 so that its 2-position is only 4.4 from K79 (Figure 5C) and the 3-aminophenyl fragment nearly overlays with the benzene ring of bound tryptophan in the crystal structure shown in Figure 5A. One of the more notable mutations along the lineage is W455R, which significantly changes both the shape and charge distribution of the active site. This mutation is predicted to flip the orientation of 1 in the active site due to formation of a bidentate hydrogen bond between R455 and the carbonyl oxygen of 1 (Figure 5D). The reduced site selectivity observed for substrate 5, which has a methyl group at the quinazolinone 5-position that would disrupt this interaction, provides experimental support for the potential relevance of R455 H-bonding to catalysis. Notably, 1 is also oriented such that the 2-position of the 3-aminophenyl fragment projects toward K79 in a pro-M conformation, consistent with the observed site-and atroposelectivity. It is less clear how S467T and S469G might influence activity. MD simulations suggest that E357, H109, and S470 maintain a persistent H-bonding network that lines the top of the substrate binding pocket. 25 The proximity of T467 and G469 to S470 could modulate this network to better accommodate the quinazolinone substrates. Notably, W466 is predicted to move even closer to bound substrate in the 3-T structure, further suggesting that this network could play a significant role in substrate binding. Relative to parent enzyme 6-TL F465P, 3-T gives 14-and 25-fold improved conversion of substrates 1 and 2 and 35-and 91-fold improved regioselectivity on these substrates, which were used for directed evolution. Steady state kinetic analysis of 3-T activity on 1 revealed that the enzyme has a kcat of 0.03 min -1 and a KM of 2.0 µM, values that are approximately an order of magnitude less than and nearly identical to those for WT RebH acting on its native substrate, 37,38 tryptophan. Despite the unique substrate topology required for atroposelective halogenation of 1, the catalytic efficiency of this system (kcat/KM = 0.012 min -1 µM -1 ) matches our previous best example of an engineered enzyme acting on a non-native substrate, tryptamine (0.012 min -1 µM -1 ), 31 which is structurally similar to tryptophan. Importantly, this high activity and selectivity translated well to several additional substrates with varied steric and electronic properties. This finding again highlights how RebH can be engineered to accept structurally diverse substrates that enable its use for enantioselective catalysis. ## Conclusion In summary, we engineered a variant of RebH that catalyzes site-and atroposelective halogenation of 3-(3-aminophenyl)-4(3H)-quinazolinones via kinetic or dynamic kinetic resolution. While enantioselective catalysis using FDHs remains rare, the high selectivity obtained for halogenation of a range of structurally diverse quinazolinone substrates illustrates the utility of FDHs toward this end. This result, combined with the fact that engineered variants of RebH also catalyze enantioselective desymmetrization of meso-methylenedianilines 21 and halolactonization of olefins bearing pendant carboxylate nucleophiles 22 , highlights the diversity of substrate topologies and mechanisms for stereochemical induction that can be accommodated by these enzymes. Access to isolable quantities of single atropisomers of halogenated 3-(3-aminophenyl)-4(3H)quinazolinones is particularly useful given the biological activity of related structures 39 and the ease with which the bromide and amine substituents can be leveraged for subsequent synthetic elaboration. Since electrophilic halogenation has been used for atroposelective functionalization of several different substrate classes, 9,40 these results suggest that atroposelective catalysis via FDH-catalyzed, site-selective halogenation could find many additional applications.

chemsum

{"title": "Directed Evolution of Flavin-Dependent Halogenases for Atroposelective Halogenation of 3-Aryl-4(3H)-quinazolinones via Kinetic or Dynamic Kinetic Resolution", "journal": "ChemRxiv"}

high_chemoselectivity_in_the_phenol_synthesis

## Abstract: Efforts to trap early intermediates of the gold-catalyzed phenol synthesis failed. Neither inter-nor intramolecularly offered vinyl groups, ketones or alcohols were able to intercept the gold carbenoid species. This indicates that the competing steps of the goldcatalyzed phenol synthesis are much faster than the steps of the interception reaction. In the latter the barrier of activation is higher. At the same time this explains the high tolerance of this very efficient and general reaction towards functional groups. ## Introduction As documented in numerous reviews , over the last eleven years homogeneous gold catalysis has emerged from early examples which documented its potential for organic synthesis of even complex molecules to an established tool in preparative organic chemistry . One of these early examples is the gold-catalyzed phenol synthesis in which the furan-ynes 1 used as substrates represent the first ene-ynetype compounds ever used in gold catalysis. While many investigations in the field focused on methodology, mechanistic research was much less widespread . The goldcatalyzed ene-yne cycloisomerization reactions are, mechanistically, very complex reactions , and the furan-yne cycloisomerization is no exception. For the latter reaction arene oxides D and oxepines C could be detected as intermediates, and these could even be trapped by Diels-Alder reactions. In addition, labelling studies were carried out and the electronic influence of substituents was investigated . Computational studies as well as side-products produced in the reaction pointed towards intermediates A and B (Scheme 1) . Moreover, interesting new pathways were opened when ynamides and alkynyl ether substrates were employed: Here A is also a possible intermediate along these pathways . Since direct experimental evidence existed only for C and D, we intended to intercept the postulated carbenoid intermediates A or B. Apart from intermolecular trapping , intramolecular trapping of such carbenoids has also been reported . One option would be to offer a competing carbonyl group, to produce a carbonyl ylide, which could then undergo a 1,3dipolar cycloaddition . The second option would be a classical cyclopropanation of an olefin. A third option would be trapping of intermediate A with an intramolecular hydroxy nucleophile . Here we report our observations when trying to apply these principles to intermediates of type A or B. ## Intermolecular olefinic trapping reagents We started with the simplest experiments, namely the intermolecular trapping of the gold carbenoid intermediates. When 3 was reacted in the presence of an activated olefin, such as norbornene or styrene, phenol 4 was formed exclusively in essentially quantitative yield, no other products could be detected (Scheme 2). Experiments with a competing carbonyl group (competing with the carbonyl group in intermediate B) were also unsuccessful. Ketone 5 , prepared by the addition of methyllithium to commercially available hex-5-enoic acid, was used as an external carbonyl group. Reaction with both tosylamide 3 and ether 6 always delivered the phenolic products 4 or 7, respectively (Scheme 3). The same result was obtained when PtCl 2 was used as the catalyst for the conversion of 3. ## Intramolecular olefinic trapping reagents The next step was to offer the styrene unit in an intramolecular manner. Substrate 8 could potentially undergo three different modes of reaction (Scheme 4). After the initial step, the intermediate E would be produced (analogous to A). Cyclopropanation of the styrene subunit by the cyclopropyl carbenoid would deliver 9. If E rearranged to the vinylcarbenoid F, the two competing reactions would be the formation of the phenol 10 and cyclopropanation to form 11. The synthesis of 8 was possible by a short route (Scheme 5). Starting from the commercially available 2-bromostyrene (12), a halogen-metal exchange and subsequent formylation according to a procedure of Fukumoto et al. gave 13. Addition of ethynylmagnesium bromide to 13 led to 14, which reacted with furan 15 under Mitsunobu conditions to afford 8. While the yields were good for the first two steps of the reaction sequence, the yield of the last step was only 32%. With AuCl 3 the phenol 10 was formed exclusively (Scheme 6). The structure was unambiguously confirmed by X-ray crystal structure analysis (Figure 1). It shows an interesting hydrogen bond-like interaction of the phenolic hydroxy group and the alkene unit. After changing the solvent from acetonitrile to CDCl 3 , and the gold(I) catalyst to [Mes 3 PAu]NTf 2 , only 10 was again observed. Thus, neither of the two oxidation states of the gold catalyst gave any product derived from the intercepted intermediate (the solvent was changed to CDCl 3 since the activity of gold(I) is significantly reduced by MeCN). ate H, which could then either afford product 17 via intramolecular 1,3-dipolar cycloaddition with the olefin, or could form the diene 19 by proton migration. The synthesis of 16 was only possible by a 9-step sequence (Scheme 8). The starting point was a Claisen condensation of ester 20 and tert-butyl acetate (21) in the presence of lithium hexamethyldisilazide as the base. Ketoester 22 was obtained in 56% yield, however, the two-fold addition of 21 could not be suppressed completely and 14% of the corresponding tertiary alcohol 30 was also obtained. Reduction of the ketone 22 with sodium borohydride and protection of the alcohol 23 with tertbutyldimethylsilylchloride delivered 24 in excellent yield. Reduction of the ester group with diisobutylaluminiumhydride gave aldehyde 25. The addition of lithiated trimethylsilylacetylene provided the propargylic alcohol 26 and reaction with 15 under Mitsunobu conditions yielded 27. Deprotection of the alkyne 27 and the silyl ether 28, followed by the oxidation of the resulting alcohol 29 finally led to 16. It was not possible to remove both silyl groups simultaneously with TBAF, longer reaction times which would be necessary for the deprotection of the hydroxy group led to decomposition of the substrate. At 0 °C and with a very short reaction time, the alkyne was deprotected selectively. Selective deprotection of the alcohol was then possible with a mixture of acetic acid/water/THF. Another route, in which the alcohol function was deprotected first, then oxidized, followed by removal of the trimethylsilyl group from the alkyne also failed. Thus treatment of 27 with acetic acid in aqueous THF gave the desired alcohol 31in quantitative yield. However, whilst Ley oxidation on the small-scale delivered ketone 32 in yields of up to 80%, on a larger scale the yield of 32 dropped dramatically to 28% and was accompanied by two side-products, 33 and 5. The latter are formed by an elimination reaction of the amide in 32. Furthermore, it was not possible to deprotect ketone 32 due to rapid decomposition. One of the diastereoisomers of 28 was identified as the anti-product 28a by an X-ray crystal structure analysis (Figure 2). The conversion of 16 with 5 mol % AuCl 3 proceeded fast and gave exclusively phenol 18. No other products could be detected (Scheme 9). The two gold(III) complexes 34 and 35 as well as the dinuclear gold(I) complex 36 gave the same result (Figure 3). When the catalyst was changed to platinum(II) chloride in acetone, a complex mixture of inseparable products was obtained. Since the two diastereoisomers 28a and 28b with the propargylic stereocenters were separable, we investigated the goldcatalyzed conversion of the pure isomers. From the NMR spectra taken during the conversion (Figure 4), it could be clearly seen that no epimerization of the propargylic position occurred. In addition to the selective transformation to the phenols 37a and 37b as the main reaction products, partial removal of the TBS group was observed (38, Figure 5). ## Intramolecular alcohol as potential trapping reagent For interception of intermediate A we also considered the option of an intramolecular hydroxy nucleophile, compound 39 (Scheme 10) would represent this type of substrate. The intermediate I would be an analogue of A. Instead of the phenol syn- The conversion of 39, catalyzed by AuCl 3 in CDCl 3 , again only produced the expected phenol 40 (Scheme 12). Not unexpectedly, the PMB-protected alcohol 45 was similarly converted to 46. PtCl 2 did not lead to a change in selectivity. ## Conclusion The complete failure of both the inter-and the intramolecular trapping experiments shows that the gold-catalyzed phenol synthesis follows a reaction pathway low in energy. These observations also nicely explain the high functional group tolerance, for example, towards olefins and alcohols.

chemsum

{"title": "High chemoselectivity in the phenol synthesis", "journal": "Beilstein"}

programmable_mismatch-fueled_high-efficiency_dna_signal_converter

## Abstract: Herein, by directly introducing mismatched reactant DNA, high-reactivity and high-threshold enzyme-free target recycling amplification (EFTRA) is explored. The developed high-efficiency EFTRA (HEEFTRA) was applied as a programmable DNA signal converter, possessing higher conversion efficiency than the traditional one with perfect complement owing to the more negative reaction standard free energy (DG).Once traces of input target miRNA interact with the mismatched reactant DNA, amounts of ferrocene (Fc)-labeled output DNA could be converted via the EFTRA. Impressively, the Fc-labeled output DNA could be easily captured by the DNA tetrahedron nanoprobes (DTNPs) on the electrode surface to form triplex-forming oligonucleotide (TFO) at pH ¼ 7.0 for sensitive electrochemical signal generation and the DTNPs could be regenerated at pH ¼ 10.0, from which the conversion efficiency (N) will be accurately obtained, benefiting the selection of suitable mismatched bases to obtain high-efficiency EFTRA (HEEFTRA). As a proof of concept, the HEEFTRA as an evolved DNA signal converter is successfully applied for the ultrasensitive detection of miRNA-21, which gives impetus to the design of other signal converters with excellent efficiency for ultimate applications in sensing analysis, clinical diagnosis, and other areas. ## Introduction Enzyme-free target recycling amplifcation (EFTRA) based on the toehold strand displacement reaction (TSDR) could convert traces of input targets into amounts of output products to construct autocatalytic circuits for exponential signal amplifcation as a signal converter, which has the advantages of high specifcity, low cost, and simple operation with less environmental interference. Nevertheless, with the lack of an accurate and sensitive measurement method, exactly measuring the conversion efficiency (N) of EFTRA remains a serious challenge, which further limits the exploitation of its inherent properties and expanded applications. Thus it is of signifcant importance to carve out an effective method for accurately monitoring the conversion efficiency of EFTRA. Herein, we utilize a tetrahedral DNA nanostructure with mechanical rigidity, chemical and structural stability, and highly ordered upright orientation 6,7 to design a simple DNA tetrahedral nanoprobe (DTNP), which was immobilized on the electrode surface to effectively capture the Fc-labeled output DNA, the product converted by the quantifed input target via EFTRA, to generate a sensitively monitorable electrochemical signal at pH ¼ 7.0. When this signal value is the same as that generated by the independent and quantifed Fc-labeled DNA (identical to the Fc-labeled output DNA), the experimental conversion efficiency of EFTRA (N) can be accurately obtained through the ratio of the concentration of Fclabeled output DNA from the input target miRNA via EFTRA and the concentration of the input target miRNA. Signifcantly, the DTNP could be regenerated at pH ¼ 10.0, achieving the continuous long-term usage of this sensing platform to accurately monitor the conversion efficiency of EFTRA and providing a new insight for DNA nanoprobe regeneration. Promoting the conversion efficiency of EFTRA to develop a high-efficiency DNA signal converter is another signifcant goal for advancing the superiority and applicability of nucleic acid amplifcation in diagnostic applications, biological research, nanobiotechnology, and bioengineering. Since the rate-limiting step of TSDR is the branch migration in the displacement domain, 14,15 the conversion efficiency of EFTRA could be affected by the reaction equilibrium of the branch migration process. Thus, suitable mismatched bases in the displacement domain of the resultant DNA can make the EFTRA reaction standard free energy (DG) more negative when compared with the completely matched reactant DNA in traditional EFTRA, which could enhance the driving force of EFTRA with increased reactivity and threshold for improving the conversion efficiency. Herein, we adopt the theoretical conversion efficiency of EFTRA (N 0 ) obtained by the well-studied thermodynamic parameters of the nucleic acid hybridization reaction from NUPACK 28 to provide some referential mismatched bases in the displacement domain of the reactant DNA, and then contrast the accurate experimental N to screen out the suitable mismatched bases for high-efficiency EFTRA (HEEFTRA). As a result, we carve out the HEEFTRA and apply it as a programmable DNA signal converter for biomarker assay, which gives impetus to exploit a new generation of nucleic acid amplifcation techniques for biosensing analysis and early disease diagnosis. ## Results and discussion The reaction mechanism and conversion efficiency monitoring of the HEEFTRA are shown in Scheme 1. The target miRNA could trigger the frst TSDR to hybridize with the toehold of the duplex (double helix) ABC (FS-AP-LS), then AP (A) will be displaced and released away, from which the next toehold in the middle of FS (C) will be exposed and then initiate the next TSDR in the presence of TS (D), accompanied by the simultaneous release of Fc-labeled output DNA LS (B) and target miRNA (T). Next, the released target miRNA can be reused to release more Fc-labeled DNA LS (B). As a result, one input of target miRNA can induce multiple outputs of Fc-labeled DNA LS (B) via the HEEFTRA (Part A). Then, the Fc-labeled output DNA LS (B) could be captured by the DTNP on the electrode surface to form triplex-forming oligonucleotide (TFO) at pH ¼ 7.0 for generating the sensitive electrochemical signal response and the TFO could be dissociated in pH ¼ 10.0 realizing the regeneration of DTNP, from which the experimental conversion efficiency of the EFTRA (N) can be accurately obtained (Part B). Ultimately, with the help of the theoretical N 0 and experimental N, the suitable mismatched bases in the displacement domain of duplex ABC (two mismatched bases as shown in Part A) are selected for achieving HEEFTRA. Signifcantly, the evolved HEEFTRA is used to construct an ultrasensitive electrochemical biosensor for the detection of miRNA-21 with a detection limit about an order of magnitude beyond that of the wild-type EFTRA down to 0.25 fM, which could be further applied in miRNA-21 assay from breast cancer cell lysates. Firstly, we carried out a series of polyacrylamide gel electrophoresis (PAGE) characterizations to prove the reaction mechanisms of the EFTRA (Fig. S1A †), DTNP (Fig. S1B †), and TFO (Fig. S2A †) and additionally harnessed the UV-vis absorption spectra to further verify the successful formation of TFO (Fig. S2B †), and the results were as expected. Next, cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS) for the self-assembly of the elaborated biosensor (Fig. S3 †) and the characterization of the electrode surface (Fig. S4-S6 †) all indicated the successful construction of this biosensor. Then TFO was formed under the optimal reaction time (90 min, Fig. S4A †) and pH (7.0, Fig. S4B †), and we also measured the binding constant (K a ) of TFO with a value of 1.56 10 7 M 1 (ESI, pages S15-S17 †), which demonstrates that the designed DTNP has a high binding affinity to the Fc-labeled DNA LS (B) and can be used to sensitively capture the LS (B). We also adopted SWV to characterize the biosensor under different conditions, and the results (Fig. S9 †) further displayed that this biosensing platform could effectively recognize the LS (B) and be successfully operated for accurately and sensitively measuring the EFTRA conversion efficiency (N). To obtain the experimental N, we frstly determined the relationship of the current responses of the biosensor to different concentrations of the quantifed independent Fclabeled LS (B) (Fig. S10 †) and different concentrations of the quantifed target miRNA (T) (Fig. S11 †) via EFTRA. When these two current response values corresponding to the quantifed LS (B) and the quantifed target miRNA respectively are the same, the concentration of the independent Fc-labeled LS (B) is identical to that of the output LS (B) from the input miRNA-21 via EFTRA. Thus the accurate experimental N could be obtained by proportioning the concentration of the quantifed independent Fc-labeled LS (B) and the concentration of the quantifed target miRNA under the same current response (ESI, pages S18-S21 †), and the results of the experimental wild-type EFTRA conversion efficiency (N) are shown in Table S2. † It is very important to choose suitable mismatches in the duplex ABC for achieving HEEFTRA since suitable mismatches can not only make the DG of the EFTRA more negative to induce a higher conversion efficiency but could also ensure the sufficient stability of the duplex ABC for the successful operation of the EFTRA. When the mismatches were introduced in FS (C) or LS (B), weak affinity of FS (C) for LS (B), target miRNA, or TS (D) was obtained, resulting in low EFTRA conversion efficiency (Scheme 1). Remarkably, the AP (A) with suitable mismatches not only ensured the stability of duplex ABC but could also be displaced by target miRNA more easily owing to the decreased interaction between AP (A) and FS (C), accompanied by the improved EFTRA conversion efficiency. According to the DG calculated from NUPACK as a duplex ABC stability evaluation index, we concluded that mismatched domains with number less than or equal to four at the middle instead of the tail of the displacement region in AP (A) (Fig. 1) could be permitted to introduce suitable mismatches. In view of the wide variety of mismatch types, we harnessed the theoretical computation of the EFTRA conversion efficiency to narrow their range to select the optimal mismatch type. Firstly, we simulated the EFTRA using a two-step reaction model for calculation of the theoretical N 0 : where DG 1 ¼ DG TBC + DG A DG ABC DG T and DG 2 ¼ DG B + DG T + DG DC DG TBC DG D , and T, A, B, C, and D represent target, AP, LS, FS, and TS, respectively. The equilibrium concentrations of all DNA species can then be derived by solving a set of equations: With different quantifed target miRNA, the experimental N and theoretical N 0 of the wild-type EFTRA were obtained (Table S2 †). By comparing these results, we found that the trends of N and N 0 to increase or decrease were almost the same. As shown in Fig. 2, with the target concentration increasing from 10 fM to 100 nM (1 10 14 to 1 10 7 M), the theoretical N 0 decreased. In contrast, the trend of the experimental N was consistent with the theoretical N 0 (target concentration, 1 pM to 100 nM (1 10 12 to 1 10 7 M)), while their trends were exactly opposite with the target concentration in the range of 10 fM to 1 pM (1 10 14 to 1 10 12 M), which can be ascribed to the TSDR, which is weakly thermodynamically favorable or even thermodynamically unfavorable within the limits of a sufficiently low concentration of the trigger of EFTRA (target) 15 and the difference between the real DG values and the predicted values. 32 Thus we chose a target concentration of 1.0 pM to deeply study the EFTRA conversion efficiency, at which the experimental N was practically maximum (Table S2 †). Next, we introduced all possible kinds of single-mismatched bases (Table S3 †) in AP to obtain the corresponding theoretical N 0 and experimental N. As shown in Fig. 3A, the experimental N changed in accordance with the trend of the theoretically predicted N 0 . Fig. 3B illustrates the corresponding experimental N (red curve) of some specifc multiple-mismatched bases in AP (A) (Table S3 †) with the high theoretical N 0 (blue curve). The experimental N increased with the elevated number of mismatches, which was consistent with the trend of theoretical N 0 to increase or decrease, further indicating that the theoretical N 0 could indeed be used as a reference to choose suitable mismatches for achieving HEEFTRA. However, excess mismatches of AP (A) would destroy the stability of duplex ABC, bringing false-positive EFTRA operation and a high current background signal in the DNA signal converter in the absence of target miRNA. Thus, PAGE was frstly used to monitor the stability of duplex ABC with different mismatches. As shown in Fig. 4A, lanes 1-11 correspond to duplex ABC with sequences of AP, AP-D1M1t, AP-D2M1c, AP-D3M1c, AP-D4M1c, AP-D1D4M2, AP-D2D4M2, AP-D3D4M2, AP-D1D3D4M3, AP-D2D3D4M3, and AP-D1D2D3D4M4, respectively (Table S3, † D and M represent "domain" and "mismatch" separately) and the obvious bands directly indicate that the stability of duplex ABC was almost unaffected after the introduction of single (lanes 2-5) and double (lanes 6-8) mismatches and decreased slightly after the introduction of triple mismatches (lanes 9 and 10). Compared with lanes 12-14 which respectively correspond to single strand FS, AP and LS, the band in lane 11 representing duplex ABC vanished, displaying that duplex ABC could not be constructed steadily after the introduction of quadruple mismatches in AP. Secondly, in the absence of target miRNA, we also employed SWV to study the current background signal of the biosensor after different multiple mismatches in AP (A) were introduced to further verify the stability of duplex ABC. As displayed in Fig. 4B, the SWV current responses of the biosensing platform with matched duplex ABC (curve a), single-mismatched duplex ABC (curve b), or double-mismatched duplex ABC (curve c) were hardly noticeable; however, after the introduction of triplemismatched bases (curve d) or quadruple-mismatched bases (curve e) in AP, the SWV signal responses all increased signifcantly. The high background signals of the biosensor with triple or quadruple mismatches further certifed that excess mismatches would obviously destroy the duplex ABC stability and be out of the DNA signal converter development. Integrating the above results of PAGE and SWV with the experimental N (red curve) shown in Fig. 3B, the most suitable mismatches in duplex ABC that endow the EFTRA with excellent experimental N and low background signal simultaneously were double mismatches (AP-D3D4M2) instead of the triple or quadruple mismatches from the theoretical prediction, which might be ascribed to the effects of the cation concentration, pH, etc. 33 Impressively, we introduced it in the EFTRA and achieved the HEEFTRA. The HEEFTRA was further harnessed for developing an ultrasensitive biosensor for miRNA analysis. After the suitable mismatches (AP-D3D4M2) were introduced in duplex ABC, based on the evolved HEEFTRA and the elaborated DTNP, the current response of this proposed biosensor dramatically increased with the increase of the target concentration from 0.5 fM to 100 nM (5 10 16 to 1 10 7 M) (Fig. 5A) and showed a good linear relationship with the logarithm of the miRNA-21 concentration (Fig. 5B), and the regression equation was expressed as I ¼ 0.2657 lg c + 4.1775 (R ¼ 0.9963) with detection limit down to 0.25 fM for miRNA-21 analysis (Table S6 †), exhibiting relatively desirable performance compared to the biosensor without mismatches (Fig. S11 †) and other methods (Table 1). Compared with the biosensor based on wild-type EFTRA, the proposed biosensor with mismatches in HEEFTRA exhibited a higher conversion efficiency (Fig. 6A), a wider range of detection concentration (Fig. 6B) and a lower detection limit for target assay (Table 1), demonstrating that the introduction of suitable mismatches in the displacement domain of the reactant DNA successfully achieved the HEEFTRA. Moreover, the detection of tumour-specifc circulating miRNA at ultrahigh sensitivity is of utmost signifcance for the early diagnosis and monitoring of cancer; 34 thus, there is a great need to develop new approaches or sensing media with improved miRNA detection limits owing to its low abundance in total RNA samples and the susceptibility to degradation. In view of this, we indeed explored a practically valuable HEEFTRA. In addition, the prepared biosensor based on the evolved HEEFTRA also exhibits excellent reproducibility, selectivity, and stability (Fig. S12 †). As displayed in Fig. S13, † the capacity of the elaborated biosensor for miRNA-21 detection was investigated with total RNA extraction solutions from human cancer cell lines MCF-7 and HeLa (ESI, pages S6 and S7 †) and the results (Fig. S13 †) were consistent with previous reports. Finally, the reversible pH switching of the biosensor has also been investigated (ESI, pages S29 and S30 †) and the regenerability of the proposed biosensor was excellent with more than seven pH switching cycles (Fig. S14 †). ## Conclusions In summary, we proposed a mismatch-fueled high-efficiency DNA signal converter named HEEFTRA and applied it in the construction of ultrasensitive biosensor. Overall, frst, we carved out an effective method to accurately and sensitively measure the conversion efficiency of EFTRA addressing the challenge of exact monitoring of the EFTRA conversion efficiency based on a DNA tetrahedral nanoprobe (DTNP) with multiple recognition sites at the lateral edges, high stability, and low surface-induced perturbation on the electrode surface. Second, through the introduction of suitable mismatches in reactant DNA, the conversion efficiency of EFTRA is obviously improved, providing a new idea for promoting the exploitation of the inherent properties and expanded applications of HEEFTRA. Third, as a practical application, the evolved HEEFTRA was applied to develop a biosensor with excellent specifcity, stability, reproducibility, and regenerability for the ultrasensitive detection of miRNA-21, achieving the assay of miRNA from cancer cell lysates. Given these advantages, the programmable mismatch-fueled HEEFTRA shows great potential as a new generation of DNA signal converter to construct biosensors for sensing analysis and clinical diagnosis, like the detection of nucleic acids, 47 ATP, 48 and proteins, 49 and even for applications in other areas 50 after some small adjustments of the specifc nucleic sequence in it. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Programmable mismatch-fueled high-efficiency DNA signal converter", "journal": "Royal Society of Chemistry (RSC)"}

exploiting_a_silver–bismuth_hybrid_material_as_heterogeneous_noble_metal_catalyst_for_decarboxylatio

## Abstract: Herein, we report novel catalytic methodologies for protodecarboxylations and decarboxylative deuterations of carboxylic acids utilizing a silver-containing hybrid material as a heterogeneous noble metal catalyst. After an initial batch method development, a chemically intensified continuous flow process was established in a simple packed-bed system which enabled gram-scale protodecarboxlyations without detectable structural degradation of the catalyst. The scope and applicability of the batch and flow processes were demonstrated through decarboxylations of a diverse set of aromatic carboxylic acids. Catalytic decarboxylative deuterations were achieved on the basis of the reaction conditions developed for the protodecarboxylations using D 2 O as a readily available deuterium source. † Dedicated to the memory of our friend and colleague Prof. István Pálinkó. ‡ Electronic supplementary information (ESI) available. See ## Introduction Carboxylic acids are of outstanding importance as inexpensive and easily accessible intermediates for the synthesis of an array of value-added products. 1 Among carboxylic acid transformations, protodecarboxylations and related decarboxylative couplings play a crucial role in the formation of C-C, C-X and C-H bonds, and hence they are appealing for the generation of molecular diversity. 2,3 The most common catalysts for protodecarboxylations contain copper, silver, gold, palladium or rhodium metals, typically as homogeneous sources in combination with various bases or ligands. For example, in copper-and rhodium-catalyzed examples, well-defined complexes are predominant over reusable heterogeneous sources. Palladium-catalyzed protodecarboxylations generally require high catalyst loading which severely limits their practical applicability. In addition, various hexaaluminate catalysts proved useful for decarboxylation of biomassderived carboxylic acids. Due to the high costs involved, only a few studies have been reported for gold-catalyzed protodecarboxylations. Silver-catalyzed reactions have also emerged in the field of protodecarboxylations and decarboxylative transformations, such as decarboxylative allylations and azidations, and exhibited a highly beneficial reactivity trend, comparable to that of the more costly gold-catalyzed protocols. However, with a few exceptions, 25 such reactions are promoted by soluble silver salts as non-reusable catalytic sources, 26,27 typically in the presence of various ligands, which can be regarded as a considerable drawback from an environmental point of view. Due to economic and environmental reasons, there is a continuously growing need for heterogeneous noble metal cat-alysts. 31 However, immobilization of metal catalysts on various prefabricated supports is often accompanied by reduced selectivity or loss of activity, and in the case of inadequate catalyst-support interactions, leaching of the metal component may lead to substantial environmental concerns. 32,33 Nowadays, in organic synthesis silver catalysis is considered as a significant methodology, which is due to its wide applicability, environmentally-benign nature and its lower costs compared with other precious noble metals such as gold, platinum or palladium. 34,35 Typical synthetic applications of silver catalysis rely on Ag(I) salts or complexes as homogeneous sources for the catalytically active metal. As concerns heterogeneous silver sources, supported nanoparticles (nanosilver) are the most widely applied. 33,39,40 Such heterogeneous materials are easily obtained via immobilization on various surfaces, however their main limitation is weak catalystsupport interactions which give rise to unsatisfactory stability and limits their practical synthetic utilities, especially under demanding reaction conditions, such as high-temperature continuous flow conditions or in the presence of coordinating ligands. On the basis of a naturally occurring mineral, called beyerite, we recently developed a heterogeneous silverbismuth hybrid material (AgBi-HM) with structurally-bound silver catalytic centers. 41 The material exhibited a layered structure and contained Ag(I) and Bi(III) cationic and carbonate anionic components with silver ion as the minor cationic component. As compared with traditionally immobilized catalysts, structurally-bound catalytic centres imply increased thermodynamic stability and robustness, and exhibit an increased tolerance against challenging reaction conditions and improved compatibility with various reactants and solvents. 42 Continuous flow reaction technology in combination with heterogeneous catalysis have attracted significant attention in recent years, and now comprise a powerful methodology for the synthesis of an array of useful products. Heterogeneous catalysts can easily be handled, recycled and reused in packedbed reactors, moreover, unlike in traditional batch processes, separation from the reaction products is really straightforward. 57 Due to the enhanced control over the most important reaction conditions (e.g. residence time and temperature), reaction selectivity can easily be improved while less waste is generated. 61,62 Moreover, in loaded catalyst columns, the continuous stream of reactants interacts with a superstoichiometric amount of catalyst species, which improves reaction rates significantly. On the downside, with increasing reactor dimensions scale-up may involve difficulties, such as insufficient intraparticle heat transfer rates, intraparticle diffusion limitations as well as susceptibility to liquid maldistribution. 66 However, if catalyst deactivation and leaching can be eliminated, the scale of production becomes a direct function of the process time without modifying the reactor geometry (i.e. scaleout). In spite of these obvious benefits, there are very few precedents for heterogeneous silver-catalysts being utilized in continuous flow processes, 70,71 which may be explained by the fact that stable and robust heterogeneous silver catalyst are at scarce. 41,42,72,73 To the best of our knowledge, protodecarboxylations promoted by heterogeneous noble metal catalysts have not yet been achieved under efficient continuous flow conditions. We speculated that our silver-containing hybrid material may act as a ligand-free heterogeneous silver catalyst for protodecarboxylations, and because of its stability and robustness, not only under batch but also under more demanding flow conditions. We intended to investigate the flow reactions in a high-temperature packed-bed reactor system to exploit extended parameter spaces, and to study the possibility of chemical intensification as compared with the batch process. Considering the outstanding significance of deuterated compounds in chemistry, biochemistry, environmental sciences and also in pharmacological research, 74,75 we were intrigued to explore not only protodecarboxylations but also decarboxylative deuterations as facile and site-specific access to valuable deuterium-labelled compounds. 76,77 Our results are presented herein. ## General information All chemicals used were analytical grade and were applied without further purification. Reaction products were characterized by NMR spectroscopy and mass spectrometry. 1 H NMR and 13 C NMR spectra were recorded on a Bruker Avance NEO 500 spectrometer, in CDCl 3 as solvent, with tetramethylsilane as internal standard at 500.1 and 125 MHz, respectively. GC-MS analyses were performed on a Thermo Scientific Trace 1310 Gas Chromatograph coupled with a Thermo Scientific ISQ QD Single Quadrupole Mass Spectrometer using a Thermo Scientific TG-SQC column (15 m × 0.25 mm ID × 0.25 μm film). Measurement parameters were as follows. Column oven temperature: from 50 to 300 °C at 15 °C min −1 ; injection temperature: 240 °C; ion source temperature: 200 °C; electrospray ionization: 70 eV; carrier gas: He at 1.5 mL min −1 injection volume: 2 μL; split ratio: 1 : 33.3; and mass range: 25-500 m/z. ## Synthesis and characterization of the AgBi-HM AgBi-HM was synthesized by using the urea hydrolysis method according to a modified version of our procedure reported previously. 42 AgNO 3 (3.73 g) and Bi(NO 3 ) 3 •5H 2 O (5.36 g) were dissolved in 50-50 mL 5 wt% nitric acid and the solutions were combined. Urea (7.05 g) dissolved in 100 mL of deionized water was next added to the mixture which was then placed into an oven for 24 h at 105 °C. The obtained material was next filtrated, washed with aqueous thiosulfate solution, water and ethanol four times, and dried at 60 °C to obtain the final product. The as-prepared material was fully characterized by means of diverse instrumental techniques as detailed earlier. 40,41 The X-ray diffraction (XRD) patterns were recorded on a Rigaku XRD-MiniFlex II instrument applying CuKα radiation (λ = 0.15418 nm), 40 kV accelerating voltage at 30 mA. The morphology of the as-prepared and treated samples were studied by scanning electron microscopy (SEM). The SEM images were registered on an S-4700 scanning electron microscope (Hitachi, Japan) with accelerating voltage of 10-18 kV. The actual Ag/Bi metal ratios in the samples were determined with energy dispersive X-ray analysis (EDX) measurements (Röntec QX2 spectrometer equipped with Be window coupled to the microscope). More detailed images, both of the as-prepared and the used samples, were taken by transmission electron microscopy (TEM). For these measurements, an FEI Tecnai™ G2 20 X-Twin type instrument was applied, operating at an acceleration voltage of 200 kV. The thermal behaviour of the catalyst samples were investigated by thermogravimetry (TG) and differential thermogravimetry (DTG) using a Setaram Labsys derivatograph operating in air at 5 °C min −1 heating rate. For the measurements, 20-30 mg of the samples were applied. The amount of metal ions was measured by ICP-AES on a Thermo Jarell Ash ICAP 61E instrument. Before measurements, a few milligrams of the samples measured with analytical accuracy were digested in 1 mL cc. nitric acid; then, they were diluted with distilled water to 50 mL and filtered. ## General procedure for the batch reactions A typical procedure for the decarboxylation and decarboxylative deuteration reactions is as follows. N,N-Dimethylformamide (DMF, 3 mL), the appropriate carboxylic acid (0.45 mmol, 0.15 M, 1 equiv.), KOH (6 mg, 15 mol%) and AgBi-HM as catalyst (60 mg, corresponding to 5 mol% Ag loading) were combined in an oven-dried Schlenk tube equipped with a magnetic stir bar. In case of decarboxylative deuteration, 10 equiv. of D 2 O (90 µL) was also added to the reaction mixture. After stirring for 24 h at 110 °C, the reaction mixture was cooled to room temperature, and the catalyst was filtered off. The crude products were diluted with diethyl ether and were washed with aqueous NaHCO 3 and brine. The combined organic layers were dried over Na 2 SO 4 , and concentrated under reduced pressure. The crude products were checked by NMR spectroscopy to determine conversion and selectivity. The products of the batch reactions were characterized by NMR and GC-MS techniques. In case deuterondecarboxylations, deuterium contents were determined from the relative intensities of the 1 H NMR indicator signals. Characterization data can be found in the ESI. ‡ ## Investigation of the catalyst reusability under batch conditions For investigation of catalyst reusability, the decarboxylation of 2-nitrobenzoic acid was carried out multiple times utilizing a single portion of catalyst. DMF (3 mL), 2-nitrobenzoic acid (0.45 mmol, 0.15 M, 1 equiv.), KOH (6 mg, 15 mol%) and AgBi-HM as catalyst (60 mg, corresponding to 5 mol% Ag loading) were combined in an oven-dried Schlenk tube equipped with a magnetic stir bar. The reaction mixture was stirred for 24 h at 110 °C. The mixture was next cooled to room temperature, and the solid material was removed by centrifugation. The liquid phase was extracted, dried and evaporated as detailed in section 2.3. The removed catalyst was washed with DMF (four times) and was dried in nitrogen flow before the next reaction cycle. Conversion and selectivity were determined after each cycle by using 1 H NMR. ## General procedure for the flow reactions To carry out the decarboxylation and decarboxylative deuteration reactions under flow conditions, a simple continuous flow set-up was assembled as shown in Fig. 1. The system consisted of an HPLC pump (JASCO PU-2085), a stainless steel HPLC column with internal dimensions of 4.6 × 100 mm as catalyst bed and a 5-bar backpressure regulator (BPR) from IDEX to prevent solvent boil over. The column encompassed 2 g of AgBi-HM as catalyst. For each reaction, the corresponding carboxylic acid (c = 0.1 M) and 15 mol% KOH were dissolved in acetonitrile (MeCN) or DMF. In order to achieve a clear solution, 20 equiv. of H 2 O was also added to the reaction mixture. In case of deuterodecarboxylation reactions, 20 equiv. D 2 O was added to the reaction mixture as deuterium source. In each run, 4 mL of product solution was collected under steady-state conditions. Between two experiments, the system was washed for 20 min by pumping the appropriate solvent at a flow rate of 0.5 mL min −1 . When DMF was used as solvent, the crude product was worked-up similarly as detailed in section 2. contact of the dye with the column and the moment when the coloured solution appeared at column the outlet was measured. ## Decarboxylation of carboxylic acids under batch conditions In order to achieve an initial picture on the catalytic activity of the silver-containing hybrid material in decarboxylation of carboxylic acids, batch reactions were explored first. The decarboxylation of 2-nitrobenzoic acid was chosen as model reaction to demonstrate the performance of the AgBi-HM in comparison with various commercially available silver and copper salts as the most typical homogeneous catalytic sources for this reaction type (Fig. 2). Based on literature data, 25,78 DMF was selected as solvent, and the reaction mixture containing the substrate (0.15 M) together with 5 mol% of the appropriate catalyst and 15 mol% of KOH as base was stirred for 24 h at 110 °C. It was corroborated, that product formation was not occurring without any catalyst present. Gratifyingly, the application of the hybrid material as catalyst resulted quantitative and selective decarboxylation to nitrobenzene. AgOAc, Ag 2 O, Ag 2 CO 3 and AgNO 3 as catalyst gave slightly lower conversions (95-97%) and 100% selectivity in each cases. In contrast to silver catalysts, copper salts performed poorer. In the presence of CuOAc and Cu(NO 3 ) 2 , conversion was 68% and 70%, respectively, whereas CuBr 2 was proven even less effective with a conversion of merely 39%. In all the copper-catalyzed reactions, potassium 2-nitrobenzoate appeared in the reaction mixture. Considering that the reaction is initiated by deprotonation of the carboxylic acid, the presence of the corresponding potassium salt as side product therefore indicates the incompleteness of the reaction. 25 As corroborated by a test reaction carried out in the presence of 5 mol% of Bi(NO 3 ) 3 •5H 2 O, the Bi(III) component of the hybrid material is inactive in decarboxylation of 2-nitrobenzoic acid. After achieving promising preliminary results, the effects of the major reaction conditions were next explored. Upon investigation of solvent effects (Table 1), the best results were achieved by using DMF (entry 1). MeCN and N,N-dimethylacetamide (DMA) also gave acceptable conversions (85% and 62%, respectively) and high selectivities (100% and 85%, respectively; entries 2 and 3). In EtOAc and toluene only trace amounts of nitrobenzene formation was detected (entries 4 and 5), whereas in N-methyl-2-pyrrolidone (NMP) and dimethyl sulfoxide (DMSO), no decarboxylation occurred (entries 6 and 7). As concerns reaction time, 24 h was required for completion, lower reaction times gave incomplete transformations (Fig. S1 ‡). As was expected, decarboxylation was not taking place at temperatures ≤50 °C, however conversion started to increase at 80 °C and reached completion at 110 °C (Fig. S1 ‡). The reaction gave the best results with substrate concentrations of 0.1 or 0.15 M (Table S1 ‡) The optimum value of the catalyst loading was 5 mol% as lower amounts resulted in decrease of the conversion (Table 2, entries 1-4). Upon investigation of the effects of the amount of the extraneous KOH (Table 2, entries 5-9), it was observed that without base the reaction gives only traces of the decarboxylated product; however only catalytic amounts are required for completion (e.g. 100% conversion was achieved with 15 mol% KOH). This is in accordance with the mechanistic proposal of Jaenicke and co-workers suggesting a negatively charged aryl-silver intermediate upon decarboxylation which is responsible for deprotonation after the base-promoted initiation of the reaction. 25 In our study, KOH was selected as base as it involved no precipitation and ensured a pumpable clear solution when being combined with the substrate which is crucial when considering the upcoming continuous flow experiments. Having established an optimal set of conditions for the decarboxylation of the model compound (5 mol% catalyst loading, 15 mol% KOH as base, DMF as solvent, 0.15 M substrate concentration, 110 °C temperature and 24 h reaction time), we set out to investigate the scope and applicability of the batch process (Table 3). Besides 2-nitrobenzoic acid (entry 1), its 5-methoxy-substituted derivative as well as 3,5-dinitrobenzoic acid underwent quantitative and selective protodecarboxylations (entries 2 and 3). The reaction tolerated well the replacement of the ortho-nitro substituent with bromine or methoxy groups, and gave good conversions (80% and 74%, respectively) and 100% selectivities in reactions of the corresponding benzoic acid derivatives (entries 4 and 5). Despite the higher steric hindrance, decarboxylation of 2,6-dimethoxybenzoic acid was also successful, although conversion was somewhat lower (65%) than in the case of the mono-substituted derivative (entry 6 vs. entry 5). Interestingly, decarboxylation of 2-chlorobenzoic acid and 2-hydroxybenzoic acid (salicylic acid) were not successful (entries 7 and 8), however 2,4-dichlorobenzoic acid proved as an excellent substrate and gave the corresponding dichlorobenzene with 92% conversion and 100% selectivity (entry 9). Selective decarboxylation of 1-naphtolic acid to naphthalene was also possible, however only with a moderate conversion of 49% (entry 10). To our delight, selective decarboxylation of heteroaromatic carboxylic acids, such as thiophene-2-carboxylic acid and nicotinic acid, proceeded with excellent conversions (100% and 86%, respectively; entries 11 and 12). Similarly high conversions (97-100%) and selectivities were achieved in decarboxylations of fused heteroaromatic substrates, such as indole-3-carboxylic acid, coumarin-3-carboxylic acid and chromone-3-carboxylic (entries 13-15). Decarboxylations of metha-and para-monosubstituted benzoic acid derivates, such as 3-and 4-nitrobenzoic acid, were also attempted, however in these cases no reaction occurred. These results are in accordance with earlier literature findings suggesting the formation of a metal-centered carboxylate intermediate which is stabilized by the electronic effects of the substituent(s) on the aromatic rings. 79,80 Moreover, decarboxylation of aliphatic carboxylic acids, such as hexanoic acid and levulinic acid, was proven unsuccessful using this methodology. Note that isolated yield was determined in some representative instances. One of the main benefits of heterogeneous catalysis is the ability to reuse and recycle the catalytic material. In order to evaluate this sustainable feature of the AgBi-HM, protodecarbxylation of 2-nitrobenzoic acid was performed repeatedly under optimized conditions utilizing the same portion of catalyst for each reactions. The used hybrid material was removed between each cycle by centrifugation and after washing and drying, it was simply reused. Gratifyingly, no decrease in catalytic activity or selectivity occurred during the first 7 consecutive catalytic cycles, and conversion was around 90% even after the 10 th reaction which implies the significant stability and robustness of the catalytic material (Fig. 3). ## Decarboxylation of carboxylic acids under continuous flow conditions After achieving convincing batch results, we next turned our attention to continuous flow operation with the aim to improve the efficacy and sustainability of the catalytic process. As detailed in the Experimental, AgBi-HM was employed in a simple packed-bed reactor setup (see also Fig. 1). Similarly as in the batch study, the effects of the reaction conditions were investigated again using the decarboxylation of 2-nitrobenzoic acid as a model reaction. The effects of solvents which gave good results in the batch reactions were explored again under flow conditions. For this, the catalyst bed was heated to 170 °C, and the solution of the substrate together with 15mol% KOH was pumped continuously at 50 µL min −1 flow rate. Gratifyingly, under these conditions, MeCN performed superior compared to DMF and DMA, and resulted selective decarboxylation with conversions of 100% and 96% at 0.1 M and 0.15 M substrate concentrations, respectively (Table 4). This is a remarkable improve-ment considering that MeCN is much more acceptable from environmental aspects than DMF which performed best under batch conditions (Table 1, entry 1). 81 Upon investigation of the effects of the residence time and reaction temperature, it was verified that a significant chemical intensification is possible under flow conditions. Due to the backpressure applied, it was possible to easily overheat the reaction mixture and to study the effects of temperatures far above the boiling point of MeCN. As shown in Fig. 4, quantitat- ive and selective decarboxylation could be achieved at 170 °C while the reaction mixture (containing the substrate in 0.1 M concentration together with 15 mol% KOH) was streamed at 50 µL min −1 flow rate. Notably, this corresponded to a residence time of only 10.5 min which is a significant improvement compared to the batch reaction time of 24 h. When residence time was decreased to approximately 3.5 min (150 µL min −1 flow rate), the conversion of the decarboxylation was still 75% at 170 °C. When residence time was kept constant at 10.5 min, a rapid decrease of conversion was observed with the temperature; for example at 100 °C conversion was only 13%. A range of aromatic carboxylic acids exhibiting diverse substitution patterns were next submitted to the optimized flow conditions (Table 5). Similarly as in the batch reactions, quantitative and selective decarboxylation was achieved in cases of 2-nitrobenzoic acid, its 5-methoxy-substituted derivative as well as 3,5-dinitrobenzoic acid (entries 1-3). To our delight, the flow protocol proved more effective in numerous reactions than the batch method. For example, 2-bromo-, 2-methoxy-as well as 2,6-dimethoxybenzoic acid furnished quantitative conversions (entries 4-6), whereas in batch, conversions were much lower. Notably, selective decarboxylations of 2-chloroand 2-hydroxybenzoic acid were achieved successfully under flow conditions (conversions were 100% and 23%, respectively; entries 7 and 8), whereas these substrates remained inert under batch conditions. 2,4-Dichlorobenzoic acid and 1-naphtolic acid were also successfully decarboxylated and gave similar conversions than in the corresponding batch reactions (entries 9 and 10). Fused heteroaromatic substrates showed excellent reactivity, and gave quantitative conversion and 100% selectivity, similarly as under batch conditions (entries 11-13). Unfortunately, flow reactions of thiophene-2-carboxylic acid and nicotinic acid could not be evaluated due to possible deposition of the substrates and/or the products within the catalyst column. (Isolated yield was also determined for some representative examples.) In order to investigate the preparative capabilities of the AgBi-HM catalyzed protodecarboxylation under flow conditions, the reaction of 2-nitrobenzoic acid was scaled-out (Fig. 5). With the aim to maximize the productivity of the synthesis, the flow rate was increased to 100 µL min −1 (approx. 5 min residence time), all further reaction parameters were kept at the previously optimized values (0.1 M substrate concentration, 15 mol% of KOH as base, MeCN as solvent, 170 °C temperature). A 20 h reaction window was explored, with conversion and selectivity being determined in every hour to obtain a clear view of the actual catalyst activity. Gratifyingly, the packed-bed system proved highly stable. No decrease in activity or selectivity occurred in the first 18 h of the experiment: conversion remained steady around 80-85% and no side product formation occurred. In the last two hours, a slight loss of catalytic activity was detected, however after 20 h, at the end of the experiment, a satisfying conversion of 71% could still be achieved. Finally, as the result of the scale-out, 1.207 g of nitrobenzene was isolated which corresponded to an overall yield of 82%. ## Characterization of used AgBi-HM samples With the aim to evaluate catalyst stability and robustness, AgBi-HM samples previously used in batch recycling experiments as well as in flow scale-out were examined extensively by various instrumental techniques. The materials were character- ized by TG, SEM (SEM-EDX), TEM and XRD measurements, and the structure of the used catalyst samples was compared to that of the as-prepared material (Fig. 6). Thermal analysis revealed that the original structure was kept up to 380 °C, and weight losses occurred in three endothermic steps which was also observed in both used catalyst samples (Fig. 6a). In the case of the AgBi-HM sample used in the batch recycling experiment, slightly greater weight loss could be observed at lower temperatures which may be explained by trace amounts of organic deposition on the surface. The X-ray patterns of both used samples seemed to be the same as was experienced in case of the as-prepared sample (Fig. 6b), there was no evidence of structural degradation visible. Identification of the X-ray patterns were accomplished on the basis of our previous work. 42 These results provided some further information about primer crystallite size of the composite calculated by using the well-known Scherrer equation. This resulted in an average primer crystallite size of 20.98 nm, not only for the as-prepared catalyst but also for the used ones. As shown earlier, 41 the SEM image of the freshlymade catalyst displayed a lamellar ( plate-like) morphology which was also observed in the used material (Fig. 6c). Additionally, this observation was also strengthened by TEM images, in which well-aggregated plates with a secondary particle size of around 100 nm could be seen for the as-prepared as well as for the used catalyst sample (Fig. S2 ‡). The SEM images also confirmed that organic contaminants in the form of larger aggregates (up to 10 µm) remained on the surface which makes more difficult to identify the original morphology. The SEM-EDX elemental maps demonstrated that the silver and bismuth ions are located evenly in the used sample as well (Fig. S3 ‡). ICP-AES measurements confirmed that the quantity of silver and bismuth ions are in arrangement with the as-prepared sample considering errors of measurements. 41 Taking into account all the characterization data, it can be ascertained that the AgBi-HM is a highly robust heterogeneous catalyst which proved to be invariable in a structural point of view after extensive and demanding use under batch or flow conditions. ## Decarboxylative deuterations Due to its relatively good availability and also because of the considerably large isotope effect, deuterium is of outstanding importance among stable isotopes used for labelling studies. 82,83 Synthetic protocols that incorporate deuterium into various organic substances have therefore many applications in medicinal, analytical or pharmaceutical chemistry. 74 Deuterium-labelled compounds are typically applied as analytical standards, for the evaluation of the metabolic pathways or in tracer studies to investigate pharmacokinetics, catalytic cycles and reaction pathways. As exemplified by Austedo®, the first deuterated drug marketed, pharmaceutical ingredients may also be potentiated by deuterium exchange. 87,88 In contrast to deuterations of C-C or C-X (X = hetero atom) multiple bonds, synthetic processes for the site-specific incorporation of a single deuterium into an aromatic ring are more challenging. 74,76,77,94,95 In most cases, these methods involve halogen/D exchange and are commonly mediated by strong bases which severely limits the functional group tolerance. 96 Furthermore, catalytic and acid-or basemediated H/D exchange reactions are also available, however, unlike halogen/D exchange reactions, these often involve selectivity issues. Inspired by these limitations, we were intrigued to explore decarboxylative deuterations of benzoic acid derivatives in the presence of the silver-containing hybrid material as catalyst. Initially, reactions were investigated in batch (Table 6), under conditions optimized for the protodecarboxylations earlier (0.15 M substrate concentration, 5 mol% AgBi-HM as catalyst, 15 mol% of KOH as base, DMF as solvent, 110 °C temperature and 24 h reaction time). As deuterium source, 10 equiv. of D 2 O was added to the reaction mixture. We were satisfied to find that with this simple protocol, deuterodecarboxylations of various nitrobenzoic acids as well as 2-bromo-, 2,6-dimethoxyand 2,4-dichlorobenzoic acid went smoothly. Excellent conversions (79-100%) and 100% chemoselectivity were achieved in all cases. In all reactions, deuteration was highly favoured over incidental hydrogen incorporation as indicated by deuterium contents of 76-100%. Continuous flow deuterodecarboxylations were next attempted in a packed bed reactor charged with AgBi-HM. Reaction conditions were simply taken from the protodecarboxylation experiments (0.1 M substrate concentration, 15 mol% of KOH as base, MeCN as solvent, 170 °C temperature, 50 µL min −1 flow rate, 10.5 min residence time). In these cases, 20 equiv. of D 2 O was used as deuterium source to achieve high deuterium contents. Gratifyingly, in all reactions investigated (Table 7), quantitative conversion and 100% chemoselectivity was achieved, and deuterium incorporation was also perfect in most cases. ## Conclusion A silver-containing hybrid material with structurally-bound catalytic centers has been exploited as heterogenous noble metal catalyst for decarboxylations of carboxylic acids under batch and continuous flow conditions. It proved to be a robust, efficiently recyclable and highly active ligand-free catalyst which outperformed the most typical homogeneous catalytic sources in the decarboxylation of 2-nitrobenzoic acid as model reaction. Although, under batch conditions the catalyst performed best in DMF as solvent, the application of a simple packed-bed flow system enabled a solvent switch to the environmentally more acceptable MeCN. After the optimization of the most important reaction conditions, the selective decarboxylation of diversely substituted aromatic carboxylic acids were achieved with high conversions either in batch or in continuous flow mode. Importantly, the application of continuous flow conditions offered a marked chemical intensification as compared with the batch reactions (10 min residence time vs. 24 h reaction time) and ensured time-efficient syntheses. The preparative utility of the flow process was verified by a 20 h scale-out run in which the multigram-scale decarboxylation of 2-nitrobenzoic acid was achieved without notable decrease in the activity and without detectable degradation of the structure of catalyst. On the basis of the reaction conditions established for the protodecarboxylations, heterogeneous catalytic batch as well as flow methodologies were developed for decarboxylative deuterations in the presence of D 2 O as a readily available deuterium source. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Exploiting a silver\u2013bismuth hybrid material as heterogeneous noble metal catalyst for decarboxylations and decarboxylative deuterations of carboxylic acids under batch and continuous flow conditions", "journal": "Royal Society of Chemistry (RSC)"}

stable_α/δ_phase_junction_of_formamidinium_lead_iodide_perovskites_for_enhanced_near-infrared_emissi

## Abstract: Although formamidinium lead iodide (FAPbI 3 ) perovskite has shown great promise in the field of perovskitebased optoelectronic devices, it suffers the complications of a structural phase transition from a black perovskite phase (a-FAPbI 3 ) to a yellow non-perovskite phase (d-FAPbI 3 ). Generally, it is pivotal to avoid d-FAPbI 3 since only a-FAPbI 3 is desirable for photoelectric conversion and near-infrared (NIR) emission. However, herein, we firstly exploited the undesirable d-FAPbI 3 to enable structurally stable, pure FAPbI 3 films with a controllable a/d phase junction at low annealing temperature (60 C) through stoichiometrically modified precursors (FAI/PbI 2 ¼ 1.1-1.5). The a/d phase junction contributes to a striking stabilization of the perovskite phase of FAPbI 3 at low temperature and significantly enhanced NIR emission at 780 nm, which is markedly different from pure a-FAPbI 3 (815 nm). In particular, the optimal a/d phase junction with FAI/PbI 2 ¼ 1.2 exhibited preferable long-term stability against humidity and high PLQY of 6.9%, nearly 10-fold higher than that of pure a-FAPbI 3 (0.7%). The present study opens a new approach to realize highly stable and efficient emitting perovskite materials by utilizing the phase junctions. ## Introduction Organic-inorganic hybrid perovskites, especially CH 3 NH 3 PbI 3 (MAPbI 3 ) and its mixed halide analogues, have attracted tremendous attention as light absorbers for solar cells. The power conversion efficiencies (PCEs) of perovskite solar cells (PSCs) have been boosted from 3.8% in 2009 to over 20% in 2016. Recently, these hybrid perovskite materials also have shown great promise as light emitters for lasers and lightemitting diodes because of their very high color purity, low material cost and simply tunable band gaps. 7,8 Bright photoluminescence (PL) and electroluminescence (EL) in the nearinfrared (NIR) and visible region have been demonstrated based on solution-processed MAPbX 3 perovskites by tuning the halide composition at room temperature. Comparing with the prototype MAPbI 3 , formamidinium lead iodide (FAPbI 3 ) has recently emerged as a promising candidate for use in PSCs because of its favorable band gap (1.47 eV) and superior photoand thermal stability. Moreover, FAPbI 3 perovskite demonstrates a true NIR emission around 815 nm, 23,24 also making it a preferable candidate for NIR emitters in terms of achieving more robust perovskite light-emitting diodes. However, FAPbI 3 was reported to be unstable in phase structure. It can crystallize into either a black trigonal perovskite phase (a-FAPbI 3 ) or a yellow hexagonal non-perovskite phase (d-FAPbI 3 ) depending on the heat-treatment temperature. It usually requires a very high phase transition temperature (up to 160 C) to obtain black a-FAPbI 3 , which is desirable for photoelectric conversion and NIR emission. Unfortunately, such high temperature can result in partial decomposition of FAPbI 3 into PbI 2 . 21,30 Moreover, the obtained black a-FAPbI 3 can turn into the undesirable yellow d-FAPbI 3 under an ambient humid atmosphere. 24, To date, most work on the FA-based perovskites focuses on the stabilization of black a-FAPbI 3 at relatively low temperature. Very recently, several groups have demonstrated that mixed cations such as MA/FA or Cs/FA can fully avoid the formation of d-FAPbI 3 . 25, But as far as we know, the stabilization of the pure FA phase has not been conclusively demonstrated. Herein, we for the frst time exploited the undesirable d-FAPbI 3 to achieve structurally stable FAPbI 3 perovskite flms at low annealing temperature (60 C) through fne and controllable stoichiometry modifcation (FAI/PbI 2 ¼ 1.1-1.5). All the obtained flms demonstrated a unique a/d phase junction together with a signifcantly enhanced phase stability and NIR emission at 780 nm, which is markedly different from the pure a-FAPbI 3 (815 nm). The hypochromic emission was attributed to a slight bandgap widening due to band bending at the a/d phase junction. Moreover, such a phase junction can substantially decrease the dielectric constant and grain size of the flms, thus effectively increasing the exciton binding energy and confning the excitons in the nanograins for high emission. Most notably, the optimal a/d phase junction with FAI/PbI 2 ¼ 1.2 exhibited superior long-term stability even under an ambient humid atmosphere and the highest PL quantum efficiency (PLQY ¼ 6.9%), nearly one order of magnitude higher than that of pure a-FAPbI 3 (PLQY ¼ 0.7%). ## Results and discussion Firstly, the phase structure of FAPbI 3 was systematically investigated by varying the molar ratio of FAI/PbI 2 in the precursor solution from 1 to 3. All the flms were readily prepared and characterized by X-ray diffraction. For clarity, we name the as-prepared flm x M-FAI in the following discussion, x being the molar ratio of FAI/PbI 2 in the precursor solution. In Fig. 1a, we show XRD patterns zoomed in the 2q peak from 8 to 24 (the complete diffraction patterns are shown in Fig. S1 and S2 †). For the 1 M-FAI flm annealed at 60 C, the diffraction peaks at 11.8 (010) and 16.3 (011) show the formation of pure d-FAPbI 3 . 24 However, when the annealing temperature reached 170 C, the observed peaks at 13.9 (111) and 19.7 (012) indicate the formation of pure a-FAPbI 3 . 24,26 Interestingly, as the precursor solution changed from 1.1 M-FAI to 1.5 M-FAI, all the obtained flms at a low annealing temperature of 60 C exhibited two obvious diffraction peaks at 13.9 and 11.8 . The former corresponds to a-FAPbI 3 , while the latter belongs to d-FAPbI 3 . Moreover, along with the increase of excess FAI, the diffraction peak at 13.9 gradually increased while the peak at 11.8 relatively decreased, suggesting a growing volume fraction of a-FAPbI 3 . When FAI/PbI 2 ¼ 2 (2 M-FAI), the (010) diffraction corresponding to d-FAPbI 3 disappeared and a new peak appeared at 10.4 . Although the exact phase formed with 2 M-FAI has not been identifed yet, this small angle diffraction peak might be related to a large interplanar spacing of some layered intermediate phase. 31 In the case of 3 M-FAI, the diffraction peaks at 10.3 and 9.9 could be assigned to a reported one-dimensional compound FA 3 PbI 5 (Fig. S3 †). 32 It should be noted that the excess FAI is not detected in all of the XRD patterns (Fig. S4 †), which is presumably due to the low crystallinity of FAI in the as-deposited flms as previously reported. 33 Another possible reason is that the excess FAI can insert into the crystal lattice and form some low dimensional perovskites as in the case of 2 M-FAI and 3 M-FAI. We also fabricated the flm from an under-stoichiometric precursor with FAI : PbI 2 ¼ 0.9 : 1, which gave a mixture of PbI 2 and d-FAPbI 3 (XRD pattern shown in Fig. S5 †). Overall, by carefully tuning the molar ratio of FAI/PbI 2 between 1.1 and 1.5 in the precursor solution, we are able to obtain a series of mixed-phase FAPbI 3 flms with controllable a/d composition. Confocal fluorescence microscopy was then utilized to characterize the crystal morphology and optical properties of the obtained FAPbI 3 flms. 34 As shown in Fig. 1b, the d-FAPbI 3 (P6 3 mc) crystals are in the shape of a hexagon, while the a-FAPbI 3 crystals (P3m1) look like a triangle (regular or distorted). 24 Generally, the shape of crystals mainly comes from the symmetry of the lattice structure. When excess FAI is present, the shape of the crystals exhibits a gradual transition from triangles for 1.1 M-FAI to regular polygons for 1.2 M-FAI and then to distorted polygons for 1.3 M-FAI to 1.5 M-FAI, all of which were quite different from those of pure a or d-FAPbI 3 . In the cases of 2 M-FAI and 3 M-FAI, the crystals become round and hexagon-like, respectively. Concomitantly, the images also qualitatively reflect on the brightness of each flm. Notably, the mixed-phase flms look brighter in the microscopy images, clearly indicating their higher PL intensity, especially for the flm with 1.2 M-FAI. We then conducted steady state PL measurements on these flms (Fig. 2a). For the FAPbI 3 flms of the pure phases, the PL spectra were in accordance with the previous observation, a broad emission from 400 to 650 nm for d-FAPbI 3 and a maximum peak at 815 nm for a-FAPbI 3 . 23,24 In comparison, all the mixed-phase flms demonstrated distinct emission profles with the same maximum peak around 780 nm, which is very different from either of the pure a or d phases or the simple mixture of the a and d phase. The flms with 2 M-FAI and 3 M-FAI showed much more blue-shifted emissions due to their relatively larger band gaps of low dimensional perovskites. 31,32 Apart from the tunable emission peak wavelength, the relative PL intensity increases and reaches its maximum at 1.2 M-FAI and then decreases with more FAI towards 3 M-FAI (Fig. 2b), which was consistent with the observed brightness change by the confocal fluorescence microscopy analysis. It was very striking that all the as-prepared mixed-phase FAPbI 3 showed much higher PL intensities than the pure phases. Especially for the 1.2 M-FAI flm, the relative PL intensity was nearly 20-fold and 218-fold higher than that of pure a-FAPbI 3 and d-FAPbI 3 , respectively. The absolute PL quantum yield (PLQY) of the 1.2 M-FAI flm was measured to be 6.9%, which is comparable to that of the MAPbI 3 flm, 34 but nearly one order of magnitude higher than those of pure a-FAPbI 3 (0.7%) and d-FAPbI 3 (0.3%). We further performed fluorescence lifetime imaging microscopy (FLIM) on these flms (Fig. S6 †). The PL decay of the 1.2 M-FAI flm exhibited the longest lifetime of 6.67 ns in all cases, which indicated a much lower defect concentration. 35,36 The distinct optical properties clearly indicated that the as-prepared mixed-phase FAPbI 3 flms, in particular for the 1.2 M-FAI flm, should have unique microstructures and electronic structures. Here, it is worth noting the residue of excess FAI in the mixed-phase FAPbI 3 flms. Although it is not detected by XRD, it becomes interesting to wonder what role the FAI residue plays in the PL properties of the mixed-phase flm. We further conducted XRD and PL measurements on the flm from the same over-stoichiometric precursor at 1.2 M-FAI but annealed at 170 C. The XRD pattern of the flm clearly showed that d-FAPbI 3 vanished because of the high annealing temperature and only a-FAPbI 3 remained mixed with the residue of excess FAI (Fig. S7a †). However, this flm gave very similar PL spectra and comparable PLQY around 0.7% to that of the pure a-FAPbI 3 flm from the stoichiometric 1 M-FAI precursor (Fig. S7b †). This result indicated that the residue of excess FAI may not play a role in affecting the PL properties of a-FAPbI 3 . In comparison, for the 1.2 M-FAI flm annealed at a low temperature of 60 C, d-FAPbI 3 was involved and mixed with a-FAPbI 3 . Consequently, great changes took place in the PL properties of the flm, which showed markedly hypochromic emission centered at 780 nm and 10-fold enhanced PLQY. These results strongly suggested that the involved d-FAPbI 3 and concomitant co-existence of a-FAPbI 3 and d-FAPbI 3 would be highly correlated with the distinct PL properties of the mixed-phase flms. To visualize such mixed-phase structures, the 1.2 M-FAI flm was investigated by high resolution transmission electron microscopy (HRTEM). As shown in Fig. 3a, well defned lattice fringes with a separation of 3.17 and 2.38 could be well indexed to the (222) plane of a-FAPbI 3 and the (031) plane of d-FAPbI 3 , respectively. 24 Such HRTEM results (other images shown in Fig. S8 †), clearly showed that a junction structure between the a and d phases is formed in the as-prepared FAPbI 3 flms. When excited with different wavelengths of light (Fig. S9a †), the 1.2 M-FAI flm always exhibits the same PL emission peak around 780 nm, regardless of whether the excitation energy was strong enough to excite d-FAPbI 3 individually. There is a 35 nm blue shift compared to that of the pure a phase. These results clearly reveal that the existence of the a/d phase junction does modify the band structure of the as-prepared mixed-phase flms, and the amplitude is observable. To confrm these observations, we performed ultravioletvisible absorption and ultraviolet photoelectron spectroscopy (UPS) measurements on the as-prepared flm with 1.2 M-FAI in comparison with that of the pure a and d phase. For the flms of pure a-FAPbI 3 and d-FAPbI 3 , the absorption onset is located around 840 nm and 510 nm (Fig. S9b †), corresponding to an optical band gap energy (E g ) of 1.47 and 2.43 eV, respectively, which is in good agreement with the literature reported. 24 For the 1.2 M-FAI flm, the absorption onset has a blue shift of 20 nm and lies at 820 nm, indicating a slightly larger E g (1.51 eV) compared to that of pure a-FAPbI 3 . The valence band maximum (VBM) of the flms was then determined by UPS measurements. As shown in Fig. S10, † the VBM is determined to be 5.74, 5.12, and 5.46 eV for a-FAPbI 3 , d-FAPbI 3 and the 1.2 M-FAI flm, respectively. The schematic energy level diagram is shown in Fig. 3b, which clearly demonstrates that both the conduction band minimum (CBM) and VBM of the 1.2 M-FAI flm are just between those of a-FAPbI 3 and d-FAPbI 3 . It was reported that band bending would occur in the phase junction through Fermi level lining up when they are in contact with each other. 37 Accordingly, the band edge of the a/d phase junction could be attributed to the valence band upward bending of a-FAPbI 3 and the conduction band downward bending of d-FAPbI 3 in their junction area. Such an a/d phase junction exists in all of the flms from 1.1 M-FAI to 1.5 M-FAI, so these flms should have the same bandgap, which is in good agreement with their similar PL spectra. For the perovskite flms, the high PL intensity is strongly related to the high exciton binding energy (E B ) because the thermal ionization of excitons with a low E B could induce PL quenching, and also involve the decrease of grain size, hence spatially confning the excitons. 11,38,39 The exciton binding energy, E B , is a modifed Rydberg energy given by a static solution to the Wannier equation: 40 Herein, m is the exciton effective mass, m 0 the electron mass, e the unit charge, 3 the dielectric constant and 3 0 the vacuum permittivity. To understand the greatly enhanced PL from the 1.2 M-FAI flm with the a/d phase junction, we then measured the dielectric constant of the flm in comparison to the flms with the pure a and d phase (Fig. S11 †). The dielectric constant of the 1.2 M-FAI flm was measured to be 4.8, much lower than those of both the pure a phase (6.6) and pure d phase (6.9). Such a markedly reduced dielectric constant can be roughly described by Lichtenecker's model for the dielectric function of a two-phase composite according to a serial mixing rule, 41 in which two phases alternate or randomly connect in parallel as evidenced by the HRTEM of the 1.2 M-FAI flm. Since the exciton binding energy is proportional to 1/3 2 as eqn (1) states, if we roughly estimated the same exciton effective mass, the exciton binding energy of the 1.2 M-FAI flm might be nearly two-fold higher than that of the pure a-FAPbI 3 and d-FAPbI 3 . In accordance with this, the reduced dielectric constant would help increase the exciton binding energy, thus avoiding luminescence quenching. 11 In the meantime, scanning electron microscopy (SEM) was used to measure the average grain size of these flms. As shown in Fig. 4, the average grain sizes of the pure a-FAPbI 3 and d-FAPbI 3 are 139 and 261 nm, respectively, whereas that of the 1.2 M-FAI flm is greatly reduced to 64 nm, only one third or one fourth of that of the pure phases (Fig. S12 †). This substantial reduction of the average grain size, on the one hand, can lead to a reduction of coherence length of the ferroelectric coupling between local dipoles and a diminished macroscopic polarization, thus causing a reduction in the dielectric constant, 42 on the other hand, it can spatially confne the excitons, thus increasing the radiative recombination in the nanograins. 11 In addition, the volume fraction of a-FAPbI 3 in the 1.2 M-FAI flm is considerably small due to low annealing temperature, which can help reduce the re-absorption by a-FAPbI 3 , and thereby likewise contribute to high PL intensity. To reveal the role of the precursor stoichiometry in the formation of such unique a/d phase junctions, we then investigated the colloidal properties of these precursor solutions. The absorbance edge of all of these precursor solutions demonstrated a gradual red shift along with the precursor varying from 1 M-FAI to 3 M-FAI (Fig. S13a †). Meanwhile, dynamic light scattering tests were conducted to characterize the colloidal size distribution associated with the formed lead polyiodide compounds in different precursor solutions (Fig. S13b †). Results with a shrinking variation trend came out showing a gradually decreased colloidal size when increasing the molar ratio of FAI/ PbI 2 . The average size of the colloid in the 1 M-FAI precursor solution was estimated to be $1262 nm, while the value of the 1.2 M-FAI precursor solution remarkably decreased to $354 nm. This trend revealed that excess FAI in the precursor solution can bring about more coordination effects and thus reduce the formation of large colloids, 31 which was consistent with the above observed small grains for the 1.2 M-FAI flm. Generally, d-FAPbI 3 is found to be more thermodynamically favorable and stable compared to a-FAPbI 3 at low temperature due to a lower formation energy. 43 However, when the size of the colloid particles decreases to sufficiently low values, the total free energy (including surface and bulk) of d-FAPbI 3 may be equal to or even higher than that of a-FAPbI 3 and the relative phase stability may reverse, thus facilitating the formation of a-FAPbI 3 even at a low annealing temperature and further allowing the formation of an a/d phase junction. 44,45 On the basis of these results, we proposed a possible mechanism concerning the perovskite crystal growth with excess FAI in the precursor solution (as shown in Fig. 5). Therefore, even when annealed at a temperature as low as 60 C, the FAPbI 3 flms with an appropriate excess of FAI could form the a and d phase synchronously, which results in the formation of an a/d phase junction. Finally, we investigated the preliminary stability of the asprepared FAPbI 3 perovskite flms with an a/d phase junction in comparison with that of the pure a-FAPbI 3 . All the flms were stored in an ambient environment without encapsulation (at 25 C and relative humidity <50%). Unexpectedly, even after 15 days, the XRD pattern of the 1.2 M-FAI flm with the a/d phase junction remained unchanged without showing any observable decomposition or phase transition (Fig. 6a). In contrast, for the as-prepared flm with pure a-FAPbI 3 , the XRD pattern (Fig. 6b) showed the peak at 12.6 corresponding to PbI 2 , indicating that FAPbI 3 does decompose when annealed at high temperature. 21,30 Within only 24 hours, the peak at 11.8 belonging to d-FAPbI 3 appeared, indicating that part of the a phase turned into the d phase. After 15 days, a-FAPbI 3 had fully turned into d-FAPbI 3 . Therefore, these results clearly demonstrated that the obtained flms with the a/d phase junction showed long-term stability against humidity. As evidenced by the XRD patterns in Fig. 1, we can roughly estimate that there is only a small volume fraction of a-FAPbI 3 in the 1.2 M-FAI flm. Considering the thermodynamics, such superior air stability could be mainly attributed to the majority of the flm being thermodynamically stable d-FAPbI 3 at low temperature and concomitant stabilization from the mixing entropy of the a/d phase. 28 In addition, such stabilization can be rationalized in terms of miscibility of the a/d phase due to almost the same volume per stoichiometric unit of the two crystals ($256 3 ). 24,28 To gain more insight into the role of the a/d phase junction in air stability, further investigation through in situ characterization and theoretical calculation is required. ## Conclusions In summary, we have realized the frst pure FAPbI 3 perovskite flms with controllable a/d phase junctions by using stoichiometrically modifed precursors, which demonstrated desirable humidity stability and signifcantly enhanced NIR emission. The enhanced NIR emission was found to be related to the unique a/d phase junction and the reduced grain size. Interestingly, such a/d phase junctions can be readily obtained at low annealing temperature (60 C) and demonstrated long-term phase stability against humidity. The stable and controllable phase-junction presented herein would pave a new way for developing highly stable and efficient perovskite light emitting materials and devices. We are currently carrying out studies on the perovskite light-emitting diodes with this aim.

chemsum

{"title": "Stable \u03b1/\u03b4 phase junction of formamidinium lead iodide perovskites for enhanced near-infrared emission", "journal": "Royal Society of Chemistry (RSC)"}

stability_of_radical-functionalized_gold_surfaces_by_self-assembly_and_on-surface_chemistry

## Abstract: We have investigated the radical functionalization of gold surfaces with a derivative of the perchlorotriphenylmethyl (PTM) radical using two methods: by chemisorption from the radical solution and by on-surface chemical derivation from a precursor. We have investigated the obtained selfassembled monolayers by photon-energy dependent X-ray photoelectron spectroscopy. Our results show that the molecules were successfully anchored on the surfaces. We have used a robust method that can be applied to a variety of materials to assess the stability of the functionalized interface. The monolayers are characterized by air and X-ray beam stability unprecedented for films of organic radicals.Over very long X-ray beam exposure we observed a dynamic nature of the radical-Au complex. The results clearly indicate that (mono)layers of PTM radical derivatives have the necessary stability to withstand device applications. ## Introduction Molecular systems are materials that intersect with many different promising felds such as organic/molecular spintronics, electronics, and organic magnetism. In this framework, organic radicals are exceptionally promising in various felds, and the research on radical thin flms and interfaces has recently flourished, due to their potential use in applications from quantum computing to organic electronics and spintronics. We have recently demonstrated that a Blatter radical derivative is a potential quantum bit and we attached it to copper contacts to investigate the influence of a substrate on the radical magnetic moment. 9 Our work indicated the need for identifying strategies in order to attach the radical to the surface preserving its magnetic moment at the interface by using different methods ranging from evaporation to preparation in a wet environment. However, the radical functionalization of a substrate is eased by choosing a specifc chemical group that has a high chemical affinity for the selected substrate. Usually thiols and disulfdes are chosen to covalently modify gold surfaces, including gold nanoparticles, with organic radicals by adsorption from solution. More recently, alkyne terminated derivatives have started to play a role. Nitroxides (TEMPO), nitronyl nitroxides and tripheylmethyl radicals have been successfully employed to prepare such paramagnetic hybrid materials. In this work, we capitalize our knowledge of radical thin flms and interfaces by studying the functionalization of gold surfaces with derivatives of the perchlorotriphenylmethyl (PTM) radical. PTM is a very persistent and stable radical that shows a long coherence time at room temperature, being a strong potential candidate for quantum technologies. 24 Previously, self-assembled monolayers (SAMs) of PTM on gold substrates have been investigated to study their transport properties. 25 The radical character of the layers was proved by several techniques (UV-vis, cyclic voltammetry, EPR, NEXAFS and UPS); however, a careful and in-depth characterization of the stability of these radical SAMs has not been carried out so far. Such a stability is a necessary precondition to use radical-based SAMs for any practical application. Here, we used a ferrocene functionalized PTM derivative with an alkyne termination (Fig. 1) that covalently attaches to a gold substrate spontaneously. The ferrocene functionalization makes the molecules interesting for current rectifcation, as seen in SAMs incorporating ferrocene acting as a redox-active moiety. We investigated also the formation of radical self-assembled monolayers (SAMs) obtained by using on-surface chemistry. Our investigations were performed using X-ray photoelectron spectroscopy (XPS). While XPS is a well-established technique to investigate the electronic structure of materials, this is not the only aspect that can be examined. 35 Because of its high sensitivity, it is also possible to quantitatively calculate the stoichiometry of the investigated systems. Further aspects can be explored: it is very sensitive to the chemical environment of the elements, allowing the occurring chemical bonds and the charge transfer from/to surfaces to be revealed. It is possible to gain information on flm stability (e.g., under X-ray beam or air exposure) and on post-growth phenomena. It is extremely well suited to investigate radical thin flms (including their radical character) when evaporated by using controlled conditions. 36 We proved in our previous work that XPS in combination with a careful and robust best ft procedure allows investigation of the radical character, with the results being in perfect agreement with electron paramagnetic resonance (EPR) measurements. 9,36-41 EPR is the technique typically used for radical characterization. However, its use for flms is limited (1) by the fact that it is an ex situ technique. Radical thin flms might not be stable enough outside the ultra-high vacuum environment where they are deposited or obtained by onsurface reaction. (2) By the choice of the substrate that might contribute to the EPR signal. 41 (3) By the substrate dimensions that are often over-dimensioned for standard spectrometers. (4) By the fact that standard EPR spectrometers do not have the necessary sensitivity to measure (sub)monolayers. Conversely, XPS has a high sensitivity further beyond many other conventional chemical techniques, as it can detect less than 10 13 atoms, 45 allowing investigations in the monolayer and submonolayer regime without requiring advanced "stateof-the-art" spectrometers, as it is the case for EPR, but a standard, commercially available, monochromatized laboratory XPS station is sufficient. In this work, we investigate the chemistry of the SAM/gold interface, demonstrating that the SAMs were successfully attached to the substrate, using also on-surface chemistry. We also show that it is possible to identify the spectroscopic lines associated with the radical character versus its diamagnetic counterpart. The work focuses on the SAM stability, under X-ray and air exposure, using a method that can be applied to any material to explore any kind of stability issue, such as gas exposure, humidity, aging, and temperature that are of paramount importance for technological applications. ## Experimental section SAM1 and SAM2 were prepared following the protocol thoroughly described in ref. 34. SAM4 was grown following a twostep reaction: (1) SAM1 was immersed in a 2 mM solution of Bu 4 NOH/THF (freshly distilled) under an argon atmosphere. The solution was left with a gentle stirring for 8 h at room temperature in the dark. Then, the substrates were removed from the flask and thoroughly rinsed with THF (distilled). ( 2) Immediately afterwards, the substrates were immersed in a 4 mM p-chloranil/THF (distilled) solution under an argon atmosphere. The solution was left for 12 h at room temperature in the dark. Finally, the substrates were removed from the flask, thoroughly rinsed with THF (distilled) and dried with a nitrogen stream. Coverage and radical formation were checked with cyclic voltammetry. An XPS Ultra High Vacuum (UHV) system (2 10 10 mbar base pressure) equipped with a monochromatic Al Ka source (SPECS Focus 500) and a SPECS Phoibos 150 hemispherical electron analyzer was used. Survey spectra were measured at 50 eV pass energy and individual core level spectra at 20 eV pass energy. Both were subsequently calibrated to the Au 4f signal at 84 eV. To minimize potential radiation damage, freshly prepared flms were measured, and radiation exposure was minimized unless differently stated in the text (i.e., stability measurements). For measurements probing air stability, X-ray beam exposure was further limited after air exposure to attribute the observed changes exclusively to the degradation by air exposure. Photon-energy dependent XPS measurements were performed at the third-generation synchrotron radiation source BESSY II (Berlin, Germany) at the Low-Dose PES end station installed at the PM4 beamline (E/DE ¼ 6000 at 400 eV). They were carried out in multibunch hybrid mode with a SCIENTA ArTOF electron energy analyzer (ring current in top up mode ¼ 300 mA). ## Results and discussion We examined two different layer preparations using the PTM radical derivative (SAM2 and SAM4) and we compared them with those obtained by depositing the diamagnetic counterpart, SAM1 (Fig. 1). The PTM radical and the diamagnetic derivative shown in Fig. 1 were synthesized as previously reported: 34 SAM2 is obtained by depositing the radical on a gold substrate from its solution. SAM4, in contrast, is obtained by frst depositing the analogous diamagnetic molecules on gold and following a two-step synthesis (i.e., anion generation and oxidation), and thus, the PTM radical is formed on the surface. 46 Fig. 2 shows the SAM2 XPS spectra of the important core levels (for the survey and the stoichiometric analysis, see Fig. S1, Tables S1 and S2 in the ESI †). The spectra are characterized by the predominance of gold signals in agreement with the deposition of a monolayer. Apart from a carbon concentration that slightly exceeds the theoretical values, which is usual in samples prepared ex situ with wet-environment techniques, the flms are remarkably clean, and no signifcant amounts of contaminants are visible. In XPS, the integrated area of the main lines corresponding to photoelectrons emitted from a given element, together with their satellites, is proportional to the concentration of that same element in the investigated system. 35,47,48 In highly resolved XPS spectra, the rich fne structure allows ftting the lines including contributions from different atomic sites of the same element which, due to a different chemical environment, are expected to show differences in their binding energies. 35,47,48 The flm stoichiometry agrees with the expected values, confrming that the radical derivative was indeed attached to the gold substrate. The C 1s spectroscopic line is characterized by a main peak at around 284.5 eV and a feature at around 286 eV. The C 1s intensity is due to photoelectrons emitted from the carbon atoms. The contributions mirror several different chemical environments. In fact, carbon atoms are not only bound to other carbon atoms, but to hydrogen, nitrogen, and chlorine atoms. Each different environment leads to a slightly different binding energy that can be identifed by using a best ft procedure (Fig. 2a). 34,37,49 The ftting procedure in XPS is driven by specifc and detailed chemical and physical arguments, and not by a mere mathematical approach. The curves are described using a Voigt profle, i.e., a convolution of a Gaussian and a Lorentzian profle. This is because different contributions influence the line shape of the XPS main features: intrinsic lifetime broadening, vibronic and inhomogeneous broadening lead to a Lorentzian profle, while experimental contributions have a Gaussian profle. The lifetime of the core hole is determined basically using the Heisenberg uncertainty principle and consequently the intrinsic peak width is determined, too. For example, the Lorentzian width for the C 1s orbital is around 80 meV, and for the N 1s orbital it is around 100 meV in organic materials. 50 The experimental setup gives a contribution assumed to have a Gaussian lineshape due to the resolution of the analyser, the non-perfect monochromaticity of the X-rays, and inhomogeneities of different nature. The ft that we use is based on the procedure adopted for closed-shell molecules. 14 The fnal ft is the result of several self-consistent iteractions of sequential fts performed considering all physical and chemical information and adding more constraints at each iteraction, with the goal of keeping the parameter dependency very low (dependency values of the last fts in this work were very close to zero). The constraints in our ft are based on the element concentration, and the binding energy constraints must adhere to electronegativity and known values in the literature, so we use the published or measured core-hole lifetimes for each element. The ft procedure must systematically hold for all samples of a specifc system, prepared and measured under the same conditions. Our procedure revealed to be extremely robust giving results in very good agreement with EPR and ab initio calculations, both for open-shell and closed-shell systems, as well. 9, To reach this result, we work on sets of samples that are large enough to be statistically signifcant. In this way, we can also identify the samples that do not correspond to the expected stoichiometry. 37,39,56,57 In the spectra, we observe the presence of shake-up satellite intensities (Fig. 2). As a result of the core-hole formation, the symmetry is reduced, and a larger number of non-equivalent carbon atoms should be considered. 58,59 The ionization at different carbon sites may give different contributions to the shake-up spectra. The S 1 satellite can be related to the frst HOMO-LUMO shake-up. 60 Its energy position with respect to the main line is lower than the optical gap, a typical effect in the HOMO-LUMO shake-up satellites of polyaromatic molecules caused by the enhanced screening of the core-hole due to its delocalization. A large number of satellite features is expected upon a photoemission event. However, their assignment is very complicated, especially for large molecules because they are not completely described by theoretical models. Such a detailed description is outside the goal of this work; therefore, we have identifed most of the higher binding energy satellite intensities under a single component, S 2 . This component is correlated with the C-Cl feature from a stoichiometric point of view. This assignment is further corroborated by the fact that the C-Cl feature and S 2 change simultaneously depending on the photon energy, as it can be easily seen in Fig. 2 and 3. The intensities of the various contributions agree with the expected stoichiometry, confrming once more that the SAM2 carbon line corresponds to the radical derivative (Table S2 in the ESI †). The Cl 2p, N 1s and Fe 2p core level spectra are also shown. Their features confrm the presence of an intact molecule: Cl 2p core level lines show the typical doublet feature (spin-orbit splitting ¼ 1.6 eV, as in the literature 64 ), and the N 1s spectrum (Fig. 2c) is characterized by contributions due to photoelectrons emitted from three different chemical environments, confrming the intactness of the triazole derivative (Fig. 2c and d). The signal of the Fe 2p shows the expected doublet (spin-orbit splitting ¼ 12.8 eV, concomitant with the values in the literature 64,65 ), and the noteworthy absence of further intensities indicates that the signal is due to electrons emitted from iron atoms in the +2 oxidation state, as it is the case for ferrocene. Note that, for the monolayers of clean ferrocene, the Fe 2p spectrum does not show any satellite intensity. In fact, this intensity depends on the ligands and it varies with their electronegativity. 70 Additionally, in the case of monolayers on metal substrates, the image-charge formed at the interface 71 may further screen the satellite intensities. The XPS intensities and line shapes indicate that the radical was attached to the surface preserving the expected stoichiometry. Thus, we can confdently infer that the synthesis and the preparation of SAM2 were successful. To support this conclusion and explore the use of XPS to identify the PTM radical, we investigated SAM1, i.e., the SAM obtained from the diamagnetic counterpart of the PTM radical derivative (Fig. 1). The essential core level spectra are shown in Fig. 3 (for the survey and the stoichiometric analysis, see Fig. S2, Tables S3 and S4 in the ESI †). In our discussion, we focus on the C 1s core level spectroscopic line. This is the line that is directly correlated with the radical character (see Fig. 1) because the unpaired electron mainly resides in the central radical carbon atom of the perchlorinated triphenylmethyl unit. The stoichiometry for SAM2 and SAM1 is different. In SAM1 the central methyl carbon atom of the PTM is bound to hydrogen. Therefore, we expect a different C 1s line broadening with respect to the radical spectra. Indeed, we observe a larger line for SAM1 (Full Width at Half Maximum (FWHM) ¼ 1.8 eV versus 1.4 eV for SAM2, under the same experimental conditions). This difference is mirrored by a larger Gaussian width required in the ft procedure (see Table S4 in the ESI †). We also observe a different binding energy. The SAM1 C 1s main line is at higher binding energy than the SAM2 main line. This indicates that the core-hole created upon photoemission is more efficiently screened in SAM2 than in the diamagnetic molecule. This can be explained considering the donor-acceptor character of SAM2 (ref. 34) where the simultaneous presence of the radical and the azidomethyl ferrocene unit stands for faster charge delocalization of the core-hole. These differences in the C 1s main line, binding energy and broadening between SAM1 and SAM2 allow using XPS to identify the radical character of the SAMs. In an XPS experiment it is possible to probe different sampling depths: 72 when changing the photon energy, the materials emit electrons with different kinetic energy which is equivalent to emitting photoelectrons with different inelastic mean free path (l). Thus, we performed a photon-energydependent experiment on SAM1, SAM2 and SAM4 using 460 and 640 eV photon energy, respectively. This corresponds to varying l between 0.17 and 0.28 nm (ref. 73 and 74) (Fig. 4 and S3 †). The experiment at 460 eV is very surface-sensitive (note that both experiments at 460 and 640 eV are very surfacesensitive with respect to the measurements so far discussed, which were performed at 1486.6 eV). We observe that, by varying the photon energy, the relative intensities of the main line and the line due to photoelectrons emitted from carbon atoms bound to the electronegative nitrogen and chlorine atoms change: the feature at higher binding energy has higher intensity at 640 eV. What is also important is that these changes are accompanied by changes in the S 2 satellite, indicating, as mentioned, that these two components are strongly correlated, corroborating our ft assignments. This change in the intensity depends on the photon energy and, thus, on the inelastic mean free path, and it is due to the surface core level shift effect, i.e., the difference of the core level photoemission between a surface atom/molecule and a bulk atom/molecule. 76,78,79 This effect is visible in organic thin flms when the molecules are not planar and carry electronegative atoms. 49,80,81 In fact, electronegative atoms shift the electronic cloud, causing a different screening of the core-hole created upon photoemission. However, this screening is different when it occurs at different depths where structural differences are signifcant, for example, in the case of upright versus flat lying molecules. 80 In the present case, the C-Cl components are stronger at 640 eV when the experiment is less surface-sensitive. We can infer structural information from this dependence: the XPS results indicate that the PTM radical is closer to the substrate with respect to the azidomethyl-ferrocene unit (as sketched in Fig. 1); therefore, its contribution is stronger when l is longer. For the photon energy of 1486.6 eV, l is comparable with the dimensions of the molecule (l ¼ 0.81 nm (ref. 73)), in which case the stoichiometry information plays the major role against the structural information, as seen in closed-shell systems like phthalocyanines. 82 Using the above results as a reference, we investigated SAM4 (Fig. 4, lower panel). This monolayer has the same theoretical stoichiometry as SAM2, but it has been obtained via on-surface radical formation from the diamagnetic molecule. We focused once more on the C 1s core level spectra. First, from the point of view of the stoichiometry, as previously performed for SAM2 and SAM1, we observe that the C 1s line shape has the same features as in the SAM2 core spectra. In this case also, we observe the same photon energy dependence at 460 and 640 eV, hinting at a similar structural adjustment of the molecule units with respect to the substrate. What is most important is that the FWHM of the C 1s line is narrower than in the case of the diamagnetic molecule, i.e., SAM4 has a narrower main line than SAM1 (see Fig. S3 †). Following our above discussion, this effect indicates a radical character of the flm. Since the radical generation occurred on the surface, this result hints at and supports the successful on-surface preparation of the radical. A ft procedure backs these observations: the same best ft procedure leads to the same intensities and binding energies for the C 1s contributions of the spectra of SAM2 and SAM4 (Fig. 4 and Tables S5-S8 in the ESI †). Cyclic voltammetry experiments support the radical character of the layers, too (see Fig. S4 in the ESI †). The redox peaks corresponding to the PTM radical 4 PTM anion and ferrocene 4 ferrocenium redox process are clearly observed. A change in photon energy as performed in the present XPS experiments also implies a change in the C 1s cross-section increasing the complexity of the screening effects. Looking at the ft results, we note that the S 1 intensity decreases with increasing the photon energy while the intensities of the S 2 satellite show the opposite behaviour (Tables S5-S8 in the ESI †). This gives a hint about the fact that the S 1 intensity is related to the dipole excitation of a core electron to the lowest unoccupied molecular orbital (LUMO) accompanied by the monopole ionization of the valence electron: this shake up contribution is near the ionization threshold region and decreases with the increase in energy, 83,84 as observed in our fts. An important aspect that we intend to address here is the stability of the monolayer in the real environment. While the PTM radical is known to be chemically stable both in solution and in powder if visible light is avoided, there is no report on the chemical and structural stability of its flms where single radical molecules are exposed to air. To tackle this issue, we kept SAM2 monolayers under air in darkness and measured them again 128 days later, always minimizing X-ray exposure during measurements. The results are shown in Fig. 5. The C 1s core level spectrum comparison between the fresh monolayer and the "aged" monolayer shows a small difference in the relative intensity of the main feature with respect to the feature at higher binding energy, while the Cl 2p spectra do not show major differences. Post-growth phenomena, such as desorption and ripening, are expected and well-known in the case of organic molecules, and expected also in radical flms, especially for those systems having low vapor pressure at room temperature and physisorbed on surfaces. 36,37,40 To investigate the origin of the difference in the C 1s core level spectra we performed a best ft analysis, following two hypotheses. In one case, we performed the ft considering that PTM might switch to the perchlorophenylfluorenyl radical (PPF) (Fig. S5 and Table S9 in the ESI †). This is a known derivative of the PTM radical generated both by heating over 300 C (ref. 85) and by photoirradiation. 86 In the second case, we considered that the stoichiometry of the monolayer stays unchanged but the carbon intensity increases due to the adsorption of carbon impurities from the environment (Fig. S6 and Table S10 in the ESI †). Both fts are plausible. A closer inspection of the survey spectra helps to interpret the results (Fig. 5, lower panel). Initially, the gold signal is stronger, i.e., its intensity decreases with time. Simultaneously the carbon signal increases, while the chlorine signal does not change. From the stoichiometric analysis of the spectra, we found that in the fresh monolayer the carbon to gold ratio (C/Au) and the chlorine to gold ratio (Cl/Au) are 0.37 and 0.04, respectively. After 128 days, they are 0.40 and 0.04, respectively. This clearly indicates that the chlorine content does not diminish and that the phenomenon playing the major role is carbon adsorption. This means that not only the PTM radical is chemically stable, but also its monolayers are stable under prolonged air exposure. This is a result of great signifcance because it fully supports the use in devices of the PTM radical and its derivatives grafted on surfaces. We also studied the stability of SAM2 against X-rays. As previously, we focus our discussion on the PTM radical analysing the C 1s and the Cl 2p core level spectra (Fig. 6). We could observe frst small changes in the spectroscopic lines after 18 hours of X-ray exposure, a 0.1 eV shift of the binding energy towards higher values and a difference in the satellite intensities. The ft analysis performed on the C 1s line confrms that these are not signifcant stoichiometric changes (Fig. S7, Tables S11 and S12 in the ESI †). We crosschecked this fnding also using synchrotron radiation and monitoring the flm in realtime over around 8 hours (Fig. 6, lower panel, photon energy: 640 eV, flux: 1 10 9 to 1 10 10 photons per s). No changes were detected. To understand what happens under very long X-ray exposure, we exposed the flms to X-rays for 52 hours and we looked at the effects (Fig. 6d). After such a long exposure, the gold signal is more intense, while the C 1s and Cl 2p lines show no decrease in the intensity. This indicates that the gold substrate is more exposed with time. Usually this result hints at changes in the flm morphology due to post-growth phenomena, such as desorption, dewetting or Ostwald ripening, which lead to the coalescence of small islands into big islands leaving a larger area of the substrate surface free. The result indicates, also in this case, some degree of dynamics, suggesting a change in the layer morphology. These experimental observations seem puzzling in the case of a strong adsorbate-substrate chemical bond. To help in understanding this phenomenon, we can look at one of the most investigated SAM systems: thiolates on gold. Investigations of thiolate-Au surfaces have demonstrated a clear dynamic nature of these surfaces, where the mobility of the adsorbate-Au complex plays an important role, both on flat surfaces as well as on nanoparticles, upon mild annealing and even at room temperature. 87 The mobility is explained in terms of the presence of defects on gold surfaces. At defect sites, the interaction between a single gold atom and a covalently attached molecule is stronger than the interaction with the environment (gold atoms and surrounding molecules, respectively) causing the motion of the complete adsorbate + Au assembly on the surface, giving rise to ripening, and even to desorption. This is a very general mechanism of surface diffusion occurring when an adsorbate is strongly bound to coinage metals such as gold. 87,89 The behaviour of the PTM-based SAMs on gold and the resulting XPS spectra observed during prolonged X-ray beam exposure would hint at the fact that such a mechanism also occurs in the present case, favoured or induced by the prolonged X-ray exposure. ## Conclusions Once more, XPS has proved to be a very powerful tool to investigate radical flms and radical/metal interfaces, uncovering phenomena not yet known. Furthermore, our XPS method to assess the stability of radical/inorganic interfaces can be applied to any system. In this work, we have investigated the stability of chemically functionalized gold surfaces with a PTM radical, either by preparing the self-assembled monolayers directly from the radical solution or, alternately, by chemical means obtaining the radical on the surface from its diamagnetic precursor. While the chemical stability of the PTM radical is well-known (PTM is considered an inert radical) here we show that the radical monolayers have unprecedented stability under ambient conditions and aggressive X-ray exposure. Extremely prolonged X-ray exposure indicates a dynamic nature of the radical-Au complex, analogously to the case of thiolate-Au surfaces. To our knowledge, this phenomenon has not yet been reported for this class of adsorbate-Au systems. Therefore, further investigations, including annealing experiments and theoretical modelling, are necessary to deepen the understanding of the dynamical aspects of this surface. We cannot exclude that similar phenomena might occur at room temperature also upon prolonged air exposure, with a reaction time of weeks, as seen for thiolate-Au nanoparticles. 87 Although further investigations on the long-term aging pattern of the PTM radical-based layers also depending on different parameters, such as temperature and visible light, are necessary, our results point out that carbon absorption from the ambient environment plays the major role when the monolayer is exposed to air for a long time. The PTM radical and its derivatives form monolayers that have unprecedented stability properties, confrming that these systems are suitable candidates for market-oriented applications.

chemsum

{"title": "Stability of radical-functionalized gold surfaces by self-assembly and on-surface chemistry", "journal": "Royal Society of Chemistry (RSC)"}

synthesis_of_α-amino_amidines_through_molecular_iodine-catalyzed_three-component_coupling_of_isocyan

## Abstract: A facile and efficient synthetic protocol for the synthesis of α-amino amidines has been developed using a molecular iodinecatalyzed three-component coupling reaction of isocyanides, amines, and aldehydes. The presented strategy offers the advantages of mild reaction conditions, low environmental impact, clean and simple methodology, high atom economy, wide substrate scope and high yields. ## Introduction Amidines are a class of organic compounds exhibiting a variety of biological activity that makes them potential candidates for drug development and discovery . Simple amidines are generally synthesized from their corresponding nitriles either by the Pinner reaction or by the thioimidate route . Recently, much attention was given to the development of new routes for the synthesis of substituted amidines . Even if these methods provide amidines in acceptable yields, they suffer from limitations such as limited structural diversity of the final products. Since multicomponent reactions (MCRs) are expected to provide a rich structural diversity, much attention was paid on the development of multicomponent-coupling strategies for the synthesis of amidines. The Ugi reaction is probably one of the best multicomponent reactions to provide huge structural diversification of the products . Thus, several modifications of the Ugi reaction were explored recently. As depicted in Figure 1, diamides, α-amino amides, and α-amino amidines can be obtained depending on the nucleophile used. However, the reaction does not lead to acceptable product yields of products without using proper catalysts except when the nucleophile is carboxylate. For instance, Figure 1: Synthesis of diamides, α-amino amides and α-amino amidines through Ugi and related MCRs. among various catalysts screened, only phosphinic acid and boric acid were found suitable for conversion of substrates into products when water was used as a nucleophile for amide preparations . In the direction of amidine synthesis using isocyanide MCRs, a few catalysts such as p-toluenesulfinic acid , metal triflates , bromodimethylsulfonium bromide , ZnO nanoparticles and BF 3 •OEt 2 were reported with varying degrees of success. All these reported methods for the preparation of α-amino amidines have their own limitations such as long reaction times, high catalyst loading and use of expensive and hazardous metal catalysts. Therefore, the development of a mild, inexpensive and efficient catalytic protocol for the amidine synthesis is highly needed. Iodine is expected to act as a Lewis acid or Brønsted acid in methanol . Apart from oxidation, catalytic iodine provides mild and efficient ways for the formation of C-C and C-N bonds . As a part of our ongoing interest towards the synthesis of new molecular libraries , we were interested in developing a one-pot MCR strategy for the synthesis of amidines. ## Results and Discussion To check the feasibility of the iodine-catalyzed amidine synthesis through the modified Ugi reaction, we carried out a model reaction of tert-butyl isocyanide (1 mmol), benzaldehyde (1 mmol), and aniline (2 mmol) using 5 mol % of molecular iodine in methanol (Table 1). The reaction worked well at ambient temperature and led to good yields of 4a. Among various solvents screened, methanol was found to be the best choice as solvent for the reaction. Furthermore, we observed that the catalyst loading could be reduced to 1 mol % without affecting the product yield. Further decreasing the amount of catalyst (0.5 mol %) still lead to a good yield of 4a, albeit with a longer reaction time (Table 1, entry 5). It was interesting to notice a significant decrease in the product yield when the catalyst was overloaded (Table 1, entries 6 and 7). When the reaction was carried out without catalyst (iodine), no product was observed (Table 1, entry 11). This observation confirmed that catalytic iodine is necessary for the success of the reaction. Next we studied the substrate compatibility of the reaction to generalize the scope of the α-amino amidine synthesis (Table 2). Aliphatic, aromatic and heteroaromatic aldehydes were used with similar success leading to high yields of amidines. It is worth to note here that aldehydes containing an alkyne moiety yielded the corresponding amidines with similar success (Table 2, entries 17 and 18). With aromatic amines, the reaction was good; with aliphatic amines (for instance benzylamine) the reaction was sluggish and the desired amidine was not obtained. The reaction worked well with a variety of isocyanides such as tert-butyl isocyanide, cyclohexyl isocyanide, and more importantly with functional groups bearing isocyanides such as ethyl isocyanoacetate and p-toluenesulfonylmethyl isocyanide (p-TosMIC) (Table 2, entries 19 and 20). Thus, the diversification of the α-amino amidine was achieved by varying the aldehyde, aromatic amine and isocyanide components of the reaction. The iodinecatalyzed protocol gave better yields (75-93%) of amidines than a recently reported p-toluenesulfinic acid catalyzed protocol (52-71%). In contrary to the p-toluenesulfinic acidcatalyzed protocol, the formation of byproducts (α-amino amides) was suppressed in our iodine-catalyzed protocol which gave rise to better yields and cleaner products. When compared with other related reports , our iodine-catalyzed protocol gave similar yields of α-amino amidines. However, it should be emphasized that our protocol with low catalyst loading (1 mol %) makes it a cleaner and lower environmental impact methodology to access α-amino amidines. Then, we tried the three-component reaction using heteroaromatic amines such as 2-aminopyridine, 3-aminopyridine and 4-aminopyridine. The desired products (amidines) were not obtained with 3-aminopyridine and 4-aminopyridine. However, we found that iodine can efficiently catalyze the three-component coupling reaction of 2-aminopyridine, aldehyde and isocyanide (Groebke-Blackburn-Bienaymé reaction) (Figure 2) . Recently, catalytic iodine (10 mol %) was found to give good yields of imidazolopyridine in a three-component reaction of 2-amino-5-chloropyridine, isocyanide, and aldehydes under reflux conditions . However, we found that the similar reaction using 2-amonopyridine could be performed at ambient temperature using 1 mol % of iodine as catalyst to achieve a satisfactory yield of product (82-85%). A probable mechanistic pathway for the formation of α-amino amidines is outlined in Figure 3 which is analogous to the established mechanism reported in the literature . Iodine can serve as a catalyst for the activation of imine. The attack of nucleophilic isocyanide on the activated imine leads to the formation of intermediate 8 or 8'. Subsequently, another molecule of amine attacks the intermediate 8 or 8' to give α-amino amidine 9 which undergoes further -hydrogen shift to provide the α-amino amidines 4 . ## Conclusion In conclusion, we have developed a simple and clean methodology for the synthesis of substituted α-amino amidines using a three-component coupling of isocyanide, aldehyde, and aromatic amines with molecular iodine as a catalyst. The current strategy provides elegant access to α-amino amidine and imidazolopyridines in high yield with significantly low catalyst loading. ## Experimental A 25 mL round bottom flask was filled with aldehyde (1 mmol), amine (2 mmol)/2-aminopyridine (1 mmol), isocyanide (1 mmol) and MeOH (5 mL). Then, I 2 (1 mol %) was added and the reaction mixture was stirred until the reaction was completed (TLC). The reaction mixture was evaporated to dryness using a rotary evaporator and the residue was purified by silica-gel column chromatography using a mixture of ethyl acetate/hexane as eluent in increasing polarity.

chemsum

{"title": "Synthesis of \u03b1-amino amidines through molecular iodine-catalyzed three-component coupling of isocyanides, aldehydes and amines", "journal": "Beilstein"}

peptide-based_capsules_with_chirality-controlled_functionalized_interiors_–_rational_design_and_ampl

## Abstract: Peptides are commonly perceived as inapplicable components for construction of porous structures. Due to their flexibility the design is difficult and shape persistence of such putative structures is diminished.Notwithstanding these limitations, the advantages of peptides as building blocks are numerous: they are functional and functionalizable, widely available, diverse and biocompatible. We aimed at the construction of discrete porous structures that exploit the inherent functionality of peptides by an approach that is inspired by nature: structural pockets are defined by the backbones of peptides while functionality is introduced by their side chains. In this work peptide ribbons were preorganized on a macrocyclic scaffold using azapeptide-aldehyde reactions. The resulting cavitands with semicarbazone linkers arrange the peptide backbones at positions that are suitable for self-assembly of dimeric capsules by formation of binding motifs that resemble eight-stranded b-barrels. Self-assembly properties and inside/outside positions of the side chains depend crucially on the chirality of peptides. By rational optimization of successive generations of capsules we have found that azapeptides containing three amino acids in a (L, D, D) sequence give well-defined dimeric capsules with side chains inside their cavities. Taking advantage of the reversibility of the reaction of semicarbazone formation we have also employed the dynamic covalent chemistry (DCC) for a combinatorial discovery of capsules that could not be rationally designed. Indeed, the results show that stable capsules with side chains positioned internally can be obtained even for shorter sequences but only for combination peptides of (L, L) and (D, L) chirality. The hybrid (L, L)(D, L) capsule is amplified directly from a reaction mixture containing two different peptides. All capsules gain substantial ordering upon self-assembly, which is manifested by a two orders of magnitude increase of the intensity of CD spectra of capsules compared with nonassembled analogs. Temperature-dependent CD measurements indicate that the capsules remain stable over the entire temperature range tested (20-100 C). Circular dichroism coupled with TD DFT calculations, DOSY measurements and X-ray crystallography allow for elucidation of the structures in the solid state and in solution and guide their iterative evolution for the current goals. ## Introduction Proteins, with their variety of biological functions and diversity of structural motifs, constitute an unlimited source of inspiration for scientists in many felds. Unsurpassed models of high binding affinity, selectivity and catalytic efficiency are found among proteins' recognition sites and enzymatic pockets. Such effectiveness comes from the presence of precisely organized confned spaces in proteins' interiors: shape-persistent cavities with well-defned positions of functional groups. Artifcial porous materials and molecules that contain cavities also aim at mimicking these structural features for similar purposes (storage, separation and catalysis). The current state-of-the-art techniques in this feld allow for the synthesis of a vast variety of porous materials either in the form of discrete capsules 1 or as infnite crystalline frameworks (MOFs, 2 COFs 3 or SOFs 4 ). However, decoration of the cavities with functional groups is still a tall order. Among porous solid materials metal-organic frameworks (MOFs) with functional organic sites, 5 MOFs mimicking natural enzymes, 6 peptide functionalized MOFs 7 and peptide-based porous materials have been reported. 8,9 Seminal examples of discrete porous moleculescapsulesthat form directional interactions with encapsulated guests were presented by the group of Ballester, 10,11 and by our group. 12 In these examples functional groups constitute integral parts of the capsules' walls, and therefore they are highly conserved and cannot be modifed without affecting the main structural framework. In the current study we aim at developing a more general approach towards functionalization of cavities using natureinspired building blockspeptides. Peptides are commonly perceived as inapplicable components for the construction of porous structures due to their flexibility and poorly defned noncovalent interactions. These features reduce shape persistence of the putative porous products and make their design difficult. We have previously shown that, despite flexibility and low predictability, short peptides can be used to form discrete porous structures provided that they are properly preorganized by positioning at semi-rigid scaffolds and stabilized by the formation of b-sheet-like binding motifs. 13,14 In this work we develop this concept towards internal functionalization of the cavities. We propose construction of capsules in which structural frameworks are formed by peptides' backbones while the functionality is provided by their side chains, positions of which are determined by chirality. This strategy opens up the possibilities of diverse functionalization using a rich pool of amino acids while retaining the integrity of the main structural framework. Such an approach is closely analogous to the design exploited by nature, which constructs enzymatic/recognition sites using amino acids' side chains, while the shapes of the binding pockets are determined by their backbones. However, our approach also introduces new elements: abiotic macrocyclic fragments (required for preorganization) and peptides of unnatural chirality 15 (for internal functionalization). The fnal structural results of this combination are only vaguely predictable based on structural motifs present in natural proteins. Therefore, in order to allow for both the rational design and an element of serendipity we employ two approaches: (a) a strategy of rational iterative improvement and complexity increase in successive generations of capsules and (b) the dynamic covalent chemistry approach (DCC) to potentially amplify the capsules overlooked by rational design. Our results show that the combination of both approaches produces a unique set of peptide-based porous capsules with functionalized interiors, of which only some could be rationally designed. ## Rational design We have previously reported the synthesis and self-assembly of cavitands composed of peptides of different lengths, sequences and chiralities, which were attached to macrocyclic scaffolds (resorcin arenes) using either methylene bridges 16 , imines 13 or acylhydrazone linkers. 14 These semi-open cavitands form noncovalent dimers by hydrogen bonding arranged in b-barrel type binding motifs that involve peptides' backbones, while their side chains are potentially available for other functions. Molecular modelling indicates that imine-based and acylhydrazone-based capsules have enough space inside the cavities to accommodate some side chains; however, all of our experimental attempts to synthesize such capsules have failed resulting in disintegration of the dimers or hydrolysis of the dynamic covalent bonds. We hypothesize that steric crowding due to an acute internal curvature or inappropriate stereochemistry is the limiting factor. Indeed the internal diameters of such artifcial capsules are in the range 13.1-14.7 (measured between opposite C a carbons), while for eightstranded b-barrels the internal diameter is at least 16 (Fig. 1a). Therefore, in the current work we designed capsules with increased diameters by employing semi-rigid semicarbazone linkers (Fig. 1a). The semicarbazone linkers can be synthesized by reaction of azapeptides with aromatic aldehydes. In the predominant extended conformation (all amides trans) the linker extends the internal diameter of the putative capsule up to ca. 17 (Fig. 1a-c) which should be sufficient to accommodate side chains. In order to control the positions of side chains the chirality of peptides was modulated. In homochiral peptide ribbons the side chains of subsequent amino acids point alternatively to the opposite surfaces of the ribbon. In heterochiral peptide ribbons all side chains are positioned at the same surface of the ribbon, but they point towards slightly different directions (Fig. 1b). Upon self-assembly of cavitands chirality is expected to control the position of the side chains. It should be noticed that numerous variables may affect this design and introduce new features. Azapeptides 17 are able to form additional hydrogen bonds 18 and exist in compact cis conformations. Therefore, besides the desired intermolecular hydrogen bonding motifs, 22,23 they may promote the formation of b-turns 24,25 and b-hairpins, 26,27 which are undesirable features in the current design. Moreover, unforeseen structural features that originate from the flexibility of peptide chains and from heterochirality may appear during selfassembly (the binding motifs of linear heterochiral peptides remain mostly unknown). In order to survey these troublesome features we employed dynamic covalent chemistry which empowers serendipity and enables amplifcation of the structures neglected by rational design. ## Synthesis Azapeptides containing single amino acids (4a and 4b) were synthesized using strategy A, which involves the reaction of Cbzhydrazine active ester 1 with amino acid amides 2a-b (Fig. 2a). Strategy A proved to be ineffective for longer azapeptides. Therefore for azapeptides 6a-c, 7a-d and a reference aminebased azapeptide 9 we used a carbonyldiimidazole (CDI) based procedure (strategy B, Fig. 2b). N-terminal aza fragment 5 was obtained in the reaction of an amino acid tert-butyl ester with carbonyldiimidazole and Cbz-hydrazine. After deprotection of the acid function further elongation was performed using standard peptide coupling protocols (with EDCI, OXYMA coupling reagents). ## First generationreconnaissance Reaction of the shortest azapeptide containing L-Phe, 4a, with tetraformylresorcin arene 8 leads to the formation of tetrasubstituted product 10a (Fig. 2c). NMR spectra of 10a in DMSO (Fig. S1-S3 †) confrm the C 4 -symmetric substitution pattern and DOSY indicates that 10a exists as a monomer in DMSO (D ¼ 1.28 10 10 m 2 s 1 , radius 0.86 nm). Crystals suitable for X-ray analysis were obtained by vapour crystallization from a chloroform/methanol mixture. In contrast to the monomeric structure observed in DMSO, the X-ray structure reveals the formation of a dimeric capsule (10a) 2 in the solid state (Fig. 3). The capsule core has an approximate C 4 symmetry with two hemispheres being symmetrically non-equivalent and adopting substantially different conformations. In one of the hemispheres the semicarbazone linker has a trans conformation of an amide bond (hemisphere trans-A), while in the second hemisphere the semicarbazone linker adopts a cis conformation (hemisphere cis-B). The binding motif involves 11-membered rings formed using peptides' backbones from hemisphere trans-A and the cisamide group (also part of the backbone) from the hemisphere cis-B. The most pronounced feature of the azapeptide capsule (10a) 2 , which makes it different from the previously known peptide capsules, is the presence of four phenylalanine side chains inside the cavity. Additionally, there is also enough space in the cavity for encapsulation of two small molecules (here two solvent molecules). The solid state structure of (10a) 2 proves that (a) formation of a capsular dimer is possible for azapeptides and (b) widening of the capsules by the application of longer linkers is indeed a promising route towards functionalization of cavities with peptides' side chains. Although this capsular structure is not retained in DMSO and 10a is only sparingly soluble in less polar solvents (e.g. chloroform) which would facilitate self-assembly, the X-ray structure of (10a) 2 was taken as a starting point for further optimization. The structure suggests that the terminal -NHMe groups and C]O groups from the neighbouring strands are not involved in the binding motif and therefore the binding is not fully complementary. Therefore, we tested the replacement of the terminal group -NHMe group with the -NH 2 group which would potentially allow for its incorporation into the binding motif to improve complementarity. ## Second generationimproved binding motif Indeed, the second-generation cavitand 10b (with a terminal -NH 2 amide group), despite its apparently higher polarity than 10a, is well soluble in chloroform and forms ordered dimeric capsules in solution. The plausible structure of the capsule (10b) 2 , which is in agreement with the spectral data, is depicted in Fig. 4d. The diffusion coefficient is 3.35 10 10 m 2 s 1 (CDCl 3 , 298 K), which corresponds to a radius of 1.22 nm in agreement with the formation of dimer (10b) 2 . The formation of the dimer is accompanied by structural ordering which is re-flected in the CD spectrum (Fig. 4a). Comparison of the CD spectra of (10b) 2 and 13 (a reference cavitand based on a simple chiral amine without dimerization properties) in chloroform shows two orders of magnitude lower intensity for 13 than for (10b) 2 , while the UV spectra of both compounds remain highly similar. In contrast, in methanol, in which the capsules dissociate, the CD spectra of 10b and reference 13 both show similar and low intensity (Fig. 4b). Theoretical CD spectra (TD DFT) calculated using the plausible model agree well with the experimental spectrum, which validates the model. 28 ROESY correlation signals indicate close intramolecular contacts between protons positioned next to the macrocyclic scaffolds (H f and OH) and protons from the terminal -NH 2 group (H l1 and H l2 ) (Fig. 4d and S7 †) which suggest an extended shape of the dimeric capsule (10b) 2 . One of the terminal amide protons -NH 2 is downfeld shifted by 2.28 ppm with respect to the second one, suggesting its involvement in a hydrogen bonding binding motif. The averaged symmetry of capsule (10b) 2 observed in NMR spectra (Fig. 4c and S4, S5 †) is D 4 , thus, both hemispheres and all peptide arms within the hemispheres are symmetrically equivalent and there are no indications of cis/ trans azapeptide isomerism. The signs and shapes of CD effects proved to be sensitive only to the inherent chirality of molecules but not to the cis/trans isomerism, and therefore, they do not allow for determination of stereochemistry of the azapeptide bonds in (10b) 2 . NMR spectra suggest external positioning of phenylalanine side chains (chemical shifts are close to typical values and there are no ROESY correlation signals that can come from internalization). This spectral evidence proves that the binding motif in the second generation capsule (10b) 2 is altered compared to the frst generation capsule (10a) 2 . Through-space contacts in (10b) 2 are in agreement with an extended conformation of peptide strands as observed in the trans-A hemisphere, and therefore in the plausible model the capsule consists of two trans-A hemispheres with the optimal arrangement of terminal and binding groups, but, as it is expected for this single amino acid sequence, with all side chains positioned externally. ## Third generationelongation The third generation capsules (11a-c) contain the -NH 2 terminal amide group and longer peptide sequences consisting of L-Phe at the frst position and L-Ala(11a), D-Ala (11b) or D-Leu (11c) at the second position. The sequences were selected in order to check the influence of chirality (L versus D) and size (Ala versus Leu) on the self-assembly and the structure of capsules. Homochiral product 11a exists in DMSO as a monomeric cavitand (D ¼ 1.40 10 10 m 2 s 1 , r ¼ 0.78 nm) and it is poorly soluble in chloroform. Poor solubility in non-polar medium is indirect evidence that 11a does not form a complementary binding motif, and it has numerous solvent-exposed hydrogen bond donors and acceptors. In contrast, heterochiral derivatives 11b and 11c are well-soluble in chloroform and form discrete dimeric species (11b) 2 and (11c) 2 of D 4 symmetry (D ¼ 3.20 10 10 m 2 s 1 , radius 1.27 nm, CDCl 3 , 298 K). Signals of the terminal amide -NH 2 group are differentiated by 2.2 ppm suggesting their involvement in the binding motif, while the presence of ROESY correlation signals between the semicarbazone linker and terminal alanine (side chain (CH 3 ) n and a-CH m ) suggests a head-to-head dimer of an extended shape (Fig. 5b and c). For capsules (11b) 2 and (11c) 2 both side chains are pointing to the same surface of the ribbon in agreement with the structural features of a heterochiral sequence, and therefore, they are expected to be positioned externally. Indeed, the chemical shifts for the signals of terminal side chains, better solubility of (11c) 2 compared with (11b) 2 and independence of the self-assembly process from the steric effects introduced by the Leu side chain prove that the side chains of the third generation of capsules are positioned outside the cavity. The X-ray structure of (11b) 2 confrms the dimeric nature (Fig. 5a); however, the solid state structure is different than the suggested structure in solution. The approximate symmetry of the capsule core in the solid state is C 4 and, as in the case of (10a) 2 , two different hemispheres trans-A and cis-B are present. In the trans-A hemisphere all peptide strands have extended conformations with side chains positioned externally. The terminal amide -NH 2 groups form hydrogen bonds with phenolic OH groups (N/O distances are 2.9-3.3 ) of the second hemisphere (cis-B). The geometry of the trans-A hemisphere and non-covalent contacts are in agreement with the ones postulated in solution. However, the geometry and noncovalent contacts observed for the cis-B hemisphere are substantially different in the solid state than in solution. In the cis-B hemisphere the semicarbazone linker adopts a cis conformation which enforces a folded conformation of the peptide strand. As a consequence the side chain of the frst amino acid (Phe) is positioned inside the cavity, and the side chain of the second side amino acid (Ala) is positioned outside the cavity and the terminal amide -NH 2 groups are not involved in the intra-capsular binding motif but involved mostly in intercapsular interactions. Although observed in the solid state, such a conformation is not present in solution (there are no relevant chemical shifts nor ROESY correlation signals). Therefore, we assume that the geometry of the trans-A hemisphere better reflects the approximate geometry in solution and the dimer's structure in solution resembles an extended dimer consisting of two trans-A-like hemispheres (Fig. 5c). The geometry of the cis-B hemisphere could be a result of interactions present only in the solid state: (a) competing inter-capsular interactions that involve the terminal amide groups and (b) preference towards a more compact (and less porous) structure due to spatial confnement. However, this solid state structure reflects the fact that due to the non-perfect complementarity and flexibility the capsule is dynamic. ## Fourth generationinternal functionalization The analysis of the third-generation capsules leads to the conclusion that the homochiral junction between positions 1 and 2 leads to destabilization of the structure. Position 2 is probably still too crowded for placement of a side chain inside the cavity. On the other hand, facile formation of capsules made of heterochiral peptides proves the efficiency of this binding motif. In the next generation capsules we combined these two observations and designed a three amino acid sequence with a heterochiral junction (L-Phe-D-Ala) for effective self-assembly and a homochiral junction (D-Ala-D-Ala) for internal functionalization. Product 12a forms a well-defned dimer (12a) 2 of D 4 symmetry in chloroform solution. The spatial arrangement in capsule (12a) 2 was analysed using ROESY and DOSY and all data are in agreement with the formation of a head-to-head dimer with peptide arms in extended conformations (D ¼ 3.00 10 10 m 2 s 1 , radius 1.36 nm, CDCl 3 , 298 K). Importantly, 1 H NMR shows that the signal of the side chain of the terminal alanine (CH 3 ) is substantially upfeld shifted and appears at 0.57 ppm, which indicates that this group is directed towards the interior of the capsule, in close proximity to aromatic walls (Fig. 6). Additionally it shows ROESY correlations with H g , H f and H i . Homologous cavitand 12b, in which the cavity-exposed Ala was replaced by more bulky Phe, also gives capsular species (12b) 2 , with internally exposed phenylalanine side chains. However, the capsules are much less ordered as indicated by broad signals in the NMR spectrum and less stable, as they disintegrate within days upon standing in solution. Thus, rational structure optimization of azapeptide capsules resulted in innercavity functionalization and proved the crucial role of chirality in both efficient self-assembly and functionalization of the inner cavity. Control experiments with azapeptides containing three amino acids with two heterochiral junctions (7c and 7d) confrm the crucial role of chirality. Thus the (L, D, L) sequence in 12c and 12d precludes the formation of capsules. The cavitand 12d is insoluble in non-polar solvents. The homologous product 12c, which possesses two hydrophobic phenylalanines in each arm, is well-soluble in chloroform, but it exists as a self-folded C 2 -symmetric cavitand in this environment (doubled number of NMR signals, D ¼ 3.7 10 10 m 2 s 1 , radius 1.10 nm, CDCl 3 , 298 K). The shape of the ECD spectrum of 12c and its comparison with the spectra of other cavitands suggest that structural changes lower the symmetry of the resorcinarene core (Fig. 7c and 4a). We assume that the formation of b-turns and a system of intramolecular hydrogen bonds is responsible for this unique conformation of cavitand 12c. The suggested self-folded structure of 12c, which involves b-turns 29 and a C 2 -symmetric boat-like conformation of the macrocycle, is presented in Fig. 7b. ## Fih generationserendipity The reversible character of reactions between aldehydes and amine-based groups makes them applicable in dynamic covalent chemistry (DCC). 30 We have previously exploited the DCC approach for the synthesis of various homo-and heterochiral capsules based on imine or acylhydrazone linkers and proved that they amplify and self-sort out of the mixtures of racemic peptides. 13,14 However, the dynamic covalent chemistry of azapeptides is not known. What is more, reversibility of the reaction between semicarbazide groups (being part of the azapeptide functionality) and aldehydes has only been studied for the simplest semicarbazide in a catalytic variant. 31 With a set of various azapeptides in hand we explored the reversibility and possibility of amplifcation of self-assembled capsules from mixtures of azapeptides. First, we determined the kinetics of the non-catalytic reactions of azapeptides with 8. The time-dependent experiments indicate that the reactions in chloroform at 70 C require several to twenty four hours to reach the stationary state (Fig. S9, S20 and S39 †). This period of time depends on the peptide length and solubility. Next, we studied the reaction between 8 (C ¼ 30 mM) and two azapeptides of different sequences but the same length (C(aza1) ¼ C(aza2) ¼ 60 mM) in CDCl 3 by NMR (Fig. 8a). It should be noted that in most cases such reaction mixtures cannot be analysed by HPLC or MS due to the instability of self-assembled capsules on a chromatographic support or in the gas phase or due to identical mass (for substrates that are stereoisomers). Therefore, due to the complexity and resolution limits of NMR in each reaction we used simultaneously only two azapeptides. The reaction of 8 with rac-7a reaches equilibrium within 2 days (Fig. S40 †) and the spectrum of the racemic mixture is virtually identical to that of the mixture with the enantiomerically pure substrate 7a (Fig. 8b) indicating very efficient chiral self-sorting. In the case of rac-6b the capsular product (11b) 2 visibly predominates, but self-sorting isn't quantitative. This result was further confrmed by a self-sorting experiment with a pseudoracemic mixture of (L, D)-6b and (D, L)-6c which could be analysed by MS (Fig. S27 †). Peaks from enantiomerically pure C 4 -symmetrical cavitands are the highest, while peaks from mixed products are also present which indicates non-quantitative chiral self-sorting. These results prove that: (1) azapeptides react with aldehydes in a reversible way and therefore they can be used as substrates in dynamic covalent chemistry and (2) self-assembly is a powerful factor that stabilises capsular structures and therefore leads to their amplifcation. With positive results for the dynamic character of azapeptide-aldehyde reactions and encouraging amplifcation of selfassembled structures we sought after new, hybrid products that can be obtained only from the multisubstrate mixtures. The reaction of 8 with a mixture of (L, L)-6a and (D, L)-6b leads to complete consumption of the substrates within 2 days and the initial formation of homodimeric capsules (capsule (11b) 2 remains in solution, while product (11a) 2 is poorly soluble, Fig. 9a). However, with time an additional set of signals start to emerge that was not present for neither of the reactions containing single azapeptides as substrates (Fig. 9a and S33 †). After 6 days the initially formed homodimeric products were quantitatively transformed into new species. Analysis of NMR spectra indicates that the fnal product is a capsular heterodimer (L, L-11a)(D, L-11b) in a head-to-head arrangement (correlation signals between terminal NH 2 and H f protons) and of roughly similar size to the previously obtained capsules (D ¼ 3.25 10 10 m 2 s 1 , radius 1.25 nm). Terminal alanine residues from one hemisphere are located inside the cavity (0.62 ppm for CH 3 protons) and have correlation signals in the ROESY spectrum with protons from the semicarbazone linker region (H g , H f , H h and H i ) from the other hemisphere. The CD effects for the lowest energy bands in the CD spectrum of (11a)(11b) have almost diminished, and the spectrum is almost an algebraic sum of the two components which indicates that the capsule consists of two hemispheres that are twisted in the opposite directions (Fig. 9c). The plausible structure of the capsule (11a)(11b) that meets all these experimental constraints is presented in Fig. 9b. With a set of peptide capsules of various geometries and chiralities we aimed at the determination of their thermal stability. Temperature-dependent CD measurements for homodimeric capsules (L-10b) 2 , (L, L-11a) 2 , and (L, D-11b) 2 , and heterodimeric capsule (L, L-11a)(D, L-11b) in tetrachloroethane (TCE was used instead of chloroform because it has higher boiling point, and ECD spectra in TCE are analogous to the ones in chloroform, S23 †) show that all spectra of homodimeric capsules remain invariable (Fig. 10a, S10, S15, S23 and S42 †). Because for nonassembled cavitands the intensities of the ECD bands are two orders of magnitude lower, it can be concluded that homodimeric capsules are thermally stable over the entire temperature range tested (20-100 C). For the heterodimeric capsule (L, L-11a)(D, L-11b) the intensities of the ECD band are initially two orders of magnitude lower (due to hemispheres of different chiralities) and they decrease with temperature indicating possible instability of the assemblies at higher temperatures (Fig. 10b and S34 †). ## Conclusions and outlook In summary, we have elaborated a strategy for the synthesis of capsules with functionalized internal cavities based on azapeptides. The strategy involves rational design of a macrocyclic compound and a semicarbazone linker, optimization of the binding motif (modifcation of the C-terminus) and modulation of chirality of peptides towards proper positioning of the side chains (homo-and heterochiral sequences of peptides). As a result we obtained capsules that self-assemble using hydrogen bonding interactions between their backbones and have side chains positioned in their cavities. Additionally, we have proven the reversible character of reactions between azapeptides and aldehydes. This fnding opens a way towards the application of azapeptides in dynamic covalent chemistry and towards the synthesis of new capsules in a combinatorial way. Indeed, from a set of combinatorial experiments an unexpected hybrid capsule was obtained, which wasn't designed in a rational way nor it could be obtained from the reaction employing only a single azapeptide as a substrate. Further self-sorting experiments of azapeptide mixtures indicate that amplifcation of well-defned products is effective only in cases when capsules are formed. These results prove that self-assembly is an effective stabilization factor that affects thermodynamic equilibria and leads to self-sorting even for highly similar and conformationally labile substrates. This observation is in contrast to the common opinion that rigidity and orthogonality of interactions are indispensable features for selective self-sorting. The design presented in this paper is closely inspired by natural eight-stranded b-barrels. However, such b-barrels, due to their small internal diameter, have limited abilities to host molecules. This limitation can also be a drawback of the azapeptide capsules. However, the current strategy allows for elongation and alternation of the positions of side chains, and the dynamic nature of the cores enables expansion. The capsules are not fully sealed due to imperfect complementarity of binding motifs, so they can have portals for guest entrance/ release. Thus, in contrast to the crowded interiors of natural b-barrels, interiors of these artifcial capsules are more accessible. Another limitation of eight-stranded b-barrels is that bulky polar side chains are avoided inside the barrels. In fact, also in these seminal examples only the smallest side chain (methyl group from Ala) was accommodated in the proximity of a hydrophobic pocket. However, the strategy suggested in this paper allows for elongation of the peptides and changes in the position of side chains, and therefore, different locations of internal side chains are also possible. What is more, the formation of hybrid capsules opens up possibilities of diversi-fcation of hemispheres and, for example, encapsulation of acid-base pairs at alternating positions mimicking natural catalytic sites (functional dyads, triads, etc.). ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "Peptide-based capsules with chirality-controlled functionalized interiors \u2013 rational design and amplification from dynamic combinatorial libraries", "journal": "Royal Society of Chemistry (RSC)"}

synthesis_of_zearalenone-16-β,d-glucoside_and_zearalenone-16-sulfate:_a_tale_of_protecting_resorcyli

## Abstract: The development of a reliable procedure for the synthesis of the 16-glucoside and 16-sulfate of the resorcylic acid lactone (RAL) type compound zearalenone is presented. Different protective group strategies were considered and applied to enable the preparation of glucosides and sulfates that are difficult to access up to now. Acetyl and p-methoxybenzyl protection led to undesired results and were shown to be inappropriate. Finally, triisopropylsilyl-protected zearalenone was successfully used as intermediate for the first synthesis of the corresponding mycotoxin glucoside and sulfate that are highly valuable as reference materials for further studies in the emerging field of masked mycotoxins. Furthermore, high stability was observed for aryl sulfates prepared as tetrabutylammonium salts. Overall, these findings should be applicable for the synthesis of similar RAL type and natural product conjugates. ## Introduction Resorcylic acid lactones (RALs, Figure 1), a compound class of benzannulated macrolides, are pharmacologically active secondary metabolites produced by a variety of different fungal species . Zearalenone (ZEN, 1) is a well-known RAL type mycotoxin for which maximum tolerated levels in food and feed were enacted and recommended, respectively, in Europe . ZEN is produced by several plant pathogenic Fusarium species, including Fusarium graminearum and Fusarium culmorum. These species, which are the most frequently occurring toxin-producing fungi of the northern temperate zone, are commonly found in cereals and crops throughout the world . Significant levels of ZEN are prevalently found in grains such as maize, wheat, and rice . It is known that ZEN can cause problems of the reproductive tract (e.g., impaired fertility) in animals . Physiological studies revealed binding of ZEN to recombinant human estrogen receptors and have furthermore shown a ZEN-induced stimulation of the growth of human breast cancer cells . Additionally, masked mycotoxins, especially altered derivatives formed through conjugation to sugar moieties or sulfate, emerge after metabolization by living plants. Due to changed chemical structures and properties compared to the parent mycotoxins, these conjugates can usually not be detected applying standard analytical techniques . Responsible biochemical transformations are catalyzed usually by enzymes within detoxification processes . Schneweis et al. reported the occurrence of ZEN-14-β,D-glucoside (5, Figure 2A) in wheat and the first chemical synthesis of this compound applying phase transfer glycosylation has been reported by Grabley et al. . ZEN-14-sulfate (6, Figure 2A) was first isolated from F. graminearum-inoculated rice and both, the 5) and ZEN-14-sulfate ( 6), (B) ZEN-16-β,D-glucoside (7) and ZEN-16-sulfate (8). Sulfates shown as sodium salts. glucoside 5 and the sulfate 6, were identified as ZEN metabolites in Arabidopsis thaliana . These conjugates are easily hydrolyzed back to the parental mycotoxin during digestion of contaminated grain, and should therefore be considered as masked mycotoxins . Recently it has been shown that ZEN treated barley, wheat and Brachypodium distachyon cells produce both the ZEN-14 and the ZEN-16-glucoside, with up to 18-fold higher levels of ZEN-16-glucoside than ZEN-14-glucoside in barley roots . We therefore intended to develop a synthetic method for regiocontrolled conjugation of ZEN. Basically, the RAL type moiety of ZEN contains two possible sites for glycosylation/sulfation, but due to the higher reactivity of the phenol group at position 14, reactions at this site are strongly favored compared to conjugation at C16-OH . Although natural products containing a RAL type moiety conjugated at the phenol group in ortho position to the carboxyl group (2'-OH) were already detected and identified , to the best of our knowledge there are no reliable synthetic procedures and strategies towards this class of compounds described in the literature so far. The synthesis of the natural glucoside delphoside by Saeed was performed using methyl ether protection at O-6 of the isocoumarin core structure during glucosylation and rather harsh unfavorable demethylation with boron tribromide in the last step . Without structure verification and characterization ZEN-16-β,D-glucoside (7, Figure 2B) was tentatively identified as a byproduct of the Königs-Knorr glucosylation of ZEN for preparation of ZEN-14-glucoside . In the course of ongoing research in the emerging field of masked mycotoxins, we were able to prepare ZEN-16-β,D-glucoside (7) and ZEN-16-sulfate (8, Figure 2B) in reasonable amounts after the development of reliable procedures that should be generally applicable to resorcylic acid lactones. Additionally the first chemical synthesis of the ZENderivative 14-O-acetylzearalenone (14-AcZEN, produced by some Fusarium strains) is reported. ## Results and Discussion The general strategy for regiocontrolled conjugation at position 2' of resorcylic acid esters and lactones is shown in Scheme 1. Regioselective protection of the more nucleophilic 4'-phenol and subsequent glucosylation or sulfation should lead to the desired products. Scheme 1: General strategy for the synthesis of RAL-2'-conjugation (Pg: protective group, pGlc: protected glucose moiety, pS: protected sulfate, Glc: β,D-glucoside). For the development of a reliable protective group strategy and subsequent reaction optimization, 2,4-dihydroxybenzoic acid isopropyl ester (9) was used as a RAL mimic. For protection of the 4-OH group we first considered an acetyl group that could be regioselectively introduced by reaction of 9 with acetic anhydride and catalytic amounts of 4-(dimethylamino)pyridine (DMAP) to obtain the acetylated RAL mimic 10 (Scheme 2A). Different methods for glycosylation were investigated using acetyl-protected glucosyl donors since diastereoselective β-conjugation, which is needed for the preparation of glucosides formed during phase II metabolism, is commonly achieved applying the participation of acyl groups at O-2 of the glycosyl donor (anchimeric effect) . Lewis acid-mediated glucosylation using the trichloroacetimidate donor 11 according to the procedure of Saeed did not lead to the desired product, which can be explained by the weak nucleophilicity of the 2-OH group of the acceptor 10. This assumption was supported by detection of the glucosyl acetamide 12, which is known to be formed by rearrangement of 11 when activated in the presence of a weak acceptor (Scheme 2B) . Königs-Knorr glucosylation, which in general is most frequently used for the glucosylation of phenols, using commercially available bromo sugar 13 activated by silver(I) salts or under phase transfer conditions led to complex product mixtures (Scheme 2C). Nevertheless, the procedure for selective acetylation of resorcylic acid esters and lactones was applied for the first synthesis of 14-O-acetylzearalenone (14) (Scheme 3). To avoid undesired cleavage of the acetyl group during glucosylation of 10, p-methoxybenzyl (PMB) protection was applied instead, since the PMB group was considered to be inert to the reactions conditions of the Königs-Knorr procedure, thus forcing conjugation at position 2. Regioselective p-methoxybenzylation of 9 was achieved by reaction with PMB-Cl and Cs 2 CO 3 in dry DMF after optimization in terms of base and solvent type (Scheme 4A). Königs-Knorr glucosylation of the PMB-protected mimic 15 afforded 16, which was deprotected using 2,3-dichloro-5,6-dicyano-p-benzoquinone (DDQ) for oxidative PMB cleavage and subsequent ester saponification to yield the desired glucoside 17 in an overall yield of 30% (Scheme 4B). Applying this procedure to ZEN (1) afforded the glucosylated intermediate 19 after Königs-Knorr glucosylation of 14-PMB-ZEN (18), but subsequent deprotection using DDQ did not afford the desired product. Also an alternative procedure for oxidative PMB cleavage using cerium ammonium nitrate (CAN) did not lead to the formation of the deprotected compound 20 (Scheme 5). Beside unreacted 19, LC-MS/MS analysis showed the formation of a product with an m/z value 16 amu higher than calculated for 19 indicating oxidation of this intermediate but no deprotection. Since the DDQ-promoted cleavage of phenolic PMB ethers can be complicated by overoxidation, especially with electron-rich phenolic compounds , we assume a significant effect of the conjugated olefinic double bond at C-6' of the resorcylic acid moiety being responsible for the observed different behavior of ZEN (1) and the ZEN mimic 9. After this second setback the protective group strategy was changed again within a third approach. Considering steric hindrance and methods for selective deprotection under relatively mild conditions led us to the use of triisopropylsilyl (TIPS) protection for regiocontrolled glucosylation of resorcylic acid esters and lactones. Regioselective silylation of 9 and ZEN (1) was readily achieved affording compounds 21 and 22 (14-TIPS-ZEN), respectively, in nearly quantitative yields by reaction with TIPS-Cl and imidazole in dry CH 2 Cl 2 . Applying this strategy we were finally able to accomplish the synthesis of the ZEN mimic glucoside 17 (Scheme 6A) as well as of the target compound ZEN-16-β,D-glucoside (7) as shown in Scheme 6B. Reasonable yields of 41% (17) and 34% (7), respectively, were obtained applying an optimized purification protocol. Additionally, TIPS protection was applied for the synthesis of ZEN-16-sulfate (8) using a procedure that was successfully applied for the synthesis of ZEN-14-sulfate (6) as described recently . Reaction of 22 with the 2,2,2-trichloroethyl (TCE) protected sulfuryl imidazolium salt 23 protected sulfate 24. TIPS cleavage and subsequent TCE deprotection using Zn/ammonium formate (HCOONH 4 ) yielded the desired product. Interestingly, when using the crude intermediate after TIPS deprotection without purification directly in the second step, we obtained the tetrabutylammonium salt of ZEN-16-sulfate (NBu 4 -8) in good yield (65% over 2 steps) as shown in Scheme 7. The stability of this compound was significantly increased compared to the corresponding sodium salt, which is of great importance in terms of preparation of reference ma-terial and long term stability of appropriate standard solutions for further investigations. ## Conclusion In summary, different methods for the regioselective protection of resorcylic acid esters and lactones were investigated for subsequent regiocontrolled glucosylation and sulfation. Whereas acetyl and p-methoxybenzyl protection led to undesired products, TIPS-protected RALs were successfully used as intermediates for the preparation of corresponding glucosides and sulfates applying the Königs-Knorr glucosylation and chemical sulfation using TCE-protected sulfuryl imidazolium salt 23, respectively. These methods were used for the first chemical synthesis of the ZEN-16-conjugates 7 and 8 in reasonable amounts for ongoing research and further investigations in the field of conjugated/masked mycotoxins. ## Supporting Information Supporting Information File 1 Experimental details (including remarks and general procedures), characterization data, copies of NMR spectra of new compounds, 2D NMR spectra of glucoside 7. [http://www.beilstein-journals.org/bjoc/content/ supplementary/1860-5397-10-112-S1.pdf]

chemsum

{"title": "Synthesis of zearalenone-16-\u03b2,D-glucoside and zearalenone-16-sulfate: A tale of protecting resorcylic acid lactones for regiocontrolled conjugation", "journal": "Beilstein"}

functionalized_graphene-based_biomimetic_microsensor_interfacing_with_living_cells_to_sensitively_mo

## Abstract: It is a great challenge to develop electrochemical sensors with superior sensitivity that concurrently possess high biocompatibility for monitoring at the single cell level. Herein we report a novel and reusable biomimetic micro-electrochemical sensor array with nitric oxide (NO) sensing-interface based on metalloporphyrin and 3-aminophenylboronic acid (APBA) co-functionalized reduced graphene oxide (rGO). The assembling of high specificity catalytic but semi-conductive metalloporphyrin with high electric conductive rGO confers the sensor with sub-nanomolar sensitivity. Further coupling with the small cell-adhesive molecule APBA obviously enhances the cytocompatibility of the microsensor without diminishing the sensitivity, while the reversible reactivity between APBA and cell membrane carbohydrates allows practical reusability. The microsensor was successfully used to sensitively monitor, in real-time, the release of NO molecules from human endothelial cells being cultured directly on the sensor. This demonstrates its potential application in the detection of NO with very low bioactive concentrations for the better understanding of its physiological function and for medical tracking of patient states. ## Introduction Nitric oxide (NO) has attracted intense interest due to its diverse and vital role in the regulation of physiological processes, such as cardiovascular systems, neurotransmission, immune responses and angiogenesis. The quantifcation of NO in vivo is essential to comprehensively unravel its function in physiology. However, measuring NO in biological systems is challenging due to its short half-life (<10 s), trace level (physiological range of NO bioactivity is as low as 100 pM or below) and the presence of complex interfering species. Methods commonly available, including chemiluminescence, absorbance, fluorescence and electron paramagnetic resonance spectroscopy, suffer from high costs or complicated pretreatments that restrict real-time measurement. Although newly developed techniques such as feld-effect transistors or surface-enhanced Raman spectroscopy allow for the real-time detection of NO outside or inside living cells, 16,17 an electrochemical NO sensor based on microelectrodes can be a powerful tool for the direct and real-time monitoring of NO release in physiology due to the ultra-fast response time, high-spatial resolution and non-invasive size. 7, In recent times, great efforts have been made to ensure that electrochemical sensors have sufficient sensitivity and selectivity to accurately detect NO. To improve sensor sensitivity, different types of catalytic materials, such as metalloporphyrins (phthalocyanine and salen), carbon nanotubes, 26 metal nanoparticles and nanowires, have been employed to develop sensitive NO sensors. With its unique properties, graphene has emerged as a rising star material and it has advanced the development of electrochemistry. Recently, graphene has shown a great potential for the sensitive detection of NO. Moreover, enhanced selectivity was achieved by coating a permselective membrane onto the sensor. 7,21,22 Nafon, Teflon and o-phenylenediamine are often used to restrict the diffusion of interferent molecules (e.g., nitrite, ascorbic acid and dopamine) to the electrode surface. 23,36,37 In comparison with isolated cells and the above electrode confguration, culturing cells on microelectrodes could reconstruct the in vivo interface between NO emitting cells and their receptor cells (e.g., endothelium and smooth muscle cells) better, and the monitoring of NO from the tightly attached cells could therefore model real physiological conditions more closely. In this case, biocompatibility is a critical issue that should be seriously considered. Pre-coating with positively charged polyelectrolytes (e.g., poly-L-lysine 38 ) or an extracellular matrix (e.g., laminin 39 and collagen 40 ) is typically utilized to boost cell attachment. However, the additional polymer and extracellular matrix for selectivity or biocompatibility enhancement will actually decrease electrode response. 7,17,41 It was also quite difficult to coat very small or irregularly shaped electrode surfaces in a controlled way. Recently, small cell-adhesive molecules, such as RGD and monosaccharides, 34,41 have been explored for interfacing electrodes with living cells, providing an approach to fabricate biocompatible sensors. So far, great success has been achieved in developing NO electrochemical sensors with sensitivities above nanomolar concentrations. However, the construction of sensing-interfaces with sub-nanomolar to picomolar sensitivities that concurrently possess excellent cytocompatibility and high selectivity remains a great challenge, especially for the high throughput detection at single cell levels (i.e., single or a few cells). Herein, we have developed a multifunctional microsensor array for the detection of NO from several cells. Inspired by the excellent catalysis of NO electrooxidation by metalloporphyrins and the two-dimensional nanostructure and high electric conductivity of graphene, we prepared novel hybrid nanosheets, based on Fe(III) meso-tetra (4carboxyphenyl) porphyrin (FeTCP) and reduced graphene oxide (rGO), through p-p interactions, which were then deposited onto an ITO microelectrode array via electrophoretic deposition. These nanosheets, named FGHNs, were further functionalized covalently with a small cell-adhesive molecule, 3-aminophenylboronic acid (APBA). This can react with the 1,2-or 1,3diols in carbohydrate moieties that exist largely on cell membranes to provide the sensor with good cytocompatibility. The as-prepared microsensor array and the electroactive area are shown schematically in Scheme 1. The microsensor demonstrated exceptional sensitivity and selectivity to NO with a detection limit of 55 pM in PBS and 90 pM in a cell medium. Human umbilical vein endothelial cells (HUVECs) can adhere and proliferate well on the electrode surface, therefore promoting interfaces with the living cells being cultured thereon and high sensitive real-time monitoring of NO release. ## Preparation and characterization of the FGHNs sheets The FGHNs were prepared by a procedure similar to that previously reported. 46 Briefly, water-soluble GO and FeTCP were ultrasonically mixed to promote the p-p interaction between them, then hydrazine hydrate was added to the above solution to reduce the GO (see Methods in ESI †). As shown in Fig. 1a, a stable FGHNs dispersion was obtained (Fig. 1a, inset III) after hydrazine reduction, with the peak at 226 nm for GO, attributed to the p-p* transition of aromatic C]C bonds, shifting to 261 nm for rGO. 47 A new absorption peak at 423 nm was also observed, which corresponds to the Soret band of FeTCP with a red shift (27 nm). The results indicate that FeTCP was adsorbed on the surface of the rGO sheets by p-p interactions. Meanwhile, when compared with FGHNs, the dispersion of rGO without the stabilization of FeTCP leads to the aggregation of rGO sheets after reduction (Fig. 1a, inset II). The attachment of FeTCP on the rGO surface was also characterized by an electrochemical method. Fig. 1b shows the cyclic voltammograms of bare ITO (blue line), FeTCP/ITO (purple line), rGO/ITO (red line) and FGHNs/ITO (black line) in 0.1 M phosphate buffered saline (PBS). Compared with ITO and rGO/ITO, a pair of redox peaks were observed in the potential range from 0.6 V to 0.0 V for FGHNs/ITO. The redox peaks should obviously belong only to FeTCP in FGHNs, which is characteristic of the electron transfer process of iron at the core of Fe(III) TCP/Fe(II) TCP (Fig. 1b, purple line). In addition, AFM results show that the average thickness of the single-layer FGHNs was determined to be about 1.45 nm (Fig. 1c). There was a 0.80 nm increment compared with that of pure rGO, the single-layer thickness of which is ca. 0.65 nm, 47 owing to the presence of FeTCP on the rGO sheet surfaces. Thus the thickness of the FeTCP layer was calculated to be about 0.40 nm, since FeTCP can locate on both sides of the rGO. 49,50 X-ray photoelectron spectroscopy (XPS) was employed to further explore the interaction between rGO and FeTCP. Compared with GO (Fig. 2a), the survey of FGHNs shows the presence of detectable amounts of N1s at about 400 eV, originating from FeTCP, which indicates that the functionalization of rGO by FeTCP occurred successfully (Fig. 2b). After hydrazine reduction, the C1s spectrum indicates that the peaks associated with C-C or C-H (284.3 eV) become predominant, while the peaks related to the oxidized carbon species (C-O, C]O) are greatly weakened (Fig. 2c and d). Meanwhile, a new peak appeared at 285.7 eV in the spectra of FGHNs, which belongs to the carbon in the C-N bonds. These results indicate that GO has been well deoxygenated into rGO and further protected by FeTCP molecules to form FGHNs. ## Electrochemical behaviors of FGHNs and APBA/FGHNs The FGHNs/ITO exhibited excellent electrochemical behavior to NO (the standard solution was prepared as previously described 26 ) and the differential pulse and cyclic voltammograms showed a peak at +0.65 V and +0.75 V (Fig. 3a), respectively. It is worth noting that FGHNs/ITO acted as the best catalyst compared to rGO/ITO and FeTCP/ITO (Fig. 3b), the sensitivities of which were calculated to be 37.6, 7.2 and 2.1 mA mM 1 cm 2 , respectively. The metalloporphyrin was used as a catalytic coating for the construction of NO electrochemical sensors to improve sensitivity and specifcity, 7,21,23 but it has poor electrical conductivity. The introduction of the underlying rGO greatly enhanced the catalytic capability to NO by providing a highly conductive bridge to facilitate rapid transport of electrons between the porphyrin and the electrode. 48,50 This attributes to the amazingly synergistic electrooxidation of NO. FGHNs/ITO were further functionalized covalently with the small cell-adhesive molecule APBA by coupling the -COOH and -NH 2 groups in FeTCP and APBA. Attenuated Total Reflection Infrared Spectroscopy (ATR-IR) of FGHNs displayed two peaks at 1720 and 1586 cm 1 (Fig. 3c, black line), which are assigned to the stretching mode of a carboxyl group in FeTCP. After the covalent bonding of APBA with FGHNs, two new peaks at 1643 and 1550 cm 1 corresponding to the characteristic amine groups were observed (Fig. 3c, red line). In addition, the presence of the B-O stretching mode at 1344 cm 1 , together with the clearly aromatic C-H stretching at 801 and 710 cm 1 , further evidenced the successful immobilization of APBA. The comparison of electrochemical responses in PBS solution (data not shown) and the cell medium (Fig. S1 †) were also investigated, and the average sensitivities of APBA/FGHNs/ITO in both conditions remained at more than 80% (the ratio of the calibration curve slopes) of FGHNs/ITO after its further functionalization and reuse (Fig. 3d, inset), indicating the additional influence of small cell-binding molecule APBA to the response of electrodes. The selectivities of FGHNs/ITO and APBA/FGHNs/ ITO toward NO in PBS were studied by investigating interferents such as ascorbic acid (AA), dopamine (DA), uric acid (UA) and NO 2 (Fig. S2 †). The calculated selectivity ratios for NO against AA, DA, UA and NO 2 were 89, 74, 111 and 89 for FGHNs/ITO, and 113, 96, 158 and 117 for APBA/FGHNs/ITO, indicating that FGHNs/ITO possesses good selectivity against these interferents, especially for negatively charged molecules, and APBA/ FGHNs/ITO has better performance than FGHNs/ITO. The good selectivity might originate from the highly inherent specifcity of the metalloporphyrin to the electrocatalytic oxidation of NO. The retained surface carboxyl groups together with boronate of APBA/FGHNs also enhanced the selectivity. To further demonstrate the selectivity of the sensor in physiological solutions, we conducted the measurements in the cell culture medium RPMI 1640 with serum (Fig. 3d). In addition to AA, DA, UA and NO 2 , other potential interferents including 5-hydroxy tryptamine (5-HT), H 2 O 2 , L-arginine (L-Arg) and acetylcholine (ACh) were tested (Fig. S3 †), with 2 mM of each interferent. The sensor exhibited a practical selectivity against most molecules, except positively charged DA and 5-HT with a small detectable signal. In addition, the response of the sensor remained about 90% of the initial current after 10 measurements, indicating high reproducibility of the sensor. ## Cell adhesion and proliferation on biomimetic APBA/FGHNs lm Construction of a cell-compatible sensing interface on which cells could be immobilized and grown directly is of great importance to detect cell-released molecules accurately, especially for those that are labile and could be metabolized rapidly, for example, free radical NO. APBA is a molecule capable of binding specifcally with the 1,2-or 1,3-diols which exist in the carbohydrates of cell membranes, therefore it can be employed as an artifcial carbohydrate-receptor for cell adhesion. After being seeded on different substrates with the same cell density and cultured for 1 h, the cells were then rinsed three times with physiological saline solution to remove loosely bound cells. It was observed that the greatest number of HUVECs were left on APBA/FGHNs/ITO (Fig. S4 †), indicating the powerful cell-adhesive capacity of APBA/FGHNs. The cell proliferation behavior was also investigated by counting the number of HUVECs cultured for 1 h, 12 h, 24 h, 36 h, 48 h and 72 h (Fig. 4a), until they proliferated and covered almost all over the electrode. The cell viability was then characterized by the fluorescent live/dead cell markers Calcein-AM and PI after being cultured for up to 86 h, and cells were almost clearly alive (Fig. 4b), further displaying the excellent cytocompatibility of APBA/FGHNs. ## Microelectrode array fabrication and NO release monitoring To realize the real-time monitoring of NO release at the single cell level, we fabricated a patterned ITO microelectrode (100 mm in diameter) array using photolithography techniques (see Methods section in ESI, † and an illustration of the process in Fig. S5 †). 51 The FGHNs and APBA were then deposited on the microelectrode by electrophoresis and covalent linkage, respectively (see Methods section in ESI, † and the electrode morphologies of ITO and APBA/FGHNs/ITO microsensor shown in Fig. S6 †). To simulate the cell-microsensor confguration during the electrochemical detection of NO release, a fast (ca. 5 s) injection of different concentrations of NO solution in a micro-capillary (150 mm internal diameter, close to the active area of the electrode) was performed in PBS and the cell medium, with the results shown in Fig. S7, 5a and S8, † respectively. The sensitivity of the APBA/FGHNs/ITO microelectrode in the RPMI 1640 (0.089 nA nM 1 , Fig. S8 †) was about 70% of that in PBS solution (0.1232 nA nM 1 , Fig. S7 †), which may result from the partial adsorption of undesired compounds in the cell medium onto the active surface of the microelectrode. The typical response time of this sensor to NO was about 400 ms (Fig. S7a, † inset) and 600 ms in PBS and the cell medium (Fig. 5a, inset), if we defne the response time as the interval between the instant at which current reaches 10% of the maximum and the instant at which current rises to 90% of the maximum. The fast response characteristic of the sensor facilitates the real-time monitoring of NO release from cells. The detection limit was calculated to be about 55 pM in PBS and 90 pM in the cell medium (S/N ¼ 3), which are to the best of our knowledge the lowest of those previously reported (Table S1 †). After seeding on the APBA/FGHNs/ITO microelectrode array, the HUVECs grew and proliferated well (Fig. S9 †), with a higher cell density formed on the active microelectrode surface 10 h later. Fluorescence imaging by live/dead cell markers Calcein-AM and PI demonstrated the excellent viability of the cell (Fig. 5b). Then the APBA/FGHNs/ITO microsensor was used to monitor NO release from the living cells (Fig. 5c, black line) cultured thereon. The production of NO was evoked by stimulating HUVECs with fast injection (1 mL) of 3 mM L-Arg, which can be enzymatically oxidated by nitric oxide synthase (NOS) to produce NO in endothelial cells. 52 Control experiments were carried out to confrm the change in the measured current was due to the oxidation of NO released from HUVECs. When L-Arg was injected onto microelectrodes without cells cultured thereon or cells were simultaneously stimulated with a specifc NOS inhibitor, L-NAME (100 mM), and L-Arg, no increase in current was detected (Fig. 5c, blue line and red line for the above two conditions, respectively), excluding the possibilities of L-Arg disturbance and other related electrochemical active interferences. NO release from different numbers of living cells were monitored by the sensor in both PBS (Fig. S10 †) and the cell medium (Fig. 5d). Obvious increases in current were observed, which were followed by a gradual decrease of the current to the baseline, and the current amplitude was raised with the increasing number of cells cultured on the sensor. We measured the average concentration of released NO in the cell medium when all the active surface of the microelectrode was covered by HUVECs, and the quantitative value was calculated to be about 16 nM, in the range of NO released by endothelial NO synthase. 17,53 To test the reusability of the microsensor array after cell culture and NO detection, we detached the attached cells and cell-secreted glycoprotein by frstly immersing it in 0.01 M aqueous NaOH solution followed by rinsing in water. It was observed that the attached cells could be easily detached from the sensor for subsequent cell culture and measurement owing to the pH-dependent reversible formation between boronate and carbohydrates. Electrochemical results showed that the microsensor remained at over 80% of its initial response after cell detection and sensor renewal 5 times, indicating a highly promising strategy for constructing reusable cell biosensors, especially when they are obtained by complicated microfabrication. ## Conclusions In summary, we have constructed a multifunctional NO microsensor array based on boronic acid and metalloporphyrin co-functionalized graphene oxide. The hybrid material FGHNs exhibited excellent sensitivity to a sub-nanomolar range of NO, while APBA conferred the sensor with high cytocompatibility to cells and practicable reusability. The sensor was further used for the sensitive and selective real-time monitoring of NO molecule release from attached human endothelial cells in a cell culture medium. In addition, the sensor was transparent and could be coupled to optical imaging techniques. Though detection of NO at very low bioactive concentrations under in vivo physiological conditions has not yet been demonstrated , the exceptional sensitivity, excellent cytocompatibility and reusability of this sensor make it a promising sensing interface to be incorporated into integrated microfluidic or implantable devices. This can therefore facilitate the real-time monitoring of NO at extremely low concentrations in diverse physiological and pathological conditions with high spatiotemporal resolution.

chemsum

{"title": "Functionalized graphene-based biomimetic microsensor interfacing with living cells to sensitively monitor nitric oxide release", "journal": "Royal Society of Chemistry (RSC)"}

machine_learning_prediction_of_uv–vis_spectra_features_of_organic_compounds_related_to_photoreactive

## Abstract: Machine learning (ML) algorithms were explored for the classification of the UV-Vis absorption spectrum of organic molecules based on molecular descriptors and fingerprints generated from 2D chemical structures. Training and test data (~ 75 k molecules and associated UV-Vis data) were assembled from a database with lists of experimental absorption maxima. They were labeled with positive class (related to photoreactive potential) if an absorption maximum is reported in the range between 290 and 700 nm (UV/Vis) with molar extinction coefficient (MEC) above 1000 Lmol −1 cm −1 , and as negative if no such a peak is in the list. Random forests were selected among several algorithms. The models were validated with two external test sets comprising 998 organic molecules, obtaining a global accuracy up to 0.89, sensitivity of 0.90 and specificity of 0.88. The ML output (UV-Vis spectrum class) was explored as a predictor of the 3T3 NRU phototoxicity in vitro assay for a set of 43 molecules. Comparable results were observed with the classification directly based on experimental UV-Vis data in the same format.The UV-Vis absorption spectrum is a key physical property of an organic compound that determines many of its optoelectronic properties and photochemical reactivity. In the human body, the combined effect of an external chemical compound (e.g., plant toxins, phytomedicines, cosmetics, agrochemicals, food additives, dyes, drugs) and exposure to light, especially ultraviolet and visible radiation may give rise to an acute unwanted response in the skin or retina, which is called chemical phototoxicity 1,2 .The prediction of UV-Vis spectra from the molecular structural formula is of general high interest to design new materials, identify potential phototoxic compounds, estimate missing spectroscopic data for known molecules, or curate databases of experimental spectra.Machine learning (ML) techniques have been reported for the prediction of optical and photophysical properties of organic compounds 3-6 . Joung et al. 3 reported a deep learning model developed with an experimental database of 30 ,094 chromophore/solvent combinations to predict several optical properties, namely, the first absorption peak position, bandwidth, and extinction coefficient, the emission peak position, bandwidth, and photoluminescence quantum yield; and illustrated the possibilities of applying ML to find target molecules with desired optical and photophysical properties. The root mean squared errors of the predicted values were found to be 26.6 and 28.0 nm for absorption and emission peak positions. A comparison between predictions of the absorption and emission spectra of coumarin 153 in ethanol using the ML model and TD-DFT calculations revealed a better performance of the ML model when compared to the theoretical calculations 3 . Another database of experimental and computational UV-Vis absorption spectra attributes was recently obtained through mining methods 7 .ML algorithms can also be trained with theoretically calculated data sets obtained, e.g., by DFT methods, for faster estimation of molecular properties [8][9][10][11] . ML models based on theoretical optical spectra pre-calculated by finite-difference time-domain (FDTD) simulations for gold nanoparticles and nanorods were reported by Pashkov et al. 4 The models were explored both to predict structural parameters for a given spectrum and to predict a spectrum for given structural parameters. Gosh et al. 5 calculated a database of 132 k excitation spectra using the PBE density functional augmented with vdW corrections, and trained neural networks with various architectures to predict the spectra from the 3D structure. Kang et al. 6 used random forests to predict the highest oscillator strength and associated excitation energy among ten excitation states of molecules from 1 and 2D descriptors of the molecule. The model was trained with the TD-DFT results of about half a million molecules. Phototoxicity is strongly related to molecular photochemistry and photostability 2 . The optimization of ADME-Tox parameters (absorption, distribution, metabolism, excretion, and toxicology) using high-throughput tools is of great importance in drug discovery 12 , and ML approaches can be used to rationalize and predict phototoxicity, representing a valuable strategy for reducing experimental tests, if an acceptable level of accuracy of the developed models is ensured. Although several efforts have been reported to model phototoxicity directly from molecular structures , the inclusion of spectroscopic information can improve predictive models 2 and add chemically sound indicators that can be theoretically calculated or learned from more easily available and larger data sets. Training ML models to predict full UV-Vis spectra requires large databases of spectra obtained under consistent conditions to predict multiple continuous variables (e.g., the molar extinction coefficients at several wavelengths). Differently, here we report the exploration of ML tools to classify organic molecules in terms of their UV-Vis absorption spectrum based on molecular descriptors. The relationship between features of the UV-Vis absorption spectrum ("photoactivity") and phototoxicity can be clearly understood from the ICH S10 guidance on photosafety evaluation of pharmaceuticals 18 , according to which a molecule is potentially photoreactive if it absorbs light in the range between 290 and 700 nm (UV/ Vis) with molar extinction coefficient (MEC) greater than 1000 L•mol −1 •cm −1 . Excitation of molecules by light can lead to generation of reactive oxygen species and this can be an indicator of phototoxicity potential 18 . If the substance does not have a MEC above 1000 L•mol −1 •cm −1 in the above-mentioned window no direct phototoxicity is anticipated in humans 18 . We retrieved data from the Reaxys ® (https:// www. reaxys. com) database 19 for > 80,000 molecules, and positive/ negative classes related to photoreactive potential were assigned from the lists of absorption maxima and molar extinction coefficients with threshold values based on the ICH S10 guidance. An external data set of molecules for which data was available both for UV-Vis absorption 19 and for in vitro phototoxicity assays 2 was used to evaluate the overlap of correlations between a phototoxicity test and the UV-Vis spectrum class (experimental or predicted by ML). However, we would like to emphasize that this study aimed at training ML models to predict features of the UV-Vis spectra from the molecular structure, rather than predicting phototoxicity or evaluating the usefulness of spectroscopic data to predict phototoxicity. ## Methods Data sets/selection of training and test sets. Molecular structures were retrieved from the Reaxys ® database (https:// www. reaxys. com) 19 with associated UV-Vis absorption maxima and molar extinction coefficient (MEC) values and were filtered for molecular weight in the range 98-1080 g/mol, only one fragment, methanol as the solvent, exclusion of molecules with metal atoms, and restriction to publication date before 2016. The molecular structures were standardized by normalizing tautomerism, mesomerism and aromaticity using the Standardizer program version 19.19.0, ChemAxon (https:// www. chema xon. com). Duplicates were removed based on InChI identifiers and stereochemistry was not considered so that stereoisomers were considered as duplicates. Compounds with a non-zero global charge, radicals or valence errors were also discarded. The final data set comprises 74,784 molecules: 37,038 molecules assigned to the positive class (POS, molecules with one or more absorption maxima between 290-700 nm with MEC ≥ 1000 Lmol −1 cm −1 ) and the remaining 37,746 molecules assigned to the negative class (NEG). The definition of the classes was based on the ICH S10 guidance on photosafety evaluation of pharmaceuticals 18 . The data set was randomly divided into a training set of 72,788 molecules (POS class: 36,036 molecules and NEG class: 36,752 molecules), a test set I of 998 molecules (POS class: 501 molecules and NEG class: 497 molecules), and a test set II of 998 molecules (POS class: 512 molecules and NEG class: 486 molecules). The test set II includes 43 molecules for which the result of the 3T3 NRU phototoxicity in vitro assay is also available from Schmidt et al. 2 Table 1 shows the distribution of UV-Vis absorption features in the data sets. ## Calculation of molecular descriptors and fingerprints. Molecular fingerprints and 1D&2D molecular descriptors were calculated with PaDEL-Descriptor version 2.21 20 , and RDKit 21 . Different types of fingerprints with different sizes were calculated and explored: 166 MACCS (MACCS keys), 307 Substructure (presence and count of SMARTS patterns for Laggner functional group classification-Sub and SubC respectively), 881 PubChem fingerprints 22 , 1024 CDK (circular fingerprints), 1024 CDK extended (circular fingerprints with additional bits describing ring features), and 1024 MorganFP 23 . The 1D&2D molecular descriptors comprise 1443 descriptors, including electronic, topological, and constitutional descriptors. Modified Distance Descriptors (Md) 10 are based on the molecular connectivity thus making no use of bond orders and atomic formal charges (avoiding the generation of a 3D conformer, the application of an aromaticity definition and the mesomerism standardization). The descriptors count the pairs of atoms in a molecule at specific "modified distances". Modified distances were defined in terms of the van der Waals radius of the atoms and Sanderson electronegativity of neighbors. Md descriptors were calculated for 1010 intervals, using a resolution of 0.017, interatomic distances up to 4 bonds, and a distance factor of 4. Estimated molecular orbital energies (E HOMO , E LUMO and GAP) were obtained with previously in-house developed ML models trained with DFT calculated data 10 -ML quantum descriptors (ML QD ). They include 10 values obtained for the three properties with different models. The calculators for some types of descriptors/fingerprints did not process all molecules, and the corresponding training sets have slightly different sizes: 72,787 for RDKit and RDKit Morgan fingerprints; 72,771 for MACCS, Sub, SubC and PubChem fingerprints; 72,747 for 1D&2D molecular descriptors; 72,770 for ML QD descriptors. ## Selection of descriptors. In the quest for QSPR models with reduced number of descriptors, descriptor selection was performed based on the importance of descriptors assessed by RF (mean decrease in accuracy measure) with the R program version 3.6.1. 24 . Machine learning (ML) methods. Classification and Regression Trees (CART) 25 operate by recursive partitioning of the initial data set with the goal of maximizing an information gain function (or variance reduction in regression trees) calculated for the various branches and terminal nodes. The best tree size is identified among sub-trees by cross-validation, or splits are not attempted if improvements above a threshold are not attained. Classification trees were built using the rpart package 26 of the R program version 3.6.1 24 with default parameter values, except for 1D&2D descriptors (the cp parameter was set at 0.05). Random forests (RF) 27 were implemented as ensembles of unpruned classification trees which are grown using bootstrap samples of the training set. Each individual tree is different because bootstrap samples vary, and randomly selected subsets of descriptors are made available for each node split. Predictions are obtained by a majority vote of the classification trees in the forest. An internal cross-validation error (or out-of-bag estimation, OOB) is directly calculated with the prediction error for the objects left out in the bootstrap procedure. The importance of a descriptor can be assessed by the mean decrease in accuracy when the values of the descriptor are randomly permuted. A probability is assigned to every prediction based on the number of votes obtained by the predicted class. RFs were grown with the R program 24 , version 3.6.1, using the randomForest library 28 . The model was manually optimized recurring to the OOB estimation, with the number of trees from 500 to 1000 and the number of available descriptors for each rule (mtry) equal to the square root or 1/3 of the total number of descriptors. Support Vector Machines (SVM) 29 map the training data into a hyperspace through a nonlinear mapping (a boundary or hyperplane) and then separate the classes of objects in this space. The boundary is positioned using examples in the training set-the support vectors. Kernel functions can be used to transform data into a hyperspace where the classes become linearly separable. In this study, SVM were implemented with the program Weka version 3.8.3 30 , using the LIBSVM package 31 . The type of SVM was set to C-SVM-classification and the radial basis function was used for the kernel function. Hyper-parameter tuning was performed with the Experimenter application in Weka using ten-fold cross-validation. C and γ values varied from 1 to 1000 and 0.003 to 0.0045, respectively. The C and γ values were finally set to 500 and 0.004, respectively, and the other parameters were used with default values. Deep Learning Multilayer Perceptron Networks ( d MLP) were trained and applied with the software library Keras 32 version 2.2.5 based on Tensorflow numerical backend engine 33 . The feed-forward neural network architecture was manually optimized in terms of the number of hidden layers (2 to 6), weights initializer (random normal and Glorot uniform), optimizer (adadelta, adam, SGD, and SGD-nesterov), learning rate (0.0001 to 0.01) and decay (0 to 0.01). The final hyper-parameter settings were selected for our study based on a best of 10 validation experiments with the training set-Table 2. Statistical measures of models' performance. Models were evaluated with external test sets and by internal validation with the training set. An OOB estimation for RF and a tenfold cross-validation for the other ML techniques procedures were employed with the training set. The following measures were calculated: true positives (TP), true negatives (TN), false positives (FP), false negatives (FN), sensitivity (SE), specificity (SP), overall predictive accuracy (Q) and Matthews correlation coefficient (MCC) were calculated with Eqs. ( 1)-( 4). ## Results and discussion Machine learning prediction of UV-Vis photoreactive potential. Several ways of representing the molecular structures were evaluated as input to RF classification models, which were trained to predict the UV-Vis spectrum class. The number of trees in the forest was set to 500, the number of descriptors available for each rule was the square root of the total number of descriptors and the other parameters were used with default values. The performance of the models was evaluated by internal validation with the training set (out-of-bag estimation, OOB) and by validation with test set I (998 molecules)-Table 3. (1) www.nature.com/scientificreports/ The models with CDK, ExtCDK, and RDKitMorganFP showed the best overall predictive accuracy for the training set in the OOB estimation and similar predictions were achieved for the test set. Both CDK and ExtCDK representations yielded slightly higher sensitivity than RDKitMorganFP, but the latter enabled the highest specificity and MCC of all models. It is also worth mentioning that the few ML QD descriptors alone provided results that, although worse, are still good. The complementary potential of several molecular representations was investigated next by combining Md, ExtCDK, RDKitMorganFP, 1D&2D and ML QD molecular descriptors/fingerprints. The criteria for generating the combinations were the complementary nature of the attributes and, in case of similar sets, those yielding better predictions individually. The results in Table 4 show that combined descriptors did not provide any significantly superior model. The impact of random fluctuations in the models was assessed by re-training with different seed initialization of random functions. Fluctuations of Q up to 2% were observed meaning that differences below this value cannot be considered significant. The best models were also validated with Y-scrambling experiments in which the percentage of correct predictions varied between 49.85% and 50.1% in 5 experiments. The RDKitMorganFP model achieved the best results for the test set with a Q of 0.88 and a MCC of 0.76 (Table 3). The 250 most important fingerprint bits of the RDKitMorganFP were identified by the RF model for the training set and were selected for training a new RF model with lower computational cost, as well as other models with different ML algorithms (SVM and d MLP)-Table 5. Reduction of attributes yielded a RF model with essentially the same quality. The receiver operating characteristic curve (ROC) obtained for the test set I with the RF model trained with 250 RDKitMorganFP attributes is displayed in Fig. 1. Superior results could not be observed with the alternative ML algorithms. The best model (RF 250 RDKitMorganFP) was further validated with the independent test set 2. The statistical parameters of the obtained predictions are in line with those obtained for the first test set. Models trained with CDK and ExtCDK fingerprints (default number of features) were also evaluated (Table 6). ## Analysis of important molecular attributes. Interpretable molecular descriptors (1D&2D and ML QD ) and fingerprints (MACCS, PubChem and Sub) were processed with machine learning algorithms to provide information on relevant structural features for the UV-Vis spectrum classification. MACCS, PubChem and Sub fingerprints are binary attributes that encode the presence or absence of a particular structural feature. The importance of attributes calculated by RF was inspected. Additional to RF, simple classification trees were grown to understand relationships between individual attributes and the potential photoreactivity of molecules. As expected, the trees are poorer predictive models than the more complex RF (Table 7) but are useful to analyze www.nature.com/scientificreports/ the importance of molecular fragments. The classification tree trained with PubChem fingerprints is shown in Fig. 2, and the trees obtained with the other attributes are in Figures S1-S4 of the Supplementary Material. In the tree obtained with PubChem fingerprints (Fig. 2) the first two rules use the presence of two conjugated C=C double bonds and the α,β-unsaturated carbonyl in fragment "O=C-C=C-C" as discriminant features. Additionally, the presence of aromatic fragments with nitrogen or oxygen substituents ("N-C:C:C-C" and "C(~O) (:C)(:C)") were also used. The features selected by the tree grown with MACCS fingerprints were similar and operated in similar ways. Features encoding the presence of aromaticity, double bonds connected to nitrogen atoms, as well as the absence of carbon atoms with at least two single bonds and at least two hydrogens were associated with the positive class. Most of the inferred rules classify compounds as POS due to the presence of specific types of sub-structures, which agrees with the chemical knowledge that chromophores give rise to UV-Vis absorption. However, some rules associate the POS class with the absence of some aliphatic fragments. An example is the presence of a tertiary carbon atom; although it does not preclude the presence of chromophores in the molecule, a tertiary carbon atom is not involved in conjugation and is statistically associated with the negative class in our data set. The 1D&2D molecular descriptors enabled a tree (in Figure S3) to infer two powerful rules based on the number of atoms in the largest pi system. They discriminate molecules with extended conjugation and fused www.nature.com/scientificreports/ bicyclic rings derived from e.g., naphthalene, indole, benzimidazole, benzoxazole, benzofuran, benzothiophene, or benzazepine. An additional rule is based on features of nitrogen atoms with two aromatic bonds. Based exclusively on the estimated energies of the HOMO and LUMO orbitals, and their gap, a single rule was established that associates the POS class to GAP < 4.626 eV. Inspection of the database revealed that the molecules with the lowest value for this descriptor (~ 2 eV) include highly conjugated aromatic systems, such as tetracarboxdiimide derivatives, also corresponding to visible light absorption in the range between 500 and 700 nm with high MEC values (> 1000 Lmol −1 cm −1 ). The ten most important attributes according to the MeanDecreaseGini parameter in the RF models are reported in Tables S1-S4 of the Supplementary Material for various fingerprints and descriptors. The most relevant attributes identified by the RF models are in line with those selected to build the trees, namely attributes accounting for the presence of aromaticity, unsaturation and conjugated systems. The importance of conjugation was highlighted by the selection of three Sub fingerprints that encode the presence of α,β-unsaturated carbonyl or carboxyl groups. The presence of tertiary carbon atoms is at the top ten in three models (MACCS, Sub and PubChem-Tables S1-S3). This feature was also used in the tree of Fig. 2 associated with the negative class. Although some of the 1D&2D descriptors have no straightforward interpretation, it is clear that different aspects of unsaturated systems are encoded by several of the most important attributes: number of atoms in the largest pi system, ratio of total conventional bond order with total path count, fraction of sp 3 carbons to sp 2 carbons, measure of relative unsaturation content, total number of bonds that have bond order greater than one. Furthermore, measures of global electronic features appear as highly relevant in positions 8 and 9. ## Analysis of outliers. The RF model trained with all RDKitMorgan fingerprints predicted the test set I with accuracy of 0.88 and MCC 0.76. The ROC curve of Fig. 1 illustrates the significance of the probabilities assigned by the RF models to the predictions. Among the 998 predicted molecules, the 15 FP and 18 FN with a probability higher than 0.8 were manually inspected to discover possible reasons for wrong predictions with high assigned probabilities. Most false negatives (12 out of 18 FN) correspond to molecules with peak wavelengths inside the photoreactivity window, but close to the lower endpoint of the interval (290-317 nm). Other 2 FN have a peak within the window but with a MEC value lower than 1500 Lmol −1 cm −1 . The other 4 FN are compounds with peaks inside the window and high values of MEC reported in the database: a 16-membered macrolide with 3 deoxy sugar moieties attached and including a pi system (1) 34 , two cyclic compounds with conjugated systems (2 35 and 3 36 ) and a quinazolinone connected to a thiazole (4) 37 , Fig. 3. Three of these four predictions could be explained by similar molecules in the training set, which were assigned to the NEG class based on the experimental data. The structures 1, 2, 3 and 4 in Fig. 3 were subjected to a similarity search against the training set, using fingerprints and Tanimoto coefficients. Molecules 1, 2 and Although compound 5 is correctly assigned to the NEG class based on the experimental data, according to our definition of the POS and NEG classes (it has no maxima in the relevant window with MEC above the threshold), it represents in fact a borderline situation because UV-Vis absorption peaks are typically broad. Compound 5 has a peak at 282 nm with a MEC value of 19,400 Lmol −1 cm −1 ; it is therefore highly probable that the MEC value at 290 nm is higher than 1000 Lmol −1 cm −1 suggesting photoreactive potential. This example highlights a limitation of the models due to the nature of the experimental data here used (consisting in lists of UV-Vis absorption maxima): the absence (in the list) of a maximum within the relevant window does not guarantee that there is no absorptivity in the window with a MEC above the threshold. It may happen that the experimental spectrum did not cover the full window or the source publication reported only the highest peak(s), and there are often absorptivities in the window with a MEC above the threshold from bands whose maxima are outside the window. The similarity of compound 6 (NEG) to compound 2 may explain the prediction of the latter as negative; but inspection of the original source for compound 2 reveals that the reported UV-Vis data are for a www.nature.com/scientificreports/ metal complex and shall not be considered for structure 2. This case illustrates how the method can be useful for the curation of experimental databases. Compound 8 may explain the false negative prediction for compound 4; the inclusion of an additional chlorine substituent in the aromatic ring added a new absorption band at a higher wavelength 37 , and the ML model apparently did not learn that effect. Concerning the 15 FP, it was observed that 4 molecules are among those in the NEG class with a peak at a wavelength only slightly below 290 nm (281-289.5 nm) and with a high MEC value. Other 3 FP have peaks at wavelengths between 269 and 277 nm with MEC values between 9000 and 28,183 Lmol −1 cm −1 . Two FP have a peak with both the wavelength and MEC very near the thresholds (wavelength 295-306 nm, MEC 661-891 Lmol −1 cm −1 ). For the other 4 FP, all the MEC values listed in the database are between 3 and 5, which suggests they were originally reported in a different unit or as log(MEC)-confirmation was possible for at least one of them that a log(MEC) value was retrieved from the original literature as MEC and the compound exhibits indeed significant absorption within the 290-700 nm window. Finally, the other 2 FP arise from a situation similar to compounds 4 and 8: a similarity search against the training set revealed that the inclusion/changing of a substituent in the aromatic system (in these cases to include methoxy and amino groups) is associated with the reporting of an absorption maxima at higher wavelengths. UV-Vis spectrum classification as a predictor of in vitro phototoxicity. For 43 molecules of test set II, additional data is available concerning the 3T3 NRU in vitro phototoxicity assay (PIV) 2 . The UV-Vis spectrum class was predicted for this subset with global accuracy 0.86, sensitivity 0.96 and specificity 0.69 (comparing to 0.89, 0.90 and 0.88, respectively, for the whole test set II-Table 6). The RF output (UV-Vis spectrum class) was evaluated as a predictor of phototoxicity. Similarly, predictions of phototoxicity were also obtained using the lists of peaks and their MEC values available in the database of experimental data ("experimental spectrum class"). This enables to compare two approaches to the assessment of phototoxicity: (a) classification of the UV-Vis spectrum from lists of peaks available in the chemical literature, and (b) machine learning prediction of the spectrum class from the molecular structural formula. The confusion matrices are in Table 8. The two confusion matrices are quite similar. The RF classification would correctly estimate 25 out of 43 molecules comparing to 27 using the experimental data. All the 12 non-toxic molecules assigned to the positive class in the experimental database were also predicted as positive by the RF model, and two toxic molecules were classified as negative both in the database and in the RF predictions. This suggests that a RF classification model for UV-Vis absorption features can assist in the estimation of in vitro toxicity similarly to experimental UV-Vis data. However, it must be emphasized that this study was not based on full spectra, but on lists of peaks extracted from the literature with their inherent incompleteness, and they certainly include noise. The large number of non-toxic molecules assigned to the positive class is a limitation of UV-Vis spectra as a predictor of phototoxicity, because other characteristics of a chemical compound, beyond light absorption, are critical for phototoxicity, namely the ability to generate a reactive species. In any case, from the perspective of photosafety evaluation, high sensitivity is more important than specificity since molecules predicted as positive would be subjected to further experimental tests. ## Conclusion The random forest algorithm, trained with 72,787 organic molecules represented by Morgan circular fingerprints, was able to classify molecules according to UV-Vis spectrum features related to photoreactive potential with accuracy up to 0.89, sensitivity of 0.90 and specificity of 0.88 for an independent test set of 998 molecules. The classes in the training and test sets were assigned based on data retrieved from the chemical literature, which consists of lists of absorption maxima with molar extinction coefficients. Application of machine learning algorithms with interpretable molecular descriptors and fingerprints provided information on relevant structural features for the classification. The rules inferred by classification trees and the importance of attributes calculated by RF revealed that aromaticity, unsaturation, conjugation, and heteroatom substituents play an important role in discriminating between positive and negative classes. Analysis of outliers (wrongly predicted molecules with high associated RF probability) highlighted three main situations: (a) absorption maxima with wavelengths near the lower endpoint of the established interval (290-700 nm) and/or MEC values close to the established threshold; (b) data noise, e.g., retrieval of log(MEC) value instead of MEC value; (c) insufficient learning of the impact of some heteroatom substituents on the absorption maxima. www.nature.com/scientificreports/ The ML assignment of molecules to the positive class (related to photoreactive potential) was a predictor of a positive outcome of the 3T3 NRU phototoxicity in vitro assay with a sensitivity of 0.84 and specificity of 0.38 in a test set of 43 molecules. Comparable results were observed with the assignment based on the experimental data available for the same set (sensitivity 0.79 and specificity 0.5). The results illustrate the potential of machine learning algorithms for the classification of molecules according to the UV-Vis absorption spectrum, to assist in photosafety evaluation.

chemsum

{"title": "Machine learning prediction of UV\u2013Vis spectra features of organic compounds related to photoreactive potential", "journal": "Scientific Reports - Nature"}

complementary_oligonucleotides_regulate_induced_fit_ligand_binding_in_duplexed_aptamers

## Abstract: Duplexed aptamers (DAs) are engineered by hybridizing an aptamer-complementary element (ACE, e.g. a DNA oligonucleotide) to an aptamer; to date, ACEs have been presumed to sequester the aptamer into a non-binding duplex state, in line with a conformational selection-based model of ligand binding. Here, we uncover that DAs can actively bind a ligand from the duplex state through an ACE-regulated induced fit mechanism. Using a widely-studied ATP DNA aptamer and a solution-based equilibrium assay, DAs were found to exhibit affinities up to 1 000 000-fold higher than predicted by conformational selection alone, with different ACEs regulating the level of induced fit binding, as well as the cooperative allostery of the DA (Hill slope of 1.8 to 0.7). To validate these unexpected findings, we developed a nonequilibrium surface-based assay that only signals induced fit binding, and corroborated the results from the solution-based assay. Our findings indicate that ACEs regulate ATP DA ligand binding dynamics, opening new avenues for the study and design of ligand-responsive nucleic acids. ## Introduction Proteins and functional nucleic acids often couple specifc ligand-binding events with distinct structural changes central to biochemical regulation. 1,2 Likewise, synthetic nucleic acid aptamers undergo distinct structural changes upon ligand binding, which can be leveraged for biosensing and applications in synthetic biology. The binding-related structural changes that occur in these ligand-binder systems can be described as proceeding via either conformational selection (MWC model 10 ) or induced ft (KNF model 11 ) (Fig. 1). In conformational selection, a binder exists at equilibrium between non-binding and binding-competent states, and is only capable of binding the ligand in the latter. Alternatively, induced ft describes a binding pathway from the non-binding to bound state actively catalyzed by the ligand. Interestingly, aptamers can be engineered with enhanced switching activity by hybridizing an aptamer-complementary element (ACE, such as a short DNA oligo) to a desired aptamer sequence, forming a duplexed aptamer (DA) that acts as a synthetic switch. The ease of engineering DAs from known aptamer sequences has led DAs to fnd numerous applications based on e.g. FRET, 12 electrochemistry, colorimetry, 23 SPR, 24 fluorescence, 25,26 and signaling cascades. 27 However, to date, ligand binding in DAs has only been modeled based on conformational selection, in which the ACE acts as an inhibitor, sequestering the aptamer into a nonligand-binding, passive duplex state. In this model, the observed affinity of a DA (K Obs d ) is a function of (i) the intrinsic affinity of the native aptamer (K Apt d ) and (ii) the hybridization free energy of the ACE-aptamer duplex (K Hyb ). In qualitative agreement with such a model (for details, see Fig. S1 †), Porchetta et al. used a native cocaine DNA aptamer and varying length ACEs (10-15 bases) to engineer and tune the relative affinity of cocaine DAs over three orders of magnitude. 32 To our knowledge, although concepts of 3-body side reactions and misfolded sensor states have been used to model DAs, 28 induced ft ligand binding in DAs, in which a DA might actively sense and Fig. 1 Ligand binding pathways in nature. In conformational selection, the binder shifts between distinct conformational states, and the ligand only stabilizes a pre-existing binding competent state. Induced fit describes a binding pathway in which the ligand catalytically reorganizes the binder into a favorable conformation, leading to the ligand bound state. catalytically bind a ligand directly from the duplex state, has not been studied. In this regard, we note that small modifcations to the length and location of ACEs have been documented to impact DA biosensors in ways that cannot be accounted for by differences in ACE-aptamer hybridization free energies. Here, we evaluate the possibility of induced ft ligand binding in DAs. We focused on DAs engineered from the Huizenga and Szostak ATP DNA aptamer introduced in 1995, 40,41 as this aptamer is the most widely studied, is well characterized, and was the frst to be implemented as a DA. 12 The native aptamer binds ATP (6 mM K Apt d ) frst through folding of the stem and loop regions, followed by the cooperative binding of two ATP molecules within the binding pocket (sites I and II, Fig. 2). Using DAs engineered from the ATP aptamer, we frst performed equilibrium solution-based assays and uncovered (i) the existence of induced ft ligand binding in ATP DAs, and (ii) that ACEs allosterically regulate induced ft ligand binding, thereby modulating the affinity of ATP DAs in an unexpected manner. To confrm these fndings, we performed a second set of experiments using a non-equilibrium surface-based assay that we developed. ## Results and discussion For the solution-based assay, ATP DA constructs were designed starting with a 32-mer Cy3-labeled variant of the ATP aptamer, together with a previously reported 12-mer ACE, 3 0 -labeled with a BHQ-2 quencher. 12 This ACE hybridizes to 12 bases at the 5 0 end of the aptamer, beginning fve bases outside the consensus aptamer sequence at the -5C nucleotide, and is termed 5 0 Q -5C:12 . Representative states in the candidate binding pathways for the 5 0 Q -5C:12 DA are shown in Fig. 2. To assess any impact of ACE length and location on DAs, we engineered DAs using successively 5 0 -truncated 5 0 Q -5C:12 ACEs, as well as a three-base 3 0truncated ACE (Fig. 3a). After confrming duplex formation for all fve ACEs (Fig. S2 †), ATP DAs were tested using an equilibrium solution-based FRET assay 29,32 (see ESI methods †), with ACE:aptamer (Q:F) ratios of 1:1 (Fig. 3b) and 3:1 (Fig. S3 †). For a 1:1 Q:F ratio, the experimentally observed affinities of the 5 DAs are poorly predicted using a conformational selectionbased analytical model of DAs (Fig. 3a and c). Based on experimentally measured hybridization DA free energies (obtained using FRET melting, 44 see ESI methods, Table S1 and Fig. S5 †), only the 5 0 Q -5C:9 data is consistent with a conformational selection model (Fig. 3c; K C.S. d , K Apt d ¼ 6 mM solid isoline), while the four other DAs tested are not. The analytical model underestimates 5 0 Q -5C:12 affinity by more than 1 000 000-fold, and predicts a 15-fold increase in apparent affinity for 5 0 Q -5C:9 over 5 0 Q -5C:10 , whereas a 28-fold decrease in affinity was observed (Fig. 3a and c). An analytical model incorporating an induced ft binding pathway alongside conformational selection (and in which each Additionally, the ACEs tested gave rise to DAs with differing ATP binding cooperativities, as determined based on the Hill slope, n, suggesting that ACEs also modulate DA allostery. Here, reduced interaction of ACEs with site II promoted a shift from cooperative to anti-cooperative binding for Q:F ratios of 1:1 (n ¼ 1.3 to 0.7) and 3:1 (n ¼ 1.8 to 0.9) (Fig. 3a, b and S3 †). Interestingly, positive cooperativity of the DA was restored using the site II-hybridizing 5 0 Q -2T:9 ACE (n ¼ 1.3 and 1.7 for 1:1 and 3:1 Q:F ratios), suggesting that ACEs hybridized to site II yield DAs capable of cooperative ligand binding, as present in the native aptamer (n ¼ 2.0 (ref. 42)). To verify the unexpected fndings obtained from the solution-based assay, we developed a surface-based assay that signals only when a DA binds a ligand via induced ft, and in which conformational selection plays no role (Fig. 4). In contrast with the solution-based FRET assay, the surface-based fluorescence assay does not operate at equilibrium. Here, fluorophore-conjugated aptamers are frst hybridized to ACEs covalently coupled on a slide surface, followed by washing off of non-hybridized aptamers, yielding surface-immobilized DAs (Fig. 4). After incubation with buffer (or buffer and ligand) for a specifed time (Dt), DA dehybridization is measured as a loss of surface fluorescence (DF Obs Rel ); owing to the low concentration of released aptamers, DA dissociation is effectively non-reversible. By assaying varying ligand concentrations, the surfacebased fluorescence assay can be used to derive the induced ft affinity (K Fit ) of a DA. We used this surface-based assay to study ATP DAs constructed from three ACEs, corresponding to three ACEs used in the solution-based FRET assay (but missing quenchers). These ACEs varied in length and degree of site II hybridization to the ATP aptamer, termed 5 0 -5C:12 , 5 0 -5C:9 , and 5 0 -2T:9 (Fig. 5a). A 3-plex microarray was constructed with the 3 ACEs, incubated with Cy3-labeled aptamer, briefly washed, and imaged with Next, sub-arrays on the microarray were incubated for 1 h each with ATP in buffer, or with buffer only, followed by a second fluorescence scan (Fig. 5b). Given the high concentrations of ATP assayed, the low dissociation rate of modifed duplexes ðk * off Þ expected for the ACEs tested here (Table S1 †), and assuming a steady state of intermediate duplexes on the surface, this assay can be modeled by Briggs-Haldane kinetics. By also assuming k * off ( k rev (Fig. 4), the experimental induced ft binding affinity of a DA (K Fit,Exp ) is equal to the Michaelis constant derived from DF Obs Rel with ATP titration (Fig. 5c). The surface-based fluorescence assay yielded a high induced ft binding affinity for the site-II hybridizing 5 0 -5C:12 ATP DA (K Fit,Exp of 67 mM) (Fig. 5a and c). However, the 5 0 -5C:9 DA, which shares a footprint with 5 0 -5C:12 but with site II left unhybridized, displayed no induced ft binding (K Fit,Exp > 10 mM, Fig. 5a and c), consistent with the solution-based FRET assay fndings (Fig. 3a, c and S4 †). Thus, despite having a much lower hybridization free energy than the 12-mer ACE, the 9-mer 5 0 -5C:9 ACE does not promote induced ft. Meanwhile, DAs engineered with the site-II hybridizing 5 0 -2T:9 ACE displayed a K Fit,Exp of 263 mM (Fig. 5a and c). This value is in good agreement with an Rel /Dt) following incubation with buffer only. In parallel, the increase in DF Obs Rel /Dt arising from incubations with increasing ligand concentrations is measured, from which the induced fit binding affinities of DAs (K Fit,Exp ) can be calculated. analytical model of the solution-based assay including both conformational selection and induced ft binding pathways (Fig. 3c and S4 †). As a negative control ligand, we also assayed microarrays with GTP, and no induced ft binding was observed (Fig. S6 †). As observed in the solution-based FRET assay for site-II hybridizing ACEs, the 5 0 -5C:12 ACE also formed a DA with positive cooperativity for ATP (n ¼ 1.8) (Fig. 5a and c). Interestingly, the 5 0 -2T:9 DA displayed no cooperativity (n ¼ 0.7); given the unstable duplex expected for an ATP-disrupted 5 0 -2T:9 DA (Table S1 †), this result may indicate that a single ATP is sufficient to displace the 5 0 -2T:9 ACE in the surface-based assay. Overall, the surface-based assay results corroborate the solution-based assay fndings, supporting an ACE-dependent induced ft binding mechanism in ATP DAs. These fndings also suggest that the ACE-based allosteric regulation of ATP DAs is relatively independent of biosensor design. ## Conclusions Taken together, our results indicate that induced ft ligand binding can be a dominating pathway in DAs, and that single nucleotide changes to ACEs can signifcantly modify ATP DA induced ft binding dynamics. These fndings point towards ACEs as underappreciated functional regulators of the binding affinity, binding cooperativity, and allostery of DAs. This work opens new avenues for tuning aptamer-based systems, with particular relevance to the use of aptamers in biosensing and synthetic biology. However, it is not yet clear if induced-ft ligand binding in DAs is rare, with the family of ATP DAs studied here being an exception to the rule, or if induced ft is a general binding pathway common to duplexed ligandresponsive nucleic acids. In this regard, the non-equilibrium surface-based fluorescence assay introduced here could be expanded to investigate additional ACEs, to test other ACEaptamer combinations, to investigate the effect of changes to structural regions of an aptamer on DAs (such as modifying the ligand binding pocket or tertiary structure-stabilizing bases), or to investigate other functionalities played by ACEs in DAs, such as perturbing ligand specifcity. 45,46 Finally, this work also highlights the ACE-specifc regulation of DA ligand binding as a novel model of the thermodynamic and structural determinants that govern transitions between ligand binding pathways. In this sense, we note that DAs may offer researchers a uniquely tractable and confgurable nucleic acid-based alternative to existing proteinbased models of allosteric regulation and collective motion in biopolymers.

chemsum

{"title": "Complementary oligonucleotides regulate induced fit ligand binding in duplexed aptamers", "journal": "Royal Society of Chemistry (RSC)"}

a_bright_fit-pna_hybridization_probe_for_the_hybridization_state_specific_analysis_of_a_c_→_u_rna_ed

## Abstract: Oligonucleotide probes that show enhanced fluorescence upon nucleic acid hybridization enable the detection and visualization of specific mRNA molecules, in vitro and in cellulo. A challenging problem is the analysis of single nucleotide alterations that occur, for example, when cellular mRNA is subject to C / U editing. Given the length required for uniqueness of the targeted segment, the commonly used probes do not provide the level of sequence specificity needed to discriminate single base mismatched hybridization. Herein we introduce a binary probe system based on fluorescence resonance energy transfer (FRET) that distinguishes three possible states i.e. (i) absence of target, (ii) presence of edited (matched) and (iii) unedited (single base mismatched) target. To address the shortcomings of read-out via FRET, we designed donor probes that avoid bleed through into the acceptor channel and nevertheless provide a high intensity of FRET signaling. We show the combined use of thiazole orange (TO) and an oxazolopyridine analogue (JO), linked as base surrogates in modified PNA FIT-probes that serve as FRET donor for a second, near-infrared (NIR)labeled strand. In absence of target, donor emission is low and FRET cannot occur in lieu of the lacking co-alignment of probes. Hybridization of the TO/JO-PNA FIT-probe with the (unedited RNA) target leads to high brightness of emission at 540 nm. Co-alignment of the NIR-acceptor strand ensues from recognition of edited RNA inducing emission at 690 nm. We show imaging of mRNA in fixed and live cells and discuss the homogeneous detection and intracellular imaging of a single nucleotide mRNA edit used by nature to post-transcriptionally modify the function of the Glycine Receptor (GlyR). ## Introduction Fluorogenic oligonucleotide probes are invaluable molecular tools for detecting and localizing RNA molecules inside live cells. These probes bind a complementary nucleic acid strand via Watson-Crick-type recognition, which elicits a fluorescence response that distinguishes target-bound from unbound molecules. A key challenge in RNA imaging is the detection of single base alterations such as C / U or A / I RNA editing; a mechanism used by cells for the posttranscriptional regulation of gene expression. 15 For applications in cells, oligonucleotide probes must exceed a certain length (typically 18 nt) in order to assure the uniqueness of the recognized target segment. However, at this length the commonly used oligonucleotide probes will bind the RNA target regardless of a single base mismatch. In theory, binary probe formats should provide for high target specifcity, because the fluorescence signal depends on the simultaneous binding of two oligonucleotide probes. 16 A commonly used approach involves two fluorescence labeled probes, which interact via fluorescence energy transfer (FRET) when adjacent hybridization brings the donor and acceptor dye in proximity. 7,17 In practice, the achievable signal-tobackground ratios are limited. The efficiency of the FRET process is high when donor emission overlaps with acceptor absorption. Given that typical Stokes shifts are smaller than 50 nm, a bright "FRET signal" (acceptor emission upon donor excitation) requires a rather narrow spectral gap between the two dyes. However, the commonly applied organic dyes have rather broad emission bands and in many instances, the donor signal will bleed through and become apparent in the acceptor channel despite the absence of target. This reduces signal-tobackground. To address this issue, the spectral gap between donor and acceptor emission was increased. This, however, affects the brightness of the FRET signal. Alternatively, a three-dye system has been used, in which one dye serves as a FRET relay. 18 Herein, we describe a binary probe concept that allows avoidance of bleed through by using a dual labeled hybridization probe that is weakly emissive in the absence of target, yet becomes fluorescent and acts as donor for FRET upon target binding. The approach bears resemblance to the dual molecular beacon method introduced by Bao. 21 Our method, however, provides a large 190 nm apparent Stokes shift and, for the frst time, distinguishes all of the three possible statesthat is (i) absence of RNA target, and presence of (ii) C / U edited RNA or (iii) unedited RNA target (Fig. 1). To demonstrate the usefulness of the probes and the high target specifcity we show imaging of mRNA in fxed and live cells and discuss the intracellular imaging of a single nucleotide mRNA edit used by nature to post-transcriptionally modify the function of the Glycine Receptor (GlyR). 15,22 ## Results and discussion In the pursuit of a probe set that allows the detection of the RNA target in both edited and unedited state, we considered the use of a fluorogenic hybridization probe as FRET donor. In an ideal scenario, the spectral gap between donor and acceptor emission should be larger than the width of the donor emission band. This however affects the efficiency of FRET. To compensate for the inevitable loss of the FRET signal intensity, donor emission should be bright. We fgured that the desired property i.e. a bright and hybridization-responsive probe is in the reach of our recently developed TO/JO FIT-probes which contain the highly responsive thiazole orange (TO) dye and the highly emissive oxazolopyridine (JO) analogue as fluorescent base surrogates. 23,24 In contrast to previous work, we explored PNAbased TO/JO probes. PNA has higher affinity for complementary nucleic acids than DNA and, therefore, permits the use of shorter hybridization probes that provide high sequence spec-ifcity required for the discrimination of single base mismatches. 25 In addition, PNA is intrinsically stable to nuclease andunlike DNAdoes not require special modifcations for usage in live cell RNA imaging. 26,27 In the unbound state, TO/JO-PNA FIT-probe 1 should have low fluorescence when TO and JO interact via formation of aggregates (Fig. 1A). Owing to the absence of target the two probes 1 and 2 cannot co-align and, therefore, emission in the FRET acceptor channel will remain low as well. Spectral crosstalk between donor emission and FRET emission will be low because the donor is not emissive in absence of target. The presence of the edited RNA target T ed will trigger the adjacent hybridization of both probes and excitation of the donor will induce fluorescence emission in the acceptor channel (Fig. 1B). We were in favour of a FRET acceptor dye that emits at wavelengths > 650 nm, where emission of the TO/JO donor ensemble is very low and selected the NIR664 dye which based on previous reports should still allow sufficient overlap between TO/JO emission and NIR664 absorption. 36 The third state, i.e. the presence of the RNA in the unedited, single mismatched state, will be marked by increases of emission in the donor channel when fluorogenic donor probe 1 hybridizes while the mismatched acceptor probe 2 remains unbound (Fig. 1C). Given our previous work on DNA probes, we expected that the JO dye in PNA probe 1 confers a high brightness. 24 The JO dye is a bright emitter of fluorescence which may become even brighter when the TO dye serves as a light collector that increases the extinction coefficient andowing to the small 15 nm shift in emission maximaefficiently transfers excitation energy to the JO dye. This fosters donor emission in absence on the unedited target (Fig. 1C) or FRET emission on the edited target (Fig. 1B). ## Design of dual labelled FRET donor As a frst step towards a hybridization probe serving as a bright, albeit responsive FRET donor, we analysed a series of TO/JO-PNA FIT probes that recognise mRNA coding for a human Glycine Receptor (GlyR) (Fig. 2A). Probes, in which the TO and the JO dyes were separated by less than 5 nucleotides afforded a rather modest enhancement of fluorescence brightness upon hybridization (Fig. 2F). The 3-fold increase in emission was inferior to the fluorescence response provided by the TO-only probe TO 4 (Br ds /Br ss ¼ 4, see also Fig. 2B). However, remarkable 10-or 30-fold enhancements of emission brightness were observed when the JO dye was linked in 5 nt (TO 4 JO 9 ) or 6 nt (TO 4 JO 10 ) distance, respectively, from the TO dye. Control measurements with the JO-only probes JO 9 (Fig. S2 †) and JO 10 (Fig. 2C) suggested that the enhanced responsiveness (Br ds /Br ss z 30) was caused by dye-dye communication. Indeed, the increase of the fluorescence responsiveness observed for TO/JO probes TO 4 JO 9 and TO 4 JO 10 correlated with striking changes of the absorption spectra (see Fig. 2D). In the single stranded form, both TO-only (TO 4 , Fig. 2B) and JO-only (JO 10 , Fig. 2C) PNA FIT probes have absorption maxima between 515 and 530 nm, and always above 500 nm. In stark contrast, the absorption spectra of the TO/JO probes TO 4 JO 9 and TO 4 JO 10 (Fig. 2D) showed blue-shifted bands at wavelengths smaller than 500 nm. The TO-JO interaction leads to strong quenching of fluorescence. The single stranded TO/JO-PNA FIT probes TO 4 JO 9 and TO 4 JO 10 emit with 25% and 6%, respectively of the brightness expected for the sum of the TO-only and JO-only probes. The hypsochromic shift of absorption and the considerable quenching of fluorescence are indicative for the formation of H-aggregates. The pronounced quenching of fluorescence observed when TO was accompanied by an appropriately positioned JO base surrogate prompted us to consider means of fostering the TO-JO contact. We reckoned that an abasic site adjacent to one of the fluorescent base surrogates ("TO base" or "JO base") could provide space needed to accommodate a TO-JO complex. To test this hypothesis, we introduced an N-(2-aminoethyl) glycine (aeg) building block (indicated as X in the sequence strings) lacking a nucleobase adjacent to the "JO-base" and repeated the TO-JO-distance screen. Of note, four of the seven tested TO/JO/X-probes afforded hybridization-induced brightness enhancement > 10. We observed a remarkable 38-fold intensifcation of brightness by using TO 4 JO 9 X 10 (Fig. 2E). For probes that lacked the aeg unit a 5-6 nt TO-JO distance was required to provide $10-fold brightness enhancement. With the aeg unit the spectrum of suitable TO-JO distances was extended to 5-8 nucleotides (and possibly beyond) required that the aminoethylglycine unit was included as JO next neighbour. Control experiments revealed that positioning of the abasic site as next neighbour of the "TO base" afforded rather modest fluorescence responsiveness which is not due to inefficient quenching in the single stranded form but rather caused by a low intensity of emission of the bound probe (Fig. 2F, TO 4 JO 10 X 5 ). The intact base stack in the vicinity of the "TO base" is probably required for efficient activation of fluorescence upon hybridization. Likewise, fluorescence activation remained low when the aeg unit was placed between the fluorescent base surrogates in 2-3 nucleotides distance suggesting that the best position for aeg placement is the immediate JO environment. Additional measurements suggested that the benefcial "aeg effect" prevails within another sequence context (Fig. S5 †). An important feature of the binary probe format (Fig. 1) is the brightness of emission from the donor dyes. We found that probe-bound TO/JO probes providing high hybridizationinduced fluorescence enhancement also afforded bright emission signals (TO 4 JO 9 : Br ¼ 39 mM 1 cm 1 ; TO 4 JO 10 : Br ¼ 60 mM 1 cm 1 ) that exceeded the intensity of the TO-only probe TO 4 (Fig. 2F). We noticed that the placement of the abasic site unit aeg reduced the brightness; for probes with Br ds /Br ss > 10 by an averaged 10%. However, the brightness of the four aegcontaining probes with Br ds /Br ss > 10 (TO 4 JO 9 X 10 , TO 4 JO 10 X 9 , TO 4 JO 11 X 12 , TO 4 JO 12 X 13 ) was still higher than the brightness of the corresponding TO-only probe TO 4 . ## Cell imaging Interactions of the probes with the lipophilic environments inside a cell may perturb dye-dye communication and, thereby, obstruct fluorescence signaling of hybridization with the RNA target. To explore the feasibility of RNA imaging with TO/JO-PNA FIT-probes, we analyzed a stable Flp-In TM 293T-Rex cell line expressing an RNA coding for the fluorescent protein mCherry that was tagged in the 3 0 -untranslated region with 45 repeats of a 28 nt long sequence (ESI, Chapter 7 †) serving as a target for the TO 7 JO 12 X 13 _F probe (Fig. 3A). This probe experienced a 12-fold enhancement of fluorescence upon hybridization (Fig. S5 †). Expression of the tagged mCherry RNA was under the control of a doxycycline-responsive promotor. Addition of doxycycline to the cultured cells induced its expression as demonstrated by the emergence of mCherry emission signals measured by fluorescence microscopy, which was absent in untreated cells (Fig. S6 †). Next we fxed and permeabilized doxycycline-induced cells before adding 100 nM TO 7 JO 12 X 13 _F probe at 37 C. 28 After 1 h, the buffer was replaced by Dulbecco's phosphate-buffered saline (DPBS). Without further washes, we monitored emission by the TOJOX probe using fluorescence microscopy (Fig. 3B). Immediately after doxycycline addition (0 min), we observed only weak intracellular fluorescence signal, whereas after 15 minutes the cells were stained intensively by TO 7 JO 12 X 13 _F suggesting the presence of intracellular RNA target. Next, we assessed imaging of live cells. For delivery of TO 7 -JO 12 X 13 _F, the cells were incubated for 10 min with a 600 nM solution of the FIT-PNA in medium that contained streptolysine O (SLO). 27,29 Much stronger fluorescence signals were observed for cells treated with doxycycline (Fig. 3D) than for untreated cells (Fig. 3C). A small number of non-induced cells (3 out of 43, Fig. 3C) shows strong fluorescence. The round shape suggests that these cells undergo apoptosis and we assume that the strong signals are due to increased uptake. A noteworthy observation is that not all of the induced cells show the same intensity of emission from the FIT probe. We assign this phenomenon to cell-to-cell variations of probe delivery. Furthermore, non-synchronized cell lines show cells in different states. This leads to variations of mRNA concentration. In spite of these considerations, the fluorescence microscopy data from fxed and live cells indicates that TO/JO-PNA FIT-probes do allow imaging of intracellular mRNA. ## 3-State specic analysis of an RNA edit via FRET Encouraged by the results of the in vivo measurements we designed a binary probe system for the homogenous detection of C / U editing of mRNA coding for the Glycine Receptor GlyRa2. The binary probe system was comprised of TO/JO-PNA FIT-probe TOJO_1 and NIR664-labeled PNA probe NIR_2a which were designed to target two adjacent segments of the GlyR mRNA. The FRET acceptor strand served the purpose to probe the editing site in RNA glyr_ed. In the absence of target RNA, fluorescence in both donor and acceptor channel was low (Fig. 4A, black curve). Addition of the matched, edited target glyR_ed was accompanied by a 5-fold enhancement of fluorescence emission in the FRET channel at 690 nm (Fig. 4A, red curve). As expected, emission in the FRET channel was weak when the unedited ¼ single base mismatched target was added (Fig. 4A, blue curve), but in this case, the binary system responded by showing an 9-fold increase of the donor emission at 543 nm. We extended the length of NIR664-labeled probe from 6 to 8 nucleobases. However, the longer PNA probes lost the ability to distinguish between edited and unedited RNA (Fig. S8 †). A closer inspection revealed a small, yet noticeable FRET signal despite the absence of RNA target (Fig. 4A, inset). We assumed that the faint FRET signal was caused by weak interactions between the two PNA probes. Recently, we have remedied the problem of PNA-PNA interactions in DNA-templated chemistry by using PNA containing positively charged side chains. 30 We, therefore, incorporated three guanidino-PNA (GPNA) 31 building blocks into NIR664-labeled probe NIR_2b. Gratifyingly, the binary probe system comprised of TO/JO_1 and NIR_2b showed negligible emission in the FRET channel unless matched target glyR_ed was added, in which case emission at 690 nm was intensifed by a factor of 14 (Fig. 4B, see also Table 1). Of note, the FRET signal remained on background level upon addition of the mismatched target glyR_uned indicative for a high sequence specifcity provided by the binary probe system. The apparent quantum yield of the TO/JO/NIR667 ensemble on edited RNA is $9.5%, which is lower than that of a single NIR664 dye ($25%). However, owing to the large extinction coefficient of the TO/JO donor system (150.000 M 1 cm 1 at 520 nm) the brightness of emission at 690 nm can still reach up to 14 mM 1 cm 1 . Both apparent Stokes shift and brightness are comparable if not superior to values reported for other large Stokes shift dyes (e.g. ATTO490LS: 150 nm Stokes shift, 12 mM 1 cm 1 brightness). Measurements of the acceptor/donor emission ratio enable a clear discrimination of the three possible states. Weak signals in the donor and FRET channel characterize the absence of target. The presence of edited target is marked by an 14-fold enhancement of the intensity in the FRET channel and a 6-fold increase of the acceptor/donor emission ratio (Table 1). Increases of the donor emission signal are the hallmark for the presence of the unedited RNA target leading to a 10-fold reduction of the acceptor/donor emission ratio. We explored whether the ability to distinguish the three possible states by means of FRET signaling is sustained in complex matrices such as cell lysate. Indeed, though background in absence of RNA target is higher in cell lysate, the intensity of FRET emission after addition of synthetic edited RNA target is unaffected suggesting that both probes and RNA template remained available for sequence selective hybridization. The increased background in both donor and acceptor emission channels reduced the fold change of acceptor/donor emission ratio. However, the 14-fold change of the FRET ratio determined when edited RNA was replaced by unedited RNA should be sufficient to detect/ visualize the RNA editing event by fluorescence microscopybased cell imaging. ## Cellular imaging of an RNA C / U edit The binary probe system was used to image an mRNA edit in HEK293T cells transiently expressing bicistronic constructs containing mRNA coding for the GlyR a2 protein either in the unedited (GlyR a2 192P ) or the edited (GlyR a2 192L ) form (Fig. 5A). We equipped the 3 0 UTR of the glyra2 mRNA with 9 MS2 tag repeats, to enable its localisation via an independent method. The genetic information for the required eGFP-tagged MS2 binding protein was added upstream of the GlyR coding sequence and separated by a self-processing A2 peptide. 32 The FIT probes were introduced into live transfected HEK293T cells by applying the SLO delivery method. Afterwards, the cells were fxed. The eGFP emission signal served as a control that iden-tifes transfected cells and localizes the target mRNA (Fig. 5B and C). This signal cannot discriminate between edited and unedited states. Furthermore, it has been reported that the MS2 Fig. 4 Fluorescence spectra of TOJO_1 and (A) NIR_2a or (B) NIR-2b in absence of RNA target (black curves) or in presence of matched RNA glyr_ed (red curves) or mismatched RNA glyr_uned (blue curves). Conditions: 0.5 mM probes, 0.17 mM RNA (when added), pH 7.2, 37 C, l ex ¼ 500 nm. system also detects 3 0 -UTR segment from degraded mRNA. 33,34 However, the signal in the FRET channel is indicative for binding of TO/JO-PNA FIT-probe TOJO_1 and NIR664-labeled PNA probe NIR_2a in the coding region, which suggests that the detected mRNA molecules are intact. Moreover, the fact that target protein coding regions were translated (Fig. S11 †) implies a sufficient stability of the mRNA target. We found that the hybridization probes stained cells transfected with the edited RNA (GlyR a2 192L ) more brightly than cells transfected with the unedited RNA (GlyRa2 192P ). For a quantitative assessment, we calibrated the intensity of signals from FIT probes with regard to the amount of mRNA target expressed by analysing the ratio between the signals in the FRET channel and those in the eGFP channel. The analysis (Fig. S10 †) showed that cells expressing the edited RNA (GlyR a2 192L ) afforded a nearly two-fold higher intensity of the FRET emission than cells expressing unedited RNA (GlyR a2 192P ). A comparison with the 6-fold change in FRET emission intensity with synthetic targets in lysate (Table 1) reveals that the ability of the probe system to discriminate between the two RNA single nucleotide variations was compromised in the cells. The decrease in sequence specifcity may be due to fxation after probe delivery. We would like to stress that the conditions used to optimize single nucleotide specifc signaling in cuvette-type experiments emulate cellular conditions only poorly. For example, high local concentrations of target and/or probe and differences in target folding will affect the T m differences between matched and single mismatched ternary probe-target complexes. Enrichment of non-responsive hybridization probes on target can provide sufficient contrast for imaging. We, therefore, performed imaging experiments in the NIR channels. Indeed, the NIR signals showed similar features as observed in the FRET channel, albeit at reduced contrast (Fig. S13 †). Though the FRET signal was brightest in cells expressing the edited target mRNA, the signal increase compared to nontransfected cells (i.e. cells lacking the eGFP signal) appears rather modest (Fig. S13 †). We speculate that the excess of unbound probes over target is too high. We estimate that transfected cells express 10 4 to 10 5 copies of GlyR a2 mRNA. Given the typical dimensions of HEK293 cells this translates to a mean concentration of target mRNA # 10 1 nM. The 600 nM probe concentration applied during SLO-mediated delivery is much higher. Though the unfavorably large probe/target ratio may be lower at specifc cell sites, we infer that typing of the editing state requires the eGFP signal to guide the analysis to the site of mRNA localization. ## Conclusions The fluorogenic probes previously used for detecting single nucleotide variations in RNA did not allow the distinction between absence of target and presence of the single mismatched target. 13,14,35 In contrast, our probe system enables to distinguish between three rather than two states. The presence of matched (edited) target is characterised by an increase of the acceptor/donor emission ratio, while decreases of this FRET ratio below background values mark the presence of the mismatched (unedited) RNA target. This feature should facilitate the analysis of intracellular RNA editing by fluorescence microscopy. In this investigation, we demonstrated that the TO/JO-PNA FIT-probes provide a high responsiveness and high brightness of emission at 540 nm that enables FRET to a NIR dye with negligible bleed through. We found that an abasic site adjacent to the JO dye nucleotide improves the responsiveness of the TO/JO-PNA FIT-probes by facilitating TO-JO interactions that have characteristics of H-aggregates. The successful fluorescence microscopic imaging of mRNA in fxed and live cells demonstrated the usefulness of the TO/JO-PNA FIT-probes. We showed that the combination of fluorescence signals from the MS2 mRNA imaging system and the FRET signal from the two hybridization probes provided an indication about the editing state of RNA expressed in HEK293T cells. However, we expect that the full potential of the method will be unleashed when the donor emission will be included in the analysis. In future studies, we will replace the interfering eGFP protein and analyse RNA editing by means of ratio imaging. ## Conflicts of interest There are no conflicts to declare.

chemsum

{"title": "A bright FIT-PNA hybridization probe for the hybridization state specific analysis of a C \u2192 U RNA edit <i>via</i> FRET in a binary system", "journal": "Royal Society of Chemistry (RSC)"}

[3_+_2]_cycloaddition_with_photogenerated_azomethine_ylides_in_β-cyclodextrin

## Abstract: Stability constants for the inclusion complexes of cyclohexylphthalimide 2 and adamantylphthalimide 3 with β-cyclodextrin (β-CD) were determined by 1 H NMR titration, K = 190 ± 50 M −1 , and K = 2600 ± 600 M −1 , respectively. Photochemical reactivity of the inclusion complexes 2@β-CD and 3@β-CD was investigated, and we found out that β-CD does not affect the decarboxylation efficiency, while it affects the subsequent photochemical H-abstraction, resulting in different product distribution upon irradiation in the presence of β-CD. The formation of ternary complexes with acrylonitrile (AN) and 2@β-CD or 3@β-CD was also essayed by 1 H NMR. Although the formation of such complexes was suggested, stability constants could not be determined. Irradiation of 2@β-CD in the presence of AN in aqueous solution where cycloadduct 7 was formed highly suggests that decarboxylation and [3 + 2] cycloaddition take place in the ternary complex, whereas such a reactivity from bulky adamantane 3 is less likely. This proof of principle that decarboxylation and cycloaddition can be performed in the β-CD cavity has a significant importance for the design of new supramolecular systems for the control of photoreactivity. ## Introduction Cycloadditions are highly useful reactions in organic synthesis providing complex cyclic structures from easily available precursors . Among different reactions, [3 + 2] cycloadditions showed applicability in the synthesis of heterocyclic 5-ring compounds , as well as in the green synthesis of a number of natural products . One of the useful synthons in [3 + 2] cycloadditions is azomethine ylide , also used in intramo-lecular reactions . Azomethine ylides can be formed by several photochemical or thermal catalytic methods , including photodecarboxylation of phthalimide derivatives of α-amino acids such as N-phthaloylglycine (1) . Phthalimide is a versatile chromophore that has been used in the synthesis of complex molecules and natural products since the pioneering work of Kanaoka et al. . Photochemical reactions of phthalimides include H-abstractions, cycloadditions and photoinduced electron transfer (PET) . We became interested in the application of photochemical H-abstraction reactions initiated by phthalimides in organic synthesis . Furthermore, H-abstractions were investigated in inclusion complexes, in the cavity of β-cyclodextrins (β-CD) . We found out that H-abstraction reactions were about ten times more efficient in the β-CD complexes than in the isotropic solution, and the macrocyclic host affected the stereochemistry of the reaction. Moreover, we studied photodecarboxylation reactions initiated by the phthalimide chromophore and applied them in cyclizations with memory of chirality and diastereoselective peptide cyclizations . Photodecarboxylations were also intensively investigated in a series of nonsteroidal anti-inflammatory drugs such as ketoprofen , due to photoallergic responses initiated by photodecarboxylation of these drugs . Stereoselectivity in photochemical reactions can be achieved by use of supramolecular chemistry . For example, stereoselectivity has been reported for photochemical reactions taking place in the inclusion complexes with CD or struc-turally modified CDs . Since β-CD is often used in pharmaceutical applications for solubilization of drugs or drug delivery , it would be interesting to investigate its effects to the photodecarboxylation reaction. Therefore, we investigated photochemical reactivity of phthalimide derivatives 1-3 (Figure 1) in solution without β-CD and in the β-CD inclusion complexes. Phthalimides 1-3 yield azomethine ylides 1AMY-3AMY that are anticipated to react with acrylonitrile (AN) in [3 + 2] cycloadditions, which should be affected by β-CD. ## Results and Discussion Phthalimide derivatives 1-3 were prepared according to procedures published in precedent literature . The synthesis involves condensation of phthalic anhydride with unprotected amino acid. The synthetic procedures and characterization of compounds are reported in Supporting Information File 1), which may be ascribed to a small size of 1 that cannot fit well in the large cavity of β-CD and form a stable complex. Molecules 2 and 3 are larger, and anticipated to form more stable inclusion complexes with β-CD. Therefore, we per- . Nonlinear regression analysis of the chemical shifts to the β-CD concentration did not provide a good quality of the fit to the model involving 1:1 complex formation (Figure S5 in Supporting Information File 1). However, the approximated association constant for 3@β-CD, K = 2600 ± 600 M −1 , is similar to the known association constants for different adamantane derivatives with β-CD (K = 10 3 -10 5 M −1 ) , in agreement with the anticipated good fit of the adamantane moiety in 3 to the β-CD cavity. After demonstrating the formation of inclusion complexes 2@β-CD and 3@β-CD, we investigated the possibility for the Scheme 2: Complexation of 2 with β-CD, and formation of a ternary complex AN@2@β-CD. Scheme 3: Photochemistry of 2 in the presence of AN, with or without β-CD. formation of ternary complexes with AN. Therefore, we titrated solutions of 2 or 3 with AN. The solutions of 2 or 3 contained a high concentration of β-CD to assure that the phthalimide derivative was in the inclusion complex 2@β-CD or 3@β-CD, respectively. The addition of AN to the CD 3 CN/D 2 O (3:7 v/v) solution of 2@β-CD induced changes in the spectra, opposite to those observed upon formation of 2@β-CD (compare Figures S1 and S6 in Supporting Information File 1). The spectral changes are in accordance with the formation of a ternary complex AN@2@β-CD (Scheme 2). However, they may also indicate that excess of AN added to the solution competitively binds to β-CD forming a complex AN@β-CD and inducing dissociation of the 2@β-CD. If we assume a model for the complex formation in the stoichiometry 1:1:1, the nonlinear regression analysis of the chemical shift changes depending on the AN concentration and provided a poor fit with the estimated K 2 value of 0-6 M −1 (Figure S7 in Supporting Information File 1). We investigated also the possibility for the formation of the ternary complex AN@3@β-CD. Therefore, we titrated the solution of 3@β-CD in CD 3 CN/D 2 O (3:7 v/v) with AN, whereupon spectral changes were observed (Figure S8 in Supporting Information File 1). The signal of the adamantane H-atom at the adamantane position 6 experienced a downfield shift, whereas the phthalimide signals experienced an upfield shifts. Although the changes were small, we tried to process them using nonlinear regression analysis and model for the complex formation with 1:1:1 stoichiometry. The fit was of poor quality, but it provided an estimation of the constant with the value of K 2 = 0-7 M −1 (Figure S9 in Supporting Information File 1). The NMR titrations did not provide a clear evidence that ternary complexes were formed. However, formation of ternary complexes should affect the photochemical reactivity of 2 and 3 and cycloadditions of the corresponding azomethine ylides with AN. Therefore, we performed irradiation of solutions containing 2 and AN, 3 and AN, or the corresponding complexes 2@β-CD and 3@β-CD with AN. Scheme 3 and Scheme 4 show products formed in the photochemical reactions, whereas ratio of photoproducts obtained is given in Table 2. Addition of β-CD did not affect the decarboxylation reaction, since the same conversion to photoproducts was observed in the presence and absence of β-CD. However, for the adamantane derivative in the presence of β-CD, the secondary photochemical H-abstraction became more efficient, resulting in a different product distribution. More efficient H-abstraction reaction in the presence of β-CD have been reported . β-CD affected the cycloaddition reaction of photogenerated AMY with AN. Upon irradiation of 2 with AN in the presence of β-CD, the cycloadduct was formed in ≈5% yield, even though the irradiation was conducted in aqueous solution. The finding suggests that formation of a ternary complex AN@2@β-CD is possible and that photolysis of 2 in such a complex yields 2AMY, which is then readily intercepted with AN in the same complex. Note that cycloadduct 11 was also detected (≈4%), upon irradiation of 3 with AN in the presence of β-CD, suggesting that the photodecarboxylation reaction and subsequent [3 + 2] cycloaddition take place in the ternary complex AN@3@β-CD. However, cycloadduct 11 was formed in a higher yield (≈20%) when 3 was photolyzed with AN in the aqueous solution without β-CD. Although a reason for the different effect of β-CD is not clear, it may be due to a lower stability of the ternary complex AN@3@β-CD, compared to AN@2@β-CD. Namely, the adamantane is a bulky moiety that occupies most of the space in the inclusion complex 3@β-CD, making formation of the ternary complex less likely. Furthermore, if AN@3@β-CD was formed, photogenerated 3AMY@β-CD may not be in the right orientation for the cycloaddition to take place, leading predominantly to the reaction of 3AMY with H 2 O. ## Conclusion Herein we have demonstrated a proof of principle that β-CD can be used as a molecular container in which two molecules can be complexed, a phthalimide derivative and acrylonitrile, forming a ternary complex. Irradiation of such a complex leads to decarboxylation and formation of the reactive intermediate, azomethine ylide, within the supramolecular host. The subsequent [3 + 2] cycloaddition within the inclusion complex gives heterocyclic cycloadducts, even though it is conducted in aqueous solvent in which ylides have short lifetimes. The reaction needs to be optimized for different substrates with the right choice of host size. However, the proof of principle provides a new idea for the development of supramolecular systems for the tuning of photochemical reactivity. ## Experimental General ## Photochemistry of 2 and 3 under different conditions A solution of 2 (30 mg, 0.11 mmol) in CH 3 CN (27.5 mL) was prepared and transferred to 7 quartz cuvettes (3.9 mL into each). Then, CH 3 CN (3.9 mL) or H 2 O (3.9 mL) was added to each of the cuvettes, followed by the addition of β-CD (200 mg, 0.176 mmol), acrylonitrile (AN, 0.4 mL, 6.1 mmol) or nothing (see Table 2). Solutions were purged with N 2 for 15 min, sealed and irradiated at the same time in a Luzchem reactor at 300 nm (8 lamps) for 30 min. After the irradiation, solutions were extracted with EtOAc (3 × 3 mL), the extracts were dried over MgSO 4 , filtered and the solvent was removed on a rotary evaporator. The crude reaction mixtures were filtered through a plug of silica gel by use of CH 2 Cl 2 /EtOAc as eluent and were analyzed by 1 H NMR and HPLC-MS (Table 2). Alternatively, a solution of 3 (44 mg, 0.135 mmol) in CH 3 CN (50 mL) was prepared and transferred to 7 quartz cuvettes (7.0 mL into each). Then, CH 3 CN (7.0 mL) or H 2 O (7.0 mL) was added to each of the cuvettes, followed by the addition of β-CD (240 mg, 0.21 mmol), AN (0.7 mL, 10.7 mmol), K 2 CO 3 (1.3 mg, 0.009 mmol), or nothing (see Table 2). Solutions were purged with N 2 for 15 min, sealed and irradiated at the same time in a Luzchem reactor at 300 nm (8 lamps) for 35 min. After the above described work up, the composition of the irradiated solutions was analyzed by 1 H NMR and HPLC-MS (Table 2). Preparative irradiation of 2 with AN and with β-CD Phthalimide 2 (110 mg, 0.403 mmol) was dissolved in CH 3 CN (50 mL) and this solution was added slowly to the solution of β-CD (4.0 g, 3.52 mmol) in H 2 O (250 mL). The solution was sonicated for 15 min and then AN (10 mL, 152.6 mmol) was added. After sonicating for additional 15 min, the solution was transferred to fifteen quartz test tubes (each containing 20 mL), purged with N 2 for 20 min and sealed. The solutions were irradiated for 1 h in a Luzchem reactor using 8 lamps with the output at 300 nm. When the irradiation was completed, the irradiated solutions were combined and extracted with pentane ( ## NMR titrations with β-CD A solution of 2 (c = 7.21 mM), or 3 (c = 2.83 mM) in CD 3 CN/ D 2 O (3:7 v/v, 1.0 or 0.5 mL, respectively) in NMR tube was titrated with a solution of β-CD (c = 20.5 mM). After each addition of β-CD, an 1 H NMR spectrum was recorded. The changes of chemical shifts depending on the β-CD concentration were processed by nonlinear regression analysis using WinEQNMR software . The titration was performed at 25 °C. ## NMR titrations with AN A solution of 2@β-CD (prepared by mixing 2 in the concentration of 2.71 mM with β-CD in the concentration of 8.20 mM), or 3@β-CD (prepared by mixing 3 in the concentration of 1.54 mM with β-CD in the concentration of 20.5 mM) in CD 3 CN/D 2 O (3:7 v/v, 1.0 mL) was titrated with acrylonitrile (AN). After each addition of AN, an 1 H NMR spectrum was recorded. The changes of chemical shifts depending on the AN concentration were processed by nonlinear regression analysis using WinEQNMR software . The titration was performed at 25 °C.

chemsum

{"title": "[3 + 2] Cycloaddition with photogenerated azomethine ylides in \u03b2-cyclodextrin", "journal": "Beilstein"}

anomalies_of_a_topologically_ordered_surface

## Abstract: Bulk insulators with strong spin orbit coupling exhibit metallic surface states possessing topological order protected by the time reversal symmetry. However, experiments show vulnerability of topological states to aging and impurities. Different studies show contrasting behavior of the Dirac states along with plethora of anomalies, which has become an outstanding problem in material science. Here, we probe the electronic structure of Bi 2 Se 3 employing high resolution photoemission spectroscopy and discover the dependence of the behavior of Dirac particles on surface terminations. The Dirac cone apex appears at different binding energies and exhibits contrasting shift on Bi and Se terminated surfaces with complex time dependence emerging from subtle adsorbed oxygensurface atom interactions. These results uncover the surface states behavior of real systems and the dichotomy of topological and normal surface states important for device fabrication as well as realization of novel physics such as Majorana Fermions, magnetic monopole, etc. Topological insulators are like ordinary insulators in the bulk with gapless surface states protected by time reversal symmetry 1,2 . These materials have drawn much attention in the recent times followed by the proposals of the realization of exotic physics involving Majorana Fermions 3 , magnetic monopoles 4 etc. In addition to such fundamental interests, the predicted special properties of these states make them useful for the technological applications ranging from spintronics to quantum computations. Although the topological states are protected by time reversal symmetry 5 , numerous experiments show instability of the topological states with aging. Plethora of contrasting scenarios, anomaly on absorption of foreign elements, etc. are observed in the experimental studies . It is evident that the real materials are complex and may not be commensurate to the theoretical predictions that makes this issue an outstanding problem in various branches of science and technology. In order to elucidate these puzzles in real materials, we studied the electronic structure of a typical topological insulator, Bi 2 Se 3 at different experimental conditions such as the behavior of differently terminated surface, evolution of the electronic structure on aging, etc. employing high resolution photoemission spectroscopy. Bi 2 Se 3 forms in a layered structure (see Fig. 1(a)) with the quintuple layers of Se-Bi-Se-Bi-Se stacked together by Van der Waals force 8 . The surface electronic structure exhibits topological order with the apex of the Dirac cone, called Dirac Point (DP) at finite binding energies due to finite charge carrier doping arising from impurities, imperfections, etc. These states often show instability with time and complex time evolutions , which has been attributed to different phenomena such as relaxation of Van der Waals bond 12 , the surface band bending 5 , dangling surface states 13 , etc. There exists contrasting arguments indicating the necessity of unusually large change in the bond length for the relaxation of Van der Waals bonds 14 . Evidently, the real materials exhibit significant deviations from theoretical wisdom albeit the electronic states being topologically protected. Here, we discover that the behavior of the topological states is dependent on surface terminations. The anomalies in the behavior of the Dirac particles actually depends on subtle interactions of the adsorbates with the surface atoms. In Fig. 1, we show the signature of Dirac cone representing the topological surface states in the angle resolved photoemission spectroscopic (ARPES) data. The high symmetry points and the Brillouin zone defined in the reciprocal space of Bi 2 Se 3 are shown in Fig. 1(b). In addition to the metallic Dirac states, several bulk bands cross the Fermi level, ∈ F indicating metallicity of the bulk electronic structure 1 , 2 ]. Curiously, the Dirac point appears at a significantly high binding energy of about 0.45 eV. To ascertain the reproducibility of these data, the sample was cleaved several times and measured in identical conditions. We discover results of two categories. (i) In one case, DP appears at around 0.3 eV binding energy as shown in Fig. 2, consistent with the earlier results 1,2, . We denote this case as 'Clv1' . (ii) In the other case denoted as 'Clv2' , the DP appears around 0.45 eV binding energy. Time evolution of the 'Clv2' spectra at 20 K is shown in Figs. 1(c)-1(e) and corresponding energy distribution curves (EDCs) in 1(f) -1(h). Ironically, the DP shifts towards ∈ F with the increase in time delay from cleaving 7 suggesting an effective hole doping with time and/or passivation of the electron doped bulk with aging. DP in the 'Clv1' spectra shown in Fig. 2 appears at around 0.3 eV for freshly cleaved surface and gradually shifts away from ∈ F with the increase in time delay suggesting an electron doping with time. Evidently, the contrasting scenario for Clv1 and Clv2 surface is curious. DP appears to stabilize at a long time delay in both cases. The experiments at 200 K exhibit similar trend in energy shift with a slightly different saturation value. All these results indicate that the cleaved surfaces are qualitatively different in the two cases with ∈ F pinned at different energies and anomalous shift of the Dirac point with time. Some element specific study is necessary to reveal the surface chemistry of these materials. We employed x-ray photoemission spectroscopy (XPS) to probe the surface chemistry 18 . The normal emission Se 3d spectrum from Clv2 surface exhibits two peak structures for each spin-orbit split peaks as denoted by ' A' (53.1 eV) and 'B' (53.4 eV) in Fig. 3(a) for the 3d 5 , λ = escape depth and θ = emission angle with surface normal). The enhancement of B with the increase in surface sensitivity suggests its surface character, thereby, assigning the feature A to the bulk Se photoemission. Thus, the cleaving of Bi 2 Se 3 leads to different surface terminations exposing Se in Clv1 case and Bi in Clv2 case. We note here that top post removal method was necessary to prepare well ordered clean sample surface exhibiting significantly strong binding between the quintuple layers. The second important observation is shown in Fig. 3(b), where the 60° angled emission spectra from Clv1 surface exhibit a decrease of B with aging and subsequent increase of two weak features at higher binding energies (marked 'SeO x ' in the figure) indicating emergence of a new kind of Se at the cost of some of the surface Se species. The changes in the normal emission spectra from Clv2 surface, however, are not distinct indicating weak influence of aging on the subsurface Se species. The features deriving the Se 3d spectra were simulated by a set of asymmetric peaks as shown in Fig. 3(c). The shaded peaks represent the bulk Se photoemission and the other features correspond to the surface Se. The peak position of the emerging Se components with time and their intensity ratio indicate their origin to Se 3d signals from SeO x species 21 . Bi 5d spectra shown in Fig. 3(d), however, exhibit quite similar lineshape at various experimental conditions employed. Experiments on freshly cleaved surface do not show signature of impurity features. The oxygen 1s signal appears to emerge with aging and gradually grows with the increase in delay time. A representative case is shown in Fig. 4(a) for Clv1 surface after normalizing by the number of scans. The Clv1 spectrum after 6 hours delay exhibits three distinct features denoted by A, B and C in Fig. 4(b). The Clv2 spectrum at about 6 hours delay exhibits relatively more intense A and weaker C with no trace of B suggesting different characteristics of adsorbed oxygens on different surfaces. In both cases, the feature A grows quickly relative to the other features as shown in Fig. 4(c). Now, the question is, if this surface modification influences Dirac states although they are protected by topological order. In Fig. 4(d), we show the evolution of the Dirac point with aging and observe that the DP in Clv1 spectra appears at around 0.3 eV, on freshly cleaved surface. With increase in time delay, it gradually shifts towards higher binding energies with time and stabilizes around 0.4 eV. The same set of experiments at 200 K exhibit DP around 0.3 eV along with a weaker energy shift saturating around 0.38 eV that opens up new possibilities in understanding the behavior of these exotic states vis a vis existing wisdoms 17,22,23 . Ironically, the Clv2 spectra exhibit a reverse scenario with DP appearing at much higher binding energy of 0.45 eV and shifting in the opposite direction implying an effective hole doping case. The time evolution of DP can be expressed as, Here, ∈ 0 = DP at long delay time, t. t k1 and t k2 are the time constants. α and β are positive and are related to the electron and hole doping, respectively. The data points at different conditions can be captured remarkably well with the above equation. Now, at t 0 = , the binding energy at DP is 0 ε α β ( − + ). Therefore, the shift of DP can be expressed as values of ∈ 0 , α and β are 0.41, 0.13 & 0.02 for 'Clv1' at 20 K; 0.38, 0.12 & 0.02 for 'Clv1' at 200 K, and 0.35, 0.16 & 0.25 for 'Clv2' at 20 K. The parameters are quite similar in all the cases except a large β for the 'Clv2' case. t k1 and t k2 are 6 hrs and 16 hours at 20 K in both the cleaved cases. t k1 becomes 4 hrs at 200 K leaving t k2 unchanged. The growth of oxygen can be expressed as exp t t k − (− / ) with t k = 16 hrs for the features B & C, and 6 hrs for A at 20 K indicating a possible link between the DP and oxygen growth. Since the time delay primarily modifies the surface, the chemical potential shift must be due to the change in electron count in the surface electronic structure 24 . Three types of oxygen species are found to grow on the surface. The feature A grows quickly and have dominant contribution on Clv2 (Bi terminated) surface. The feature B is absent in Clv2 spectra and have large contribution in Clv1 spectra (Se-terminated surface) suggesting its bonding with surface Se leading to electron doping, while the feature A corresponds to oxygen bonded to Bi leading to hole doping. The charge densities on the surface are calculated using full potential density functional theory and shown in Fig. 5. The Bi terminated surface exhibit highly extended electron density spreading over a larger spatial distance away from the surface compared to the Se-terminated case. The Fermi surface corresponding to the Bi-terminated surface is significantly larger than that corresponding to Se-terminated case. Thus, DP is expected to appear at higher binding energy in Bi-terminated case compared to the Se-terminated case as found experimentally. Se and O belong to the same group of the Periodic table with O being the topmost element with higher electronegativity. Therefore, O on Se surface forms SeO x complex (signature of SeO x appears in the Se 3d spectra). Se-O bonding will attract electron cloud from the neighborhood reducing the electron density in the Bi-Se neighborhoods as manifested in the left panel of Fig. 5 by increase in spatial charge density contours around oxygen sites relative to the pristine case. Since the conduction band consists of p electrons, this would lead to an effective electron doping in the conduction band. On the other hand, oxygen on Bi-terminated surface would form BiO x complexes leading to more charge localization in the vicinity of oxygens (see right most panel of Fig. 5) reducing the Fermi surface volume; an effective hole doping scenario. In summary, we studied the surface electronic structure of a topological insulator, Bi 2 Se 3 employing high resolution photoemission spectroscopy. We observe the sensitivity of the Dirac states on surface termination. The surface states and the impurities appear to play a complex role leading to complex Fermi surface reconstructions emerging as a shift of the Dirac cone. These materials have been drawing much attention due to their potential technological applications in addition to the fundamental issues of realizing magnetic monopoles, the observation of Majorana fermions, etc. The results presented here reveal the complex microscopic details of the surface states necessary in the realization of such ambitious projects in real materials. ## Method High quality single crystals of Bi 2 Se 3 were prepared by Bridgeman method and characterized using x-ray Laue diffraction. Angle resolved photoemission (ARPES) measurements were carried out employing Gammadata Scienta R4000 WAL electron analyzer, monochromatic He I and Al Kα sources with photon energies 21.2 eV and 1486.6 eV, respectively. An open cycle helium cryostat was used to cool down the same. The energy and angle resolution were set to 10 meV and 0.1° for the angle resolved photoemission measurements with monochromatic He I light source to have adequate signal to noise ratio without compromising the resolution necessary for this study. Although the experiment setup could be operated at 300 meV resolution at 1486.6 eV photon energy, x-ray photoemission spectroscopic measurements were carried out at 380 meV resolution, which is better than the lifetime broadening of the core level features and enables collection of the spectra with good signal to noise ratio at a relatively short time. The pressure of the experiment chamber during the measurements with the photon sources on was better that 5 × 10 −10 torr. The sample was cleaved several times in situ at the experimental temperature using a top post glued on top of the sample. The well ordered sample surface has been verified by bright sharp low energy electron diffraction spots. X-ray photoemission spectra from freshly cleaved sample was found clean with no oxygen or carbon related signals. The electronic band structure of a slab of Bi 2 Se 3 was calculated employing state of the art full potential linearized augmented plane wave method using Wien2k software 25 . The bulk electronic structure exhibits a gap consistent with its insulating behavior 26,27 and the results from slab calculations show signature of Dirac cone in the presence of spin-orbit coupling.

chemsum

{"title": "Anomalies of a topologically ordered surface", "journal": "Scientific Reports - Nature"}

composite_photocatalysts_containing_bivo4_for_degradation_of_cationic_dyes

## Abstract: The creation of composite structures is a commonly employed approach towards enhanced photocatalytic performance, with one of the key rationales for doing this being to separate photoexcited charges, affording them longer lifetimes in which to react with adsorbed species. Here we examine three composite photocatalysts using either WO 3 , TiO 2 or CeO 2 with BiVO 4 for the degradation of model dyes Methylene Blue and Rhodamine B. Each of these materials (WO 3 , TiO 2 or CeO 2 ) has a different band edge energy offset with respect to BiVO 4 , allowing for a systematic comparison of these different arrangements. It is seen that while these offsets can afford beneficial charge transfer (CT) processes, they can also result in the deactivation of certain reactions. We also observed the importance of localized dye concentrations, resulting from a strong affinity between it and the surface, in attaining high overall photocatalytic performance, a factor not often acknowledged. It is hoped in the future that these observations will assist in the judicious selection of semiconductors for use as composite photocatalysts.Bismuth vanadate (BiVO 4 ) has attracted much attention as a highly responsive, visible-light driven, photocatalyst due to its comparatively narrow band gap (E G ) energy of 2.4 eV, (as compared to TiO 2 , which remains a benchmark, with an E G of 3.0-3.2 eV). BiVO 4 has been widely used in the photocatalytic degradation of organic compounds in waste water as well as for O 2 evolution under sunlight irradiation [1][2][3][4] .One of limitations of the photocatalytic efficiency in BiVO 4 is the recombination of photogenerated electrons and holes, with the free carrier life time reported to be about 40 ns by Abdi et al. 5 . In order to enhance these lifetimes, the creation of electronic barriers, facilitating spatial separation of the photogenerated electrons and holes, has been embraced by researchers as they can retard this recombination, providing more opportunities for free holes and electrons to participate in reduction and/or oxidation reactions, such as for the degradation of organic materials 4,6 . Doping semiconductors, either with metal or non-metals, has also been demonstrated to be an effective method for enhancing photocatalytic performance 4,6,7 . The composite approach relies on exploiting band-edge offsets, directing the electrons and holes into different materials, thereby providing spatial separation.Recently, many publications have coupled BiVO 4 with other metal oxides such as Bi 2 O 3 8, 9 , V 2 O 5 10, 11 , TiO 2 12, 13 , WO 3 11, 14-16 , CdS 17 , CuCr 2 O 4 18 and CuWO 4 19, for water purification and water spitting applications. The results show that the semiconductor composite photocatalysts were more active than individual catalysts for all photocatalytic degradation of organic pollutants -notwithstanding the fact that it is less likely that composites providing lower photocatalytic performances would be published. While most of these reports on composite catalysts acknowledge the role of band edge offsets, detailed investigations of reaction mechanisms are often not undertaken, nor has the nature of these energy offsets been thoroughly scrutinized 8,10,13,18,20 .Here we present a systematic study, pairing BiVO 4 with three different metal oxide semiconductors to create different offsets scenarios. CeO 2 /BiVO 4 , TiO 2 /BiVO 4 and WO 3 /BiVO 4 provide these different valence band (VB) and conduction band (CB) edge offsets, with the CB and VB potential edges of BiVO 4 being more positive than those of CeO 2 , in between those of TiO 2 , and more negative than those of WO 3 ; as shown in Fig. 1 6,7,[20][21][22] . These materials are selected based on literature values, as it is expected that this will result in different charge transfer (CT) processes being able to take place. In each case BiVO 4 has the narrowest bandgap and hence the broadest absorption, so the total amount of light harvested is consistent for each of the composites. As the above band edge potentials are obtained from different sources, and measured under different conditions, optical band gap and Mott-Shotkey measurements are conducted in order to confirm the predicted offsets. A number of mechanisms for photocatalytic degradation of dyes by semiconductors are summarised in Fig. 2, including (1) direct photolysis of dyes (2) dye photosensitization (i.e. charge injection from a photoexcited dye leaving unstable dye anion/cation) and (3) photocatalytic degradation of dyes attack by radical generated by photoexcitation of semiconductor (SC) . Typically direct photolysis is very slow, impractical as a means to 3i) and (3ii) involve direct oxidation/reduction of the dye by the photoexcited catalyst, while (3ii-3vi) proceed via radical intermediates. For each class of catalytic mechanisms, a schematic (yellow dotted line) represents how they are expected to vary with composition, with further explanation provided in text. remove organic contaminants from waste water, and indeed is the reason why researchers look towards using photocatalysts to enhance degradation rates 24 . Figure 2a shows schematically how mechanism (1) operates. It is expected that mechanism (1) should be independent of the catalyst used, or the percentage of each catalysts used in composite systems, provided there is no light harvesting competition (Methylene Blue, MB, absorbs most strongly at 664 nm, well beyond the absorption onset of BiVO 4 ). There are two variants of mechanism (2) which may occur, depending on the redox potentials of the dye compared to the band edges of the catalyst. Mechanism (2i) is photoanodic sensitization, with the excited dye injecting an electron into the semiconductor leading to an oxidized radial cation (Dye +• ), while (2ii) is photocathodic sensitization, generating a reduced radical dye (Dye −• ). Furthermore, these injected charges may lead to generation of other radical species. The chemical driving force between the dye redox potentials and those of the semiconductor will largely determine the rate of mechanism (2), and as such, with everything else being equal, should be proportional to the geometric mean of the components in composites. Degradation mechanisms (3i-vi), shown in Fig. 2b, rely on photoexcitation of the semiconductor. Generated h + and e − pairs (in the VB and CB of the semiconductor respectively) then react either directly (3i and 3ii) or generate radical intermediate species, which can degrade organic species such as the aforementioned dyes (3iii-3vi). There are a number of reactions which may take place, once again depending on the energies of the semiconductor band edges and the environment in which it exists 22, : Thus, the degradation will greatly depend on the adsorption of the dye, O 2 , OH − , H + and/or H 2 O on the surface of semiconductor. Photogenerated e − and h + can of course also recombine (either radiatively or non-radiatively) within the material. It is therefore beneficial to increase the lifetimes of these species to increase the probability that they will participate in one of the above reactions (3i-vi). Mechanisms in this third family may display a synergistic response, where composites can lead to higher reaction rates than either of the two component materials. In addition to examining rates of photocatalytic degradation for the different heterostructures, in this study, we also examine the mechanisms. This is done in two ways; firstly, through reactive species quenching studies, to differentiate between mechanisms involving OH • , O 2 •− , h + , e − and secondly with action spectra, taken to deconvolute the responses of the constituent materials. Additionally, a range of physical characterization techniques were employed, such as X-Ray Diffraction (XRD), X-ray photoelectron spectroscopy (XPS), Mott-Schottky experiments, and Scanning Electron Microscopy (SEM). Rhodamine B (RhB) and Methylene Blue (MB) are commonly used as model dyes for photodegradation experiments, being broadly representative of organic compounds in their class 22,24 , while being strongly light absorbing (making the remaining concentration easy to monitor by UV-Vis spectroscopy). In addition, since they absorb visible light, mechanisms (1) -direct photolysis 25 , and (2) -sensitization may occur in these systems, in addition to (3). Photoexcited MB can undergo a one electron reduction by other MB molecules to produce Leuco-methylene blue (MB/MB •− = −0.23 V vs. NHE) 23 . ## Results WO 3 /BiVO 4 , TiO 2 /BiVO 4 and CeO 2 /BiVO 4 composite powders, along with the respective pure materials, were synthesized by wet chemical methods with different mole ratios (1:4, 2:3, 1:1, 3:2 and 4:1), and used to produce photocatalyst films on glass by a doctor blading method 29 . In a preliminary study, the photocatalytic activities these investigated by degradation of MB and RhB under simulated solar irradiation. The concentration of remaining of dye shows a first order degradation relationship with respect to time, which can be fitted to the Langmuir-Hinshelwood (LH) kinetic model 30 (shown in Fig. S1). It was found that 1:4, 1:1 and 2:3 were the optimal ratios of WO 3 :BiVO 4 , TiO 2 :BiVO 4 and CeO 2 :BiVO 4 respectively for both MB and RhB under visible light irradiation, (Fig. S2). It should be noted that in each case, the composites outperformed the individual component materials, as predicted for type (3) mechanisms. These optimal ratios were chosen for further investigations, and are referred to simply as WO 3 /BiVO 4 , TiO 2 /BiVO 4 and CeO 2 /BiVO 4 , respectively from this point on. The crystal structures of the composites were characterized by XRD, as shown in Fig. 3, with each XRD pattern showing only characteristic diffraction peaks of monoclinic BiVO 4 and either monoclinic WO 3 , tetragonal TiO 2 (anatase) or cubic fluorite CeO 2 respectively, as expected. XRD patterns of the individual component materials can also be found in Fig. S3. Furthermore, XPS analysis was carried out to study the surface chemical composition of the above composites, and take a more detailed look at the interactions between BiVO 4 and the other metal oxide in each of the composites, as shown in Figs S4-S6. Due to proximity, electronic interactions were observed for these composites when examined by XPS . It was therefore concluded, from XRD and XPS that these systems were intimately formed composites, but did not contain any significant doping or any a mixed oxide phases. The morphologies of WO 3 /BiVO 4 , TiO 2 /BiVO 4 and CeO 2 /BiVO 4 composite powders and films were investigated using SEM (Fig. 4), with films shown to be highly porous and approximately the same thickness, ~2.8 μm, by cross-sectional SEM (Fig. 4c,f,i), in line with profilometry values mentioned previously. In each case, the two component materials appear to be well distributed among each other, as it particularly evident in the Energy Dispersive X-Ray Spectroscopy (EDS) maps. EDS data shows Bi and V tracking with one another, as expected, while other regions are rich in either W, Ti or Ce respectively, corresponding to WO 3 , TiO 2 or CeO 2 particles in the relevant composites. The scale of these differences suggests again that there are well-formed composites, with each containing image (Fig. 4a,d,g) two distinct materials in close proximity to one another. Further analysis of EDS (Figs S7-S9) is included in the supplemental information, along with TEM images (Fig. S10). ) . As mentioned, IPA, BQ and EDTA were introduced into photocatalysis experiment as scavengers of HO • , O 2 •− and h + , respectively. The photocatalytic degradation rate constant (k app ) for the composites and pure materials (both for MB (Fig. 5a) and RhB (Fig. 5b)), with the above quenchers (1 mM) were investigated under simulated solar irradiation. The addition of BQ almost completely quenched the dye degradation of both the pure BiVO 4 and CeO 2 indicating that O 2 •− is the major active species in these systems. For the photodegradation of dyes with the TiO 2 and WO 3 , a large suppression was noted upon the addition of IPA. As such, HO • was shown to be the main oxidation specie in pure TiO 2 and WO 3 photocatalysis systems. The photocatalytic systems with added EDTA showed similar degradation rates compared to the system without scavengers, suggesting a limited role played by holes in the valence band. This may, in part, due to these reactions proceeding via reactions with OH − , which is presumed to be in short supply at pH 5. Additionally, these studies suggest multiple mechanisms are at play in the degradation of MB and RhB dyes, with further studies being done to confirm this. Two major active species' in the photodegradation process of MB and RhB dyes in the presence of TiO 2 /BiVO 4 composite were seen, O 2 •− and HO • , while the main species of CeO 2 /BiVO 4 and BiVO 4 /WO 3 composite systems were O 2 •− and HO • , respectively. The generation of these radical species is explained in the context of their band edge energies, with the conduction band edge energies confirmed by Mott-Schottky measurements (Fig. S11) and valence band edges inferred based on optical bandgaps (see Fig. 6, Fig. S12 and Table S2). These experiments also highlight the possibility of a photocatalyst system being poisoned, a challenge which must be addressed in any real-world application of this technology. Figure 6 shows the normalised rate constants for photodegradation of MB dye over the different photocatalysts under monochromatic light (using band pass filters) along with their absorption spectra. MB was chosen for this study as its absorption has little overlap with that of the photocatalysts in question, as opposed to RhB. As expected, in the absence of any catalyst (Fig. 6a) MB degraded most rapidly under 650 ± 20 nm light, close to its maximum absorption (664 nm) as a result of direct photolysis (mechanism 1, from Fig. 2a) 22,41,42 . A similar response to these wavelengths were seen with all catalysts as well. As was mentioned previously, mechanism (1) degradation should be largely independent of the catalyst used, while mechanism (2, specifically 2i in this case) will be influenced by the chemical potential of free electrons in the CB. It is therefore concluded that that mechanism (2) plays a minor role, if at all, in the total photodegradation process here. The absorption onsets of the WO 3 , TiO 2 and CeO 2 were found to be 500, 390 nm and 435 nm respectively, while the absorption onset of all composites were dictated by that of BiVO 4 , at about 525 nm (shown in Fig. 6b-d). It is seen that the main mechanism of MB degradation in each case is due to the photogeneration of reactive radical species. As mentioned above, composite formation is shown to have synergistic effect on degradation rates, suggesting effective interactions between the various metal oxide species. For each composite, the responses to light of higher than band gap energies were enhanced as compared to the individual components, however this was more dramatic for some cases than others. As mentioned previously, BiVO 4 was selected as the common material as this results in equivalent total light harvesting for the composites. This hypothesis was validated looking at the absorption onsets of the composites, which were all indeed close to that of pure BiVO 4 . This also provides us with the ability to deconvolute the responses of the coupled materials. At 500 nm BiVO 4 absorbs, whereas CeO 2 or TiO 2 do not (this is within the range of the absorption onset for WO 3 , however direct comparison the response to this 500 nm light, as compared to 400 nm or 450 nm, where WO 3 more strongly absorbs light, is illustrative). Interestingly, while all composites showed a synergistic response at this wavelength, in the case of TiO 2 and WO 3 , shorter wavelengths (400 and 450 nm) result in less marked differences as compared to pure BiVO 4 . On the other hand, CeO 2 displays similar synergistic behaviour at all wavelengths from 400 to 500 nm. Based on the combined results of quenching studies, monochromatic illumination Mott-Schottky experiments (Fig. S11) and optical band gap measurements (Fig. S12), the primary degradation mechanisms are proposed for the individual materials and WO 3 /BiVO 4 , TiO 2 /BiVO 4 , and CeO 2 /BiVO 4 composites and are shown in Fig. 7. As mentioned, the main photodegradation mechanism seen with BiVO 4 involves the transfer of a photoexcited e − from the CB of BiVO 4 , along with a proton, to adsorbed an O 2 on the semiconductor surface, generating HO 2 ## • . Meanwhile, quenching studies showed OH • to the only major reaction species for the pure WO 3 photocatalyst system, which be produced by oxidation at VB of WO 3 . As OH − exists only at low concentrations in slightly acid conditions, the photocatalytic activity exhibited here was lower than other materials. For the WO 3 /BiVO 4 composite, (Fig. 7e) the CB and VB band edges of BiVO 4 are more negative than the CB and VB of the other component. Both BiVO 4 and WO 3 can be activated by visible light irradiation and generated e − and h + pairs, although at 500 ± 20 nm, the relative portion of light harvested by the WO 3 should be considerably less compared to BiVO 4 than under 400 or 450 nm illumination. According to the band edge analysis, electrons transferred to the CB of WO 3 cannot reduce O 2 to form radical species, as they can from the BiVO 4 CB. This is reflected in quenching studies, with the deactivation of the O 2 −• species mechanism in BiVO 4 as it is paired with WO 3 , representing an unintended consequence of composite formation. On the other hand, the photo-generated h + at the VB of WO 3 can transfer to the VB of BiVO 4 , which can oxidize H 2 O and HO − to HO • . Adding to the complex combination of effects in the composite, the dye is seen to have a very strong affinity towards WO 3 . This is further explored by inspection of adsorption/desorption equilibrium of the dye, as measured in the dark (MB in Fig. 8 and RhB in Fig. S13), which shows the affinity of the dye towards WO 3 to be quite strong as compared to BiVO 4 . In spite of an overall increased photocatalytic performance of the composite, as compared to pure WO 3 or pure BiVO 4 , it can be seen that the mechanism of this enhancement is not straight forward. A similar trend was seen for RhB (see Fig. S13). The role of surface area and isoelectric points (IEP) was explored (Fig. 8 above, as well as Fig. S14). This shows dye loading to be (1) highly dependent upon the materials' isoelectric point and (2) for composites, loading is close to a stoichiometric mixture of the loading of the individual component materials. The effect of the affinity for the dye for the various metal oxides is pivotal towards our understanding of the photocatalytic behaviour of composites, as described in the following paragraphs. Under illumination TiO 2 , can generate HO • by reaction of adsorbed H 2 O and h + in its VB. Additionally, adsorbed O 2 can produce O 2 •− , by reaction with free e − in the CB. Figure 7f shows the mechanisms observed in BiVO 4 /TiO 2 composites, where the CB and VB of BiVO 4 are both located between the CB and VB of TiO 2 . This represents an interesting case, particularly under 450 nm and 500 nm illumination, where only BiVO 4 is photoexcited and there are no obvious pathways whereby CT to TiO 2 can occur. Photocatalytic performance is however substantially higher than for pure BiVO 4 . This appears to again stem from the affinity of the dye for TiO 2 , whereby a high concentration of these dyes is maintained in proximity to the BiVO 4 catalyst, which is the primary producer of the active radical species. This affinity is seen in the dark adsorption/desorption equilibrium (Fig. 8) and explained by the by the zeta potential of TiO 2 (Fig. S14). Thus, the major process of this system should be driven by the visible light havering of BiVO 4 which can produce HO 2 • and/or O 2 •− at its CB. As CeO 2 has a higher CB edge potential than BiVO 4 , free electrons have a substantial free energy to generate O 2 •− from O 2 , which was seen to be the main photocatalytic mechanism at play here (Fig. 7d). The comparatively low photocatalytic activity of CeO 2 (versus BiVO 4 ) is in part attributed to low visible light absorption. 400 nm monochromatic illumination shows however that this is not the only factor, along with dark adsorption/desorption measurements (Fig. 8) which reveal the low affinity of the dye for CeO 2 , despite its high surface area. As seen in Fig. 7g, the CB and VB potential edges of BiVO 4 are more positive than those of CeO 2 . The tendency of CT is that photoexcited e − in the CB of CeO 2 can be transferred to BiVO 4 , where they can react with absorbed O 2 to create O 2 •− radicals. From the quenching studies, HO • was also observed to play role in this system, suggesting that these HO • were probably generated from intermediate species and/or the generated O 2 •− and HO 2 • radicals 22,24 . Therefore, the main mechanism is production of HO • at the VB of the BiVO 4 in the composite that leads to showing the lowest photocatalytic activity as compared to the CeO 2 /BiVO 4 and TiO 2 /BiVO 4 composites. ## Discussion A series of composites were constructed, containing WO 3 , TiO 2 or CeO 2 with BiVO 4 . This last component ensured each system had similar overall light harvesting. XRD, XPS, SEM and TEM analyses confirm the formation of the composite without significant doping. Furthermore, highly porous films were formed, with a good dispersion of the different materials among each other. Superior photocatalytic performance for degradation of MB and RhB under simulated solar light irradiation was observed for all composites, relative to their component materials. Monochromatic illumination and active species quenching studies were conducted and revealed a great deal of information about the different routes by which radical species were produced, leading to the degradation of these dyes. Quenching studies, indicated that O 2 •− is a main reactive specie generated by CeO 2 /BiVO 4 systems, while BiVO 4 /WO 3 primarily degraded dyes via HO • , and both O 2 •− and HO • were active for the TiO 2 /BiVO 4 composite. Typically enhanced free carrier lifetimes resulting from a spatial separation of charges in different (neighbouring) materials is cited as a reason for enhanced performance, however here we see a number of other factors also had a bearing on the behaviour of composite systems. Specifically, it was seen that the affinity of the dye to the partner semiconductor plays a major role as it provides a high local concentration, where generated radicals have a higher probability of being able to react as compared to in the general solution. This was demonstrated clearly in the case the BiVO 4 -TiO 2 composite, under 450 and 500 nm illumination of where no CT mechanisms are favourable, yet the degradation rate constant is roughly twice as large as for pure BiVO 4 . The WO 3 -BiVO 4 composite was also seen to be an interesting case, with the band edge configuration deactivating one radical species generation pathway. Again, the overall performance increased, which again be attributable to high dye affinity for WO 3 . This study highlights the complexities of composite design for photocatalysis, and the need to consider a range of factors, such as zeta potential and surface area, when designing new systems. It also demonstrated that band edge offsets can lead to deactivation of radical species generating pathways, as well as the more desirable increase in free charge carrier lifetime. We hope that it can serve as a basis for further design of new composite materials for water purification and other applications. ## Methods Preparation of photocatalyst powder. The synthesis of all photocatalysts were carried out by solution phase synthetic methods. For the synthesis of BiVO 4 powder, bismuth nitrate and ammonium vanadate in dilute nitric acid solution were used as starting precursors for precipitation at room temperature. CeO 2 /BiVO 4 and TiO 2 /BiVO 4 composites were prepared by two-step methods 20,43 . Firstly, BiVO 4 powder was synthesised by a precipitation method. Subsequently the BiVO 4 powder was added into either cerium nitrate or titanium isopropoxide solutions to prepare composites by precipitation or sol-gel methods, respectively. BiVO 4 /WO 3 composites were also synthesised by a two-step precipitation method, where WO 3 powder was first synthesised using tungsten nitrate as the starting material. The WO 3 powder was added into a solution with of Bi and V precursors. The choice of synthetic order relates to pH regulation for each step and stability of the partner material under these conditions. Detailed descriptions of the synthesis of each of the CeO 2 /BiVO 4 , TiO 2 /BiVO 4 and BiVO 4 /WO 3 composites in the supporting information. Fabrication of photocatalyst films. All films for photocatalytic testing and electrochemical measurements were prepared by a doctor blading technique 29 . Briefly, 0.5 g of the photocatalyst powder was used, along with 75 µL Triton X-100, 25 µL acetic acid and 4 mL ethanol. The slurry was ground with a mortar and pestle for 10 min, during which ethanol was added in small aliquots to break up larger agglomerates. The paste was then sonicated for a further 30 min. Films were obtained by blading the slurry (~100 µL) on either glass slides (film size = 20 mm × 40 mm) for photocatalytic experiment or (~12.5 µL) on FTO glass working electrodes (film size = 10 mm × 10 mm) for Mott-Schottky experiment. This was carried out on a heated (50 °C) surface, using Scotch Tape (3 M, 50 μm) as a spacer. These films were then annealed in air at 450 °C for 2 h, to produce mechanically stable films. The thicknesses of the resulting films were 2.8 ± 0.1 µm, as measured with a stylus profilometer (VeecoDektak 150). Physical characterization of photocatalyst materials. X-Ray Diffractometry (JDX-3530, JEOL, Japan) was completed using Cu K α radiation (λ = 1.546 nm). The detection range was 10° and 70° with the step size of 0.2° (2θ/s) in continuous scanning mode. Film morphologies were investigated by a Scanning Electron Microscope (SEM, JSM-7500FA, JEOL, Japan) with the accelerating voltage and emission current as 5.0 kV and 10 μA, respectively. The chemical composition and electron structure of the bare CeO 2 , BiVO 4 and BiVO 4 /CeO 2 composite were measured by X-ray photoelectron spectroscopy (XPS, PHOIBOS 100 hemispherical energy analyser from SPECS) using Al, K α radiation (1486.6 eV) in fixed analyser transmission mode. The binding energies were calibrated with reference to C 1 s line at 284.6 eV for hydrocarbon contamination Films were optically characterized using a Shimadzu 3600 UV-Vis-NIR spectrophotometer, with integrating sphere attachment. Transmission and reflection measurements were made in order to determine the absorption. Images were taken on an FEI Helios G3 CX using the STEM3 + detector at 30 kV. EDS data was acquires using an Oxford instruments X-max N 150 mm 2 SDD detector with acquisition and processing performed with the AZtec application suite. Samples were prepared by sonicating a dilute dispersion of the material in ethanol and dropping on a carbon coated copper grid. Photocatalytic testing. Photocatalytic activities of the synthesized photocatalyst powder and films were studied by the degradation of either Rhodamime B (~25 µM, RhB) or methylene blue (~50 µM, MB) solutions under simulated solar illumination (AM1.5 G, 1 sun equivalent, 100 mW cm −2 ). For monochromatic experiments, ±20 nm band pass filters were applied to the above light source. Photocatalyst films were placed put in a reactor with the dye solution. Prior to irradiation, the films were left overnight in the dye solution under continuous magnetic stirring in order to attain an adsorption/desorption equilibrium. At irradiation time intervals of 30 min, the dye solution was collected and measured using an UV-vis spectrophotometer (Agilent 8453 photodiode array). After that, the collected solution was put back into the reactor, and the photodegradation reaction resumed (<3 min intermission). An indirect chemical probe method was employed to investigate the mechanisms of dye degradation. Scavengers of various possible active species, including isopropanol (IPA, an HO • quencher), benzoquinone (BQ, O 2 •− quencher) and ethylenediamine tetra acetic acid (EDTA, a h + scavenger), were added at 1 mM during the photoreaction . Each experiment was repeated in triplicate, with the average value shown along with one standard deviation as an error bar.

chemsum

{"title": "Composite Photocatalysts Containing BiVO4 for Degradation of Cationic Dyes", "journal": "Scientific Reports - Nature"}

reprogramming_of_seed_metabolism_facilitates_pre-harvest_sprouting_resistance_of_wheat

## Abstract: Pre-harvest sprouting (PHS) is a worldwide problem for wheat production and transgene antisensethioredoxin-s (anti-trx-s) facilitates outstanding resistance. To understand the molecular details of PHS resistance, we analyzed the metabonomes of the transgenic and wild-type (control) wheat seeds at various stages using NMR and GC-FID/MS. 60 metabolites were dominant in these seeds including sugars, organic acids, amino acids, choline metabolites and fatty acids. At day-20 post-anthesis, only malate level in transgenic wheat differed significantly from that in controls whereas at day-30 postanthesis, levels of amino acids and sucrose were significantly different between these two groups. For mature seeds, most metabolites in glycolysis, TCA cycle, choline metabolism, biosynthesis of proteins, nucleotides and fatty acids had significantly lower levels in transgenic seeds than in controls. After 30-days post-harvest ripening, most metabolites in transgenic seeds had higher levels than in controls including amino acids, sugars, organic acids, fatty acids, choline metabolites and NAD + . These indicated that anti-trx-s lowered overall metabolic activities of mature seeds eliminating pre-harvest sprouting potential. Post-harvest ripening reactivated the metabolic activities of transgenic seeds to restore their germination vigor. These findings provided essential molecular phenomic information for PHS resistance of anti-trx-s and a credible strategy for future developing PHS resistant crops.Pre-harvest sprouting (PHS) is characterized by the premature germination of seeds occurring in the spikes and becomes a serious problem in many wheat-growing areas worldwide with prolonged pre-harvest rainfall and high humidity 1 . PHS leads to reduced yields, lower end-product quality of the grains and hence great economic losses. In China, for instance, about 25 million hectares of wheat are affected by PHS every year accounting for 83% of the whole wheat-growing area. Therefore, PHS resistant varieties are developed to solve the problem especially for the wheat-producing areas where long period of rainfall occurs frequently in the harvest season.Pre-harvest sprouting is a complex phenotype resulting from interactions between wheat genotypes 2 with biotic and/or abiotic environmental factors 3 . The untimely breakdown of seed dormancy is considered to be the major event for PHS to occur and therefore improvement of PHS resistance is often accompanied with prolonged seed dormancy to pass the harvest stage 4 .Thioredoxin h (trx-h) gene is important for wheat seed germination and thus PHS 5 . Trx-h initially found in wheat kernels in 1979 6 is now found widely present in the higher plants. This gene acts as an important regulator for seed germination by facilitating the reduction of intramolecular disulfide bonds in storage proteins of cereals, such as wheat and barley. During seed germination, trx-h also promotes the activation of α -amylase, pullulanase and proteases by weakening the inhibitive effect of inhibitor proteins on amylases and proteases 7 . Overexpressing trx-h gene in barley accelerated germination of the embryos and activated both α -amylase and starch pullulanase 8,9 . On the other hand, underexpressing trx h9 gene in wheat lowered the activities of Trx protein, α -amylase and pullulanase slowing seed germination 10 . It is particularly worth-noting that the transgenic wheat underexpressing trx h9 gene has also shown outstanding PHS resistance 10 . Trx-s is another member of the thioredoxin gene family initially cloned from Phalaris coerulescens 5 . In fact, trx-s and trx-h have more than 90% homology in their cDNA sequences and similar biological functions for their expression products 11 . By using pollen-tube pathway, antisense thioredoxin s (anti-trx-s) was transferred into wheat to have successfully developed transgenic wheat with PHS resistance as well 12 . It was reported that anti-trx-s inhibited the endogenous trx-h expression and lowered α -amylase activity between day-30 post anthesis and 10 days post-harvest ripening resulting in high PHS resistance in the transgenic wheat . It was also found that the introduction of anti-trx-s gene inactivated starch hydrolases and slowed hydrolysis of storage proteins in seeds imbibed for 3 to 4 days. Systems biology approaches offer excellent opportunity to understand pre-harvest sprouting in terms of protein expressions and metabolism in a more holistic manner. Proteomic analyses already showed that transferred anti-trx-s gene caused down-regulation of many proteins in wheat seed kernels involving protein biosynthesis/degradation, starch degradation, gene expression regulation, lipid and energy metabolisms 19 . Anti-trx-s also caused up-regulation of proteins in kernels involving α -amylase activity suppression and disulfide bond formation compared to wild-type 19 . Several proteins related to stress resistance (such as antioxidant and disease resistance) were further up-regulated in the transgenic wheat kernels 19 . Moreover, transgenic wheat showed differential gene expression in trx-h, serpin, heat shock protein 70 (hsp70), and WRKY transcription factor 6 (WRKY6) compared to wild-type 19 . These results also imply that anti-trx-s gene may induce comprehensive metabolic changes in the transgenic wheat seeds. However, it remains unknown what metabolic changes such transgene causes, at which seed development stages and how these transgenic effects on seed metabolic activities are related to PHS. Metabonomics ought to be a useful approach for understanding the dynamic metabolic changes since metabonomic analysis measures the metabolite composition (metabonome) of a given biological system and its dynamic responses to both endogenous and exogenous factors . Such approach has been proven to be powerful in disease diagnosis 23 , in understanding metabolic variation between different rice varieties 24 and metabolic responses to gene modifications 25 . Metabonomic analysis has increasingly become a powerful approach in understanding the effects of biotic and abiotic stressors on plant physiology and biochemistry . So far, however, there have been no reports about the effects of anti-trx-s on the wheat seed metabonome, to the best of our knowledge, though these effects are expected to be insightful for developing PHS resistant wheat. It is also conceivable that PHS and its resistance ought to be associated with the development dependence of wheat seed metabolic phenome since sprouting of wheat seeds generally go through four grain filling periods including milk stage, dough development stage, mature seeds and post-harvest ripening period 15 . In this study, we analyzed the seed metabonomic phenotypes (metabotypes) of transgenic wheat with anti-trx-s and wild-type at four different time-points of seed development (milk stage, dough development stage, mature seed and post-harvest ripeness period) using NMR spectroscopy in conjunction with multivariate statistical analysis. We also analyzed the developmental dependence of the fatty acid composition of these seeds using GC-FID/MS method. We further conducted integrative analysis on the metabonomic and proteomic differences between the PHS susceptible and resistant seeds. Our objectives are (1) to define the metabonomic changes induced by introduction of anti-trx-s and (2) to understand the molecular aspects of the PHS resistance acquired through introduction of such gene which will offer important information for further development of PHS-resistant wheat varieties. ## Results Taking into consideration of the developmental periods of wheat seeds 15 , in this study, we analyzed metabonomic features of wheat seeds harvested at about day-20 post anthesis (20-dpa, milk stage), day-30 post anthesis (30-dpa, dough development stage), day-40 post anthesis (40-dpa, mature seed) and 30 days post-harvest ripening (30-dpr), respectively. We also considered the metabonomic data in an integrative manner with our proteomic results. Our results showed that this anti-trx-s transgenic wheat seeds had lower trx-h expression and α -amylase activity but higher PHS resistance compared with the corresponding wild-type wheat seeds as reported previously . Our proteomic analysis results also showed that anti-trx-s gene introduction led to significant expressional changes of a number of metabolism-related proteins in wheat seeds 19 at 40-dpa. This was highlighted by up-regulation of protein disulfide isomerase (PDI), serpin, xylanase inhibitor protein I precursor (XIP-I) but down-regulation of glucose-6-phosphate isomeraseprecursor (GPI), aldolase (ALD), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphoenolpyruvate carboxylase (PEPC), glutamate dehydrogenase (GDH), malate dehydrogenase (MDH), pullulanase, aspartate aminotransferase (AST), adenylylsulfate kinase (APK), α -and β -amylase 19 . Seed metabolite composition for transgenic and wild-type wheat. 1 H NMR spectra of seed extracts (Fig. 1) showed rich but obvious different metabolite profiles for anti-trx-s transgenic and wild-type wheat seeds at 30-and 40-dpa, respectively. Both 1 H and 13 C signals were unambiguously assigned to individual metabolites (Table S1) based on the literature data , in-house databases and publicly available databases 33 with further confirmation using a series of 2D NMR spectral data. More than 40 metabolites were identified including 16 amino acids and their metabolites (Ala, Thr, Val, Ile, Leu, Glu, Gln, Asp, Asn, Phe, Tyr, Trp, GABA, His, Arg and 4-guanidinobutyrate), 10 organic acids (trans-aconitate, citrate, α -ketoglutarate, succinate, malate, fumarate, formate, acetate, lactate, isobutyrate), 2 plant secondary metabolites namely 4-hydroxy-3-methoxyphenylacetate (HMPA) and ferulate, 6 nucleotide metabolites (uridine, adenosine, guanosine, allantoin, NAD + and ATP), 5 choline metabolites (choline, ethanolamine, betaine, dimethylamine and trimethylamine), 4 sugars (sucrose, glucose, raffinose, D-ribose-5-phosphate) and ethanol (Fig. 1, Table S1 and Table 1). 15 fatty acids were also detected and quantified from these seeds (Table 2). Visual inspection showed that compared with wild-type, transgenic seeds had higher levels of fumarate and GABA at 30-dpa (Fig. 1a,b) but lower levels of Trp and Ala at 40-dpa (Fig. 1c,d). To obtain more detailed metabolic changes induced by anti-trx-s gene, multivariate data analyses were performed on the NMR data of these seeds. Anti-trx-s gene induced metabolic changes in transgenic wheat seeds. PCA (principal component analysis) scores plots showed clear development-associated metabonomic changes for both transgenic and wild-type wheat seeds reflected by a metabolic trajectory from 20-dpa, 30-dpa, 40-dpa and 30-dpr (Figure S1). Some obvious metabonomic differences were also observable between the transgenic and wild-type wheat seeds at 30-dpa, 40-dpa and 30-dpr (Figure S2B-D) although such difference appeared to be less prominent at 20-dpa (Figure S2A). To obtain the detailed information on significant metabolic alterations induced by anti-trx-s, pairwise OPLS-DA (orthogonal projection to latent structure-discriminant analysis) was conducted between the extracts of transgenic wheat and wild-type seeds harvested at 20-dpa, 30-dpa, 40-dpa and 30-dpr, respectively; significantly differential metabolites between these two groups were tabulated in Table 1. At 20-dpa, OPLS-DA model parameters indicated only some limited metabolic differences were present between transgenic wheat and wild-type seeds. Careful analysis of all variables individually 34 showed that at this development stage, only malate in the transgenic wheat seeds had significant level difference (p = 0.037) compared to the wild type (Table 1). At the other three time points of seed development (i.e., 30-, 40-dpa and 30-dpr), in contrast, OPLS-DA model parameters (Fig. 2, Table 1) all showed significant seed metabonomic differences between the transgenic and wild-type wheat. The coefficient-coded loadings plots indicated that introduction of anti-trx-s caused significant elevation of seed fumarate, Asp, GABA and Phe at 30-dpa together with level decreases for sucrose, Glu and Gln compared with wild-type wheat (Fig. 2a, Table 1). At 40-dpa, anti-trx-s transgenic wheat seeds had significantly lower levels for 2 sugars (sucrose and glucose), 9 amino acids (Ala, Thr, Val, Ile, Leu, Glu, Gln, Trp and Arg), a TCA intermediate (citrate) and a secondary metabolite (HMPA), 3 choline metabolites (ethanolamine, choline and betaine) and 6 nucleotide derivatives (uridine, adenosine, guanosine, allantoin, NAD + and ATP) than the wild type (Fig. 2b, Table 1). Furthermore, after 30 days post-harvest ripening (30-dpr), the transgenic wheat seeds had significantly lower level of guanosine but higher contents of sucrose, Val, citrate, HMPA, choline metabolites (ethanolamine, choline and betaine) and NAD + , being in contrast to the seed metabonomic features at 40-dpa. Moreover, introduction of anti-trx-s caused elevation of wheat seed raffinose, Asp and 4-guanidinobutyrate compared with the wild-type (Fig. 2c, Table 1). We further calculated the ratios of concentration changes for the transgene-altered metabolites (against the wild-type wheat) at four different time points. The results indicated that compared with the wild type, a clear metabolic switch for the transgenic wheat occurred between 40-dpa and 30-dpr. At 40-dpa, for instance, the levels of Trp, Ala, Thr, Gln and Phe in the transgenic wheat seeds showed about 54%, 35%, 28%, 26% and 26% level decrease, respectively, compared with the wild-type (Fig. 3). In contrast, after 30 days post-harvest ripening, the levels of most seed amino acids in transgenic wheat were similar to these in wild type with exceptions for Asp and Val whose levels were higher in transgenic wheat (Fig. 3). Furthermore, much more significant differences between sucrose and glucose levels in transgenic wheat and wild-type were observed at 40-dpa than at the other three time-points; their levels were more than 30% lower in transgenic wheat seeds than in wild type (Fig. 3). After 30 days post-harvest ripening, in contrast, the levels of sucrose and raffinose were about 15% and 34% higher in transgenic wheat seeds than in wild-type (Fig. 3). Moreover, at-40 dpa, citrate and malate levels were lower in transgenic wheat seeds than wild type whereas, at 30-dpr, the levels of these metabolites and fumarate were higher in transgenic wheat seeds (Fig. 3). Ethanolamine, choline and betaine levels also showed a decrease at 40-dpa but an elevation at 30-dpr (Fig. 3) in transgenic wheat seeds compared with wild type. Anti-trx-s gene induced fatty acid changes in transgenic wheat seeds. Composition of wheat seed fatty acids was dominated by palmitate (C16:0), stearate (C18:0) and their unsaturated forms. Linoleic acid (C18:2n6) was the most abundant wheat seed fatty acid (34.1-46.1 μ mol/g fresh weight) accounting for about 60% of total fatty acids (Table 2). Palmitic acid (C16:0, 11.7-15.0 μ mol/g) and oleic acid (C18:1n9, 6.2-8.9 μ mol/g) accounted for about 20% and 10% of total seed fatty acids, respectively. Linolenic acid (C18:3n3, 2.5-3.4 μ mol/g) and stearic acid (C18:0, 1.2-1.4 μ mol/g) were about 5% and 2% of total fatty acids, respectively. 1. Significantly differential seed metabolites between anti-trx-s transgenic (TG) and wild type wheat (WT) a . a The level of malate was lower in transgenic wheat than that in wild type seeds at 20-dpa (t-test, p = 0.037). b The coefficients were obtained from OPLS-DA results, and positive and negative signs indicate positive and negative correlation in the concentrations, respectively. Results showed outstanding dynamic changes in fatty acids induced by introduction of anti-trx-s with an obvious switch after 40-dpa. Compared with wild type seeds, the contents of total fatty acids (ToFAs), saturated fatty acids (SFAs), unsaturated fatty acids (UFAs), monounsaturated fatty acids (MUFAs), polyunsaturated fatty acids (PUFAs) and n6-type fatty acids were all decreased in transgenic wheat seeds at 40-dpa whereas the contrary was observed at 30-dpr (Table 2). Specifically, at 40 dpa, transgenic wheat seeds had significantly lower contents of C16:1n7 and PUFAs (C18:2n6, C18:3n3 and C20:4n6) than the wild-type (Table 2). At 30 dpr, in contrast, the concentration of C15:0, C16:0, C18:1n9, C20:1n9, C18:2n6 and C20:4n6 were significantly higher in transgenic wheat seeds than wild type (Table 2). Furthermore, at 40-dpa, the n6-to-n3 ratio for fatty acids was higher in transgenic wheat seeds than in wild type whereas the PUFA-to-MUFA ratio in transgenic wheat was lower than in wild type at 30-dpr (Table 2). ## Discussion Wheat seeds with anti-trx-s gene expression showed excellent resistance to pre-harvest sprouting (PHS) 35,36 . The PHS resistant wheat differed significantly from the PHS susceptible wild-type in the seed metabonomic phenotypes at several seed development stages (Fig. 2 and Table 2). Such differences were outstanding especially when seeds reached maturity (at 40-dpa) and post-harvest ripening for 30 days (30-dpr). It is known that the period from day-5 pre-mature to day-10 post-mature is critical for wheat PHS to occur 15 . Our results showed that whilst most of the seed metabolic activities were inactivated at harvest time (40-dpa), these activities were restored after post-harvest ripening for a month (Fig. 4). This indicates that appropriate regulation and control of the seed metabolic activity during seed development is critical for pre-harvest sprouting to occur and thus for development of PHS resistant wheat. This current study showed that introduction of anti-trx-s led to different metabonomic changes at different development stages in wheat seeds. In the early seed developmental stage (20-dpa or milk stage), only malate level in anti-trx-s transgenic wheat differed from that in wild type (Fig. 3). At this stage, the expression level of trx-h was only about 15.6% lower in the transgenic wheat than in wild type (Yumai-70) 15 . When anti-trx-s was transferred to another wild type cultivar Yangmai-5, no difference in α -amylase activity was observable between the transgenic and wild type wheat seeds at 20-dpa 37 . This implies that, at 20-dpa, the effects of introduction of S1. Figure 3. The ratios of concentration changes for metabolites in transgenic wheat seeds against that in wheat of wild type seeds. Metabolites with statistically significant variations are listed in Fig. 2 and Table 1. anti-trx-s have only limited impacts on seed metabolism simply because the major metabolic events at this stage are programmed towards biosynthesis of materials needed for seed development. The transgenic effects on seed malate level may indicate that Krebs cycle was slower to some extent in the transgenic wheat during this early grain fill period. When seeds reached dough development stage (at 30-dpa), much more metabolic differences were observable between the transgenic and wild type wheat seeds. GABA, Asp, Phe and fumarate were significantly up-regulated in the transgenic line whereas Glu, Gln and sucrose were significantly down-regulated (Fig. 2, Table 1). At this stage, about 23% decreases in the trx-h expression level were observed in the anti-trx-s transgenic wheat seeds compared with the control 15 . This indicated that biosynthesis of storage proteins and consumption of biogenic Glu was faster in transgenic wheat than in wild type offering nitrogen source for protein biosynthesis. Significant lower level of sucrose in transgenic seeds implied that anti-trx-s slowed the starch degradation to favor starch storage compared with wild type. At the dough development stage, therefore, the major metabolic events for both transgenic and wild type wheat seeds appeared to be biosynthesis of storage matters including starch and proteins although anti-trx-s expression seemed to accelerate such biosyntheses to some extent. When wheat seeds reached maturity at 40-dpa, germinating ratio of transgenic wheat seeds was significantly lower than wild type 36 . Significant lower levels for 26 seed metabolites in the transgenic wheat than in wild type (Fig. 2, Tables 1,2) indicated comprehensive reduction of metabolic activities in anti-trx-s transgenic wheat seeds. This is agreeable with 36.5% decreases in α -amylase activity in anti-trx-s transgenic wheat seeds compared with the wild type control Yangmai-5 at 40-dpa 37 . These results are also consistent with the proteomic results for this wheat with significant up-regulations of XIP-I and PDI but down-regulations of GAPDH, GPI, ALD and PEPC 19 . This is reasonable since XIP-I suppresses α -amylase activity in wheat seeds 38 whilst GAPDH, GPI and ALD are enzymes involved in glycolysis and thioredoxin can activate GAPDH 39 . Therefore, the markedly reduction of sucrose and glucose levels (Table 1, Fig. 2), XIP-I up-regulations and down-regulations of GAPDH, GPI and ALD 19 suggests that both the seed starch degradation and glycolysis were slowed down in transgenic wheat at this particular stage (Fig. 4). PEPC catalyzing conversion of oxaloacetate into TCA intermediates can be hydrolyzed by enzymes having disulfide bonds 40 . At 40-dpa, expression of anti-trx-s gene in the transgenic wheat lowered thioredoxin h protein expression hence down-regulation of PEPC 19 . This and significantly lower citrate levels in the transgenic wheat seeds than in controls (Table 1) also suggested that TCA cycle was slowed down at 40-dpa in the transgenic wheat (Fig. 4). Significant level decreases for many amino acids (Glu, Gln, Arg, Ala, Trp, Val, Leu, Ile and Thr) in transgenic wheat at 40-dpa suggest that expression of anti-trx-s gene affected protein metabolism. This and up-regulations of serpin and PDI 19 , two important proteins inhibiting breakdown of seed proteins, in transgenic wheat (Fig. 4) indicated that anti-trx-s expression slowed degradation of reserve proteins in seeds at this stage. Furthermore, down-regulations of GDH, AST and APK 19 in the transgenic wheat (relative to wild type) indicated that anti-trx-s expression also slowed protein biosynthesis since these were essential enzymes for sulfydryl-dependent protein biosynthesis 41,42 . Lower choline and ethanolamine levels in transgenic wheat seeds than in wild type are probably related to changes in membrane biosynthesis with choline and ethanolamine as the main fragments of membrane phospholipids in plants 43 . These two fragments can be produced through phospholipid hydrolysis catalyzed by phospholipase D (PLD). In fact, PLD antagonist did inhibit seed germination in Arabidopsis 44 . Our observation of choline and ethanolamine decreases resulting from introduction of anti-trx-s implies probable PLD inactivation in transgenic wheat to inhibit untimely seed germination (i.e., PHS) and/or down-regulation of glycerate-3-phosphate due to slowed glycolysis (Fig. 4). The decreases of adenosine, guanosine and uridine in transgenic wheat seeds were probably associated with the reduced overall metabolic activities of seeds including nucleic acid degradation and de novo biosynthesis purines and pyrimidines. The level reduction of allantoin further supports this notion with it as a catabolic metabolite of purines. Since wheat seed germination accompanies initiation of DNA and RNA biosynthesis 45 , it is reasonable that the decreased adenosine, guanosine and uridine levels in transgenic wheat seeds observed here are related to inhibition of potential seed germination. The change of HMPA associated with introduction of anti-trx-s indicates some roles of plant secondary metabolism in seed germination and such clearly warrants further investigations. Lower levels of main fatty acids in transgenic wheat seeds than in wild type (Table 2) are probably also related to slower overall metabolic activities in transgenic seeds although slowing glycolysis is one of the major reasons since de novo biosynthesis of fatty acids in plants is critically dependent on production of acetyl-CoA from glycolysis. After 30 days postharvest ripening (30-dpr), 17 metabolites in anti-trx-s transgenic wheat seeds showed significant increase compared with wild type (Fig. 2, Tables 1,2) including NAD + , a TCA intermediate (citrate), two sugars (glucose, raffinose), three amino acid metabolites (Val, Asp, 4-guanidinobutyrate), three choline metabolites (choline, ethanolamine, betaine), a secondary metabolite (HMPA) and six fatty acids (C15:0, C16:0, C18:1n9, C18:2n6, C20:1n9 and C20:4n6). This indicates that break of seed dormancy occurs and transgenic wheat seeds start having more metabolic activities than corresponding wild type. Interestingly, 10 metabolites with lower levels in transgenic wheat at 40-dpa had higher levels at 30-dpr than the wild type controls (Tables 1,2). Nevertheless, at 30-dpr, the α -amylase activity together with the ratio and rates of germination were all similar for anti-trx-s transgenic and wild type seeds 36,37 indicating both lines had similar starch degradation potentials required for germination. Compared with wild type, therefore, transgenic wheat seeds activate their metabolism to reduce the inhibiting effects of anti-trx-s gene so as to break dormancy and to initiate germination. Furthermore, 30 days postharvest ripening was sufficient for transgenic wheat seeds to recover the germination vigor by recovering the required metabolic activity of seeds. To sum up, PHS-resistance of transgene anti-trx-s resulted largely from reprogramming seeds metabolism especially at later seed development (dough development, seed maturing) and post-harvest ripening stages. The expression of antisense thioredoxin s gene suppressed the overall metabolic activities of seeds when seeds reach maturity (40-dpa) (Fig. 4) so as to prevent from seed germination (i.e., pre-harvest sprouting). This metabolic reprogramming is critically important at about 40-dpa to prepare the seed metabolic activities since wheat seeds are most susceptible PHS at this stage 15 . During 30-days post-harvest ripening, however, the transgenic seeds re-adjusted their metabolic activities by up-regulating a number of pathways to overcome the anti-trx-s effects. This re-adjustment of metabolic activities is also critically important for seeds to break dormancy and germinate. Although anti-trx-s gene expressed throughout wheat growing period, its effects on seed metabolism was less obvious at milk (20-dpa) probably because biosyntheses of proteins, starch and nucleic acids were the dominant events in seed metabolism at such stage. These results offered essential biochemistry information for pre-harvest sprouting of wheat and for PHS resistance of anti-trx-s. ## Methods Plant materials. Seeds from the transgenic wheat with antisense trx-s gene and wild-type (Yumai 18) were harvested, respectively, at day-20 post anthesis (20-dpa, milk stage), day-30 post anthesis (30-dpa, dough development stage) and day-40 post anthesis (40-dpa) when seeds reached complete maturity. Wheat seeds harvested at 20-dpa and 30-dpa were snap-frozen in liquid nitrogen and stored at − 80 °C until further analysis. Seeds at 40-dpa were divided into two groups with one group snap-frozen in liquid nitrogen followed with storage at − 80 °C (designated as 40-dpa in this study). The other group of samples were kept at room temperature for 30 more days to have post-harvest ripening (and designated as 30-dpr here) followed with snap-frozen in liquid nitrogen and storage at − 80 °C. Nine to eleven independent biological replicates were employed for 30-dpa, 40-dpa and 30-dpr whereas only six biological replicates were possible for samples harvested at 20-dpa. ## Seed metabolite extraction. Each sample was individually ground to fine power in liquid nitrogen with a mortar and a pestle followed with extraction with an optimized method for seeds 46 . In brief, above seed powder (150 ± 3 mg) was individually weighted into an Eppendorf tube followed with addition of 600 μ L aqueous methanol (66%, v/v) and 3 mm tungsten carbide bead (Qiagen, Germany). The mixture was subjected to homogenized with a tissuelyser (Qiagen, Germany) followed with 15 min intermittent sonication in an ice bath (1 min sonication and 1 min break, repeated for 15 times). After centrifugation for 10 min (16099 × g, 4 °C), the supernatant was collected. This extraction process was further repeated twice and these three supernatants obtained were combined. After removal of methanol under vacuum, samples were lyophilized. The freeze-dried extracts were re-dissolved into 600 μ L phosphate buffer (0.1 M, 50% D 2 O, pD 7.4) prepared from K 2 HPO 4 /NaH 2 PO 4 47 containing 0.002% TSP-d 4 . Following 10 min centrifugation (16099 × g, 4 °C), 550 μ L of such solution from each sample was then transferred into a 5 mm NMR tube for metabolite analysis. ## NMR Measurements. All 1 H NMR spectra were recorded at 298 K on a Bruker AVIII 600 spectrometer (600.13 MHz for 1 H) equipped with an inverse detection cryogenic probe (Bruker Biospin, Germany). A standard noesypr1d pulse sequence (RD-90 o -t 1 -90 o -t m -90 o -acquisition) was used to record one-dimensional 1 H NMR spectra with the 90 o pulse length of about 10 μs and t 1 of 3 μs. Water peak was saturated with a continuous wave irradiation during the recycle delay (RD, 2s) and mixing time (t m , 80 ms). A total of 64 transients were collected with 32 k data points over a spectral width of 12 kHz. For resonance assignment purposes, a set of 2D NMR spectra were recorded and processed as previously reported 48,49 for selected samples including 1 H-1 H COSY, 1 H-1 H TOCSY, 1 H-JRES, 1 H- 13 C HSQC and 1 H-13 C HMBC spectra. Spectral processing and multivariate date analysis. An exponential line-broadening factor of 1 Hz was applied to each free induction decay (FID) prior to Fourier transformation (FT), and the spectra were phaseand baseline-corrected followed by referenced to TSP at δ 0.00. NMR spectral region between 0.5 and 8.5 ppm was divided into segments with width of 0.003 ppm using AMIX (v3.9.2, Bruker Biospin) whilst both methanol region at δ 3.339-3.384 and water region at δ 4.610-5.220 were removed. The spectral areas of all buckets were normalized to the fresh weight of wheat seed power so that resultant data represented absolute metabolite concentration in the form of "peak intensity per weight unit of seeds". PCA (principal component analysis) and OPLS-DA (orthogonal projection to latent structure-discriminant analysis) 50 were carried out on the normalized NMR data using SIMCA-P+ software (v11.0, Umetrics, Sweden). In OPLS-DA models, one orthogonal and one predictive component were calculated using the unit-variance (UV) scaled NMR data as X-matrix and the class information as Y-matrix. The model qualities were described by the explained variances for X-matrix (R 2 X values) and the model predictability (Q 2 values) with further assessment with CV-ANOVA approach 51 where intergroup differences were considered as significant with p < 0.05. The results were displayed in the forms of scores plots showing group clustering and loadings plots indicating variables (metabolites) contributing to inter-group differences. In these plots, loadings were back-transformed 52 and the variables were color-coded according to the absolute values of the correlation coefficients (|r|) 53 using an in-house developed script. In such plots, variable (i.e., metabolites) with hot colors (e.g., red) have more significant contributions to the group classification than the cold colored ones (e.g., blue). In this study, the metabolites exhibiting statistically significant changes were obtained at the level of p < 0.05. The ratios of changes for metabolites at four time-points were also calculated against their levels in wild-type wheat seeds in the form of (C TG -C WT )/C WT , where C TG stands for the metabolite concentrations in transgenic wheat seeds whereas C WT stands for the metabolite concentrations in wild-type wheat seeds. GC-FID/MS analysis of fatty acids in the wheat seeds. Fatty acid composition in wheat seeds were measured in the methylated forms with a previously reported method 54,55 with some minor modifications. About 20 mg of seed powder of wheat harvested at 40-dpa and 30-dpr was used with addition of 500 μ L HPLC-grade CH 3 OH and one 3 mm tungsten carbide bead (Qiagen, Germany). Seeds at 20-and 30-dpa were not measured here due to limitation of samples availability. After vortex mixing, the mixture was homogenized with a tissuelyser (Qiagen, Germany) at 20 Hz for 90 s. Such treatment was repeated thrice with 3 min rest between each homogenization. 100 μ L of homogenized mixture from each sample in a Pyrex tube was added with 20 μ L internal standards in hexane (containing 1 mg/mL methyl heptadecanoate, 0.5 mg/mL methyl tricosanate and 2 mg/mL 3,5-di-tert-butyl-4-hydroxytoluene BHT) , then 1 mL mixture of methanol and hexane (4:l v/v). After cooling down the sample tubes above in a home-made liquid nitrogen bath for 10 min, 100 μ L of precooled acetyl chloride was added carefully. Tubes were screw-capped and kept at room temperature in the dark for 24 hours followed with cooling down in an ice-bath for 4 min. 2.5 mL of 6% K 2 CO 3 solution was then added (with gentle shaking) for neutralization. 200 μ L hexane was added to extract the methylated fatty acids thrice and the combined supernatants were evaporated to dryness. The extracts were re-dissolved in 100 μ L hexane for GC-FID/MS analysis. Methylated fatty acids were quantified on a Shimadzu 2010 Plus GC-MS spectrometer. A flame ionization detector (FID) was used for quantification and a mass spectrometer with an electron impact (EI) ion source was employed for identification purpose. A DB-225 capillary GC column (10 m × 0.1 mm × 0.1 μ m) was used with helium as carrier gas with 1 μ L sample injected. The injection port and detector temperatures were set to 230 °C. The oven temperature was increased from 55 °C to 205 °C at 25 °C per min, kept at 205 °C for 3 min and then increased to 225 °C at 10 °C per min. The temperature was then kept at 225 °C for further 3 min. All MS spectra over the m/z range of 45-450 were acquired with EI at 70 eV. Methylated fatty acids were identified by comparing retention times and mass spectra of 37 fatty acid standards. The results were expressed as micromole fatty acids per gram wheat seed fresh weight (FW). ## Proteomics analysis. Seed proteomics analysis for both transgenic and wild type wheat was done at 40-dpa as previously reported 19 .

chemsum

{"title": "Reprogramming of Seed Metabolism Facilitates Pre-harvest Sprouting Resistance of Wheat", "journal": "Scientific Reports - Nature"}